feat(desktop): add SSH server connections

refactor(ai): give named providers dedicated modules (#47726 )
feat(app): refine timeline detail settings (#47226 )
2026-09-07 09:26:26 +00:00 · 2026-09-07 15:30:39 +08:00 · 2026-09-06 23:56:39 -05:00 · 2026-09-07 04:54:28 +00:00 · 2026-09-07 14:17:34 +10:00 · 2026-09-06 23:42:25 -04:00
689 changed files with 31396 additions and 6715 deletions
@@ -91,7 +91,7 @@ jobs:

      - uses: ./.github/actions/setup-bun
        with:
-          bun-version: 1.4.0
+          bun-version: 1.4.2

      - name: Setup git committer
        id: committer
@@ -113,7 +113,7 @@ jobs:
        id: build
        run: ./packages/cli/script/build.ts ${{ (github.ref_name == 'beta' && '--sourcemaps') || '' }}
        env:
-          BUN_COMPILE_RELEASE: bun-v1.4.0
+          BUN_COMPILE_RELEASE: bun-v1.4.2
          OPENCODE_VERSION: ${{ needs.version.outputs.version }}
          OPENCODE_RELEASE: ${{ needs.version.outputs.release }}
          GH_REPO: ${{ needs.version.outputs.repo }}
@@ -183,7 +183,7 @@ const table = sqliteTable("session", {
 - Keep `SessionRunner`, model resolution, tool registry, permissions, and filesystem Location-scoped. Omitted `Location.workspaceID` means implicit-local placement; explicit workspace identity remains reserved for future placement semantics.
 - Preserve one explicit `llm.stream(request)` call per Physical Attempt and reload projected history before durable continuation. A logical Step may use generic pre-output retries, one full-context retry after continuation rejection, incomplete-stream continuation, or one overflow-compaction rebuild. Generic retries retain the logical step number and do not consume another agent-step allowance. Do not delegate orchestration to an in-memory tool loop.
 - Keep local Session drains process-local until clustering is implemented. `SessionRunCoordinator` joins explicit same-Session resumes, coalesces prompt wakeups, and allows different Sessions to run concurrently. A write-ahead execution claim marks a process-local busy period for restart recovery: terminal completion, failure, or user interruption releases it, while shutdown interruption and process death preserve it. Startup recovery resumes claimed top-level Sessions with durable per-execution attempt accounting. The claim is a recovery marker, not clustered ownership, fencing, or an exactly-once guarantee.
- Keep delivery vocabulary explicit. Prompts steer by default. Steers deliver in enqueue order at safe step boundaries, stopping before compaction or move control items. At an idle boundary, steers take priority; otherwise exactly one queued item delivers before the runner reevaluates continuation. Inbox items may be cancelled or changed between queue and steer before delivery. Promoting new user input resets the selected agent's step allowance; a batch of steers resets it once.
+- Keep delivery vocabulary explicit. Prompts steer by default. At safe step boundaries, steered compaction takes priority up to the first steered move control; other steers retain enqueue order. At an idle boundary, steers take priority; otherwise exactly one queued item delivers before the runner reevaluates continuation. Inbox items may be cancelled or changed between queue and steer before delivery. Promoting new user input resets the selected agent's step allowance; a batch of steers resets it once.
 - One step is one logical LLM call; its durable record covers only the model-visible span. Do not write "provider turn", and do not use bare "turn" for a single call: "turn" is reserved for the future assistant-turn unit containing all steps from prompt promotion until the session would go idle.
 - Keep event replay ownership separate from clustered Session execution ownership.
 - Keep the Instructions algebra and built-ins in `src/instructions`; keep instruction producers with their observed domains, and keep Session History selection plus `InstructionState` and `InstructionEntry` persistence Session-owned. `InstructionDiscovery` observes ambient global and upward-project instructions. The runner composes built-ins, discovery, guidance, and entries explicitly in `loadInstructions`; there is no instruction registry.
@@ -33,6 +33,7 @@
      "name": "@opencode-ai/ai",
      "version": "1.17.20",
      "dependencies": {
+        "@aws-sdk/credential-providers": "3.1057.0",
        "@opencode-ai/schema": "workspace:*",
        "@smithy/eventstream-codec": "4.2.14",
        "@smithy/util-utf8": "4.2.2",
@@ -338,7 +339,6 @@
      "version": "1.18.4",
      "dependencies": {
        "@ai-sdk/alibaba": "1.0.17",
-        "@ai-sdk/amazon-bedrock": "4.0.112",
        "@ai-sdk/anthropic": "3.0.82",
        "@ai-sdk/azure": "3.0.88",
        "@ai-sdk/cohere": "3.0.27",
@@ -350,7 +350,6 @@
        "@ai-sdk/provider": "3.0.8",
        "@ai-sdk/provider-utils": "4.0.23",
        "@ai-sdk/vercel": "2.0.39",
-        "@aws-sdk/credential-providers": "3.1057.0",
        "@ff-labs/fff-bun": "0.10.5",
        "@ff-labs/fff-node": "0.10.5",
        "@lydell/node-pty": "catalog:",
@@ -433,6 +432,7 @@
        "@types/node": "catalog:",
        "@typescript/native-preview": "catalog:",
        "app-builder-lib": "26.15.7",
+        "drizzle-kit": "catalog:",
        "drizzle-orm": "catalog:",
        "effect": "catalog:",
        "electron": "42.10.1",
@@ -695,6 +695,8 @@
        "effect": "catalog:",
      },
      "devDependencies": {
+        "@opencode-ai/client": "workspace:*",
+        "@opencode-ai/plugin": "workspace:*",
        "@tsconfig/bun": "catalog:",
        "@types/bun": "catalog:",
        "@typescript/native-preview": "catalog:",
@@ -1035,12 +1037,9 @@
    },
  },
  "trustedDependencies": [
-    "esbuild",
-    "tree-sitter-powershell",
-    "protobufjs",
    "electron",
-    "web-tree-sitter",
-    "tree-sitter-bash",
+    "esbuild",
+    "protobufjs",
  ],
  "patchedDependencies": {
    "@pierre/trees@1.0.0-beta.4": "patches/@pierre%2Ftrees@1.0.0-beta.4.patch",
@@ -1065,6 +1064,7 @@
    "@opentui/solid": "catalog:",
    "@types/bun": "catalog:",
    "@types/node": "catalog:",
+    "bun-types": "1.4.2",
    "effect": "catalog:",
    "solid-js": "catalog:",
  },
@@ -1102,7 +1102,7 @@
    "@tanstack/solid-virtual": "3.13.37",
    "@tsconfig/bun": "1.0.9",
    "@tsconfig/node22": "22.0.2",
-    "@types/bun": "1.3.13",
+    "@types/bun": "1.4.0",
    "@types/cross-spawn": "6.0.6",
    "@types/luxon": "3.7.1",
    "@types/node": "24.12.2",
@@ -1186,7 +1186,7 @@

    "@ai-sdk/mistral": ["@ai-sdk/mistral@3.0.51", "", { "dependencies": { "@ai-sdk/provider": "3.0.14", "@ai-sdk/provider-utils": "4.0.40" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-83eXY6p0lUFhSuMvNDmTKDuMciK5XDAWDlNh5c0L80tKjmtCFRItA1MZHp4IKe1r7eK8Rb5nN7qtxqMLUFRIRw=="],

-    "@ai-sdk/openai": ["@ai-sdk/openai@3.0.67", "", { "dependencies": { "@ai-sdk/provider": "3.0.10", "@ai-sdk/provider-utils": "4.0.27" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-oAiGC9eWG7IgtdsdS74bOCnAAHarAfTJhWN9x5INwnWPekL802AvF+0I5DvLzIF1MIRmNw4N8mPSL/GUVbX9Mw=="],
+    "@ai-sdk/openai": ["@ai-sdk/openai@3.0.84", "", { "dependencies": { "@ai-sdk/provider": "3.0.14", "@ai-sdk/provider-utils": "4.0.38" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-cmgbeJL0bbY0yTJH4/AdmP5E7MjWRL9G8UdhIi0JlV/So03o82ORJofW8OzwCZPTORVQblFbpZXYGDcUd9NdUQ=="],

    "@ai-sdk/openai-compatible": ["@ai-sdk/openai-compatible@2.0.41", "", { "dependencies": { "@ai-sdk/provider": "3.0.8", "@ai-sdk/provider-utils": "4.0.23" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-kNAGINk71AlOXx10Dq/PXw4t/9XjdK8uxfpVElRwtSFMdeSiLVt58p9TPx4/FJD+hxZuVhvxYj9r42osxWq79g=="],

@@ -3032,7 +3032,7 @@

    "@types/braces": ["@types/braces@3.0.5", "", {}, "sha512-SQFof9H+LXeWNz8wDe7oN5zu7ket0qwMu5vZubW4GCJ8Kkeh6nBWUz87+KTz/G3Kqsrp0j/W253XJb3KMEeg3w=="],

-    "@types/bun": ["@types/bun@1.3.13", "", { "dependencies": { "bun-types": "1.3.13" } }, "sha512-9fqXWk5YIHGGnUau9TEi+qdlTYDAnOj+xLCmSTwXfAIqXr2x4tytJb43E9uCvt09zJURKXwAtkoH4nLQfzeTXw=="],
+    "@types/bun": ["@types/bun@1.4.0", "", { "dependencies": { "bun-types": "1.4.0" } }, "sha512-K+lZULY23vRgK/CfTjFIV+tyifaNdSMlPh9j+6mQ/cLfpOznLyAuzgV/JQysyECpkBQLVMSyvjlr2fBUSA9wFQ=="],

    "@types/cacache": ["@types/cacache@20.0.1", "", { "dependencies": { "@types/node": "*", "minipass": "*" } }, "sha512-QlKW3AFoFr/hvPHwFHMIVUH/ZCYeetBNou3PCmxu5LaNDvrtBlPJtIA6uhmU9JRt9oxj7IYoqoLcpxtzpPiTcw=="],

@@ -3484,7 +3484,7 @@

    "bun-pty": ["bun-pty@0.4.8", "", {}, "sha512-rO70Mrbr13+jxHHHu2YBkk2pNqrJE5cJn29WE++PUr+GFA0hq/VgtQPZANJ8dJo6d7XImvBk37Innt8GM7O28w=="],

-    "bun-types": ["bun-types@1.3.13", "", { "dependencies": { "@types/node": "*" } }, "sha512-QXKeHLlOLqQX9LgYaHJfzdBaV21T63HhFJnvuRCcjZiaUDpbs5ED1MgxbMra71CsryN/1dAoXuJJJwIv/2drVA=="],
+    "bun-types": ["bun-types@1.4.2", "", { "dependencies": { "@types/node": "*" } }, "sha512-bxV1FgK7yBIzjRe5zBozIM4Bem11ZJcCXSrjWRG3YWLt8yFDePu4cLjpebO8OvPeIE9trbyPF4fuj3Cia4Fj3w=="],

    "bundle-name": ["bundle-name@4.1.0", "", { "dependencies": { "run-applescript": "^7.0.0" } }, "sha512-tjwM5exMg6BGRI+kNmTntNsvdZS1X8BFYS6tnJ2hdH0kVxM6/eVZ2xy+FqStSWvYmtfFMDLIxurorHwDKfDz5Q=="],

@@ -5864,6 +5864,8 @@

    "@ai-sdk/amazon-bedrock/@ai-sdk/anthropic": ["@ai-sdk/anthropic@3.0.81", "", { "dependencies": { "@ai-sdk/provider": "3.0.10", "@ai-sdk/provider-utils": "4.0.27" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-B1JDd9Ugq9R5AgIaW3674lhGCMMYJcPUxnrZh8fzbGojgg4QvHFRv6eZahGQAUsmGHbcf74G9bdSBDLWQGY2GA=="],

+    "@ai-sdk/amazon-bedrock/@ai-sdk/openai": ["@ai-sdk/openai@3.0.67", "", { "dependencies": { "@ai-sdk/provider": "3.0.10", "@ai-sdk/provider-utils": "4.0.27" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-oAiGC9eWG7IgtdsdS74bOCnAAHarAfTJhWN9x5INwnWPekL802AvF+0I5DvLzIF1MIRmNw4N8mPSL/GUVbX9Mw=="],
+
    "@ai-sdk/amazon-bedrock/@ai-sdk/provider": ["@ai-sdk/provider@3.0.10", "", { "dependencies": { "json-schema": "^0.4.0" } }, "sha512-Q3BZ27qfpYqnCYGvE3vt+Qi6LGOF9R5Nmzn+9JoM1lCRsD9mYaIhfJLkSunN48nfGXJ6n+XNV0J/XVpqGQl7Dw=="],

    "@ai-sdk/amazon-bedrock/@ai-sdk/provider-utils": ["@ai-sdk/provider-utils@4.0.27", "", { "dependencies": { "@ai-sdk/provider": "3.0.10", "@standard-schema/spec": "^1.1.0", "eventsource-parser": "^3.0.8" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-ubkAJ+xODouwtmN1tYlvTPphH1hPOBfZaEQe8U7skGvFAnIRs9PPpsq57bC2+Ky/MB4yzhd6YOsxTAx9sGpazw=="],
@@ -5872,8 +5874,6 @@

    "@ai-sdk/anthropic/@ai-sdk/provider-utils": ["@ai-sdk/provider-utils@4.0.27", "", { "dependencies": { "@ai-sdk/provider": "3.0.10", "@standard-schema/spec": "^1.1.0", "eventsource-parser": "^3.0.8" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-ubkAJ+xODouwtmN1tYlvTPphH1hPOBfZaEQe8U7skGvFAnIRs9PPpsq57bC2+Ky/MB4yzhd6YOsxTAx9sGpazw=="],

-    "@ai-sdk/azure/@ai-sdk/openai": ["@ai-sdk/openai@3.0.84", "", { "dependencies": { "@ai-sdk/provider": "3.0.14", "@ai-sdk/provider-utils": "4.0.38" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-cmgbeJL0bbY0yTJH4/AdmP5E7MjWRL9G8UdhIi0JlV/So03o82ORJofW8OzwCZPTORVQblFbpZXYGDcUd9NdUQ=="],
-
    "@ai-sdk/azure/@ai-sdk/provider": ["@ai-sdk/provider@3.0.14", "", { "dependencies": { "json-schema": "^0.4.0" } }, "sha512-5X1k57JBJ4H7H1QjX7CnJYAB1I19r/trVZTMcSms7/kLNZ8RaU4Nt2agcwZzv82Hfx6Q7/TOLU7agAKeFfc8cA=="],

    "@ai-sdk/azure/@ai-sdk/provider-utils": ["@ai-sdk/provider-utils@4.0.38", "", { "dependencies": { "@ai-sdk/provider": "3.0.14", "@standard-schema/spec": "^1.1.0", "eventsource-parser": "^3.0.8" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-/HHGmtKllqjg1OLc023v9w9kK3laW7Z6TzfZukYQWCsGBbzB9p60zTvvpXFVcs44NZBVXL3viOa1HRKUbeee8g=="],
@@ -5920,9 +5920,9 @@

    "@ai-sdk/mistral/@ai-sdk/provider-utils": ["@ai-sdk/provider-utils@4.0.40", "", { "dependencies": { "@ai-sdk/provider": "3.0.14", "@standard-schema/spec": "^1.1.0", "eventsource-parser": "^3.0.8" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-OL5IrpUm9Y8Dwy+w/vvFwPotS6m52O9W0op2oXgXdCROMJIBalBI0oro6OIBYkPxvm5Xg02GSkoQN25RlR0bnw=="],

-    "@ai-sdk/openai/@ai-sdk/provider": ["@ai-sdk/provider@3.0.10", "", { "dependencies": { "json-schema": "^0.4.0" } }, "sha512-Q3BZ27qfpYqnCYGvE3vt+Qi6LGOF9R5Nmzn+9JoM1lCRsD9mYaIhfJLkSunN48nfGXJ6n+XNV0J/XVpqGQl7Dw=="],
+    "@ai-sdk/openai/@ai-sdk/provider": ["@ai-sdk/provider@3.0.14", "", { "dependencies": { "json-schema": "^0.4.0" } }, "sha512-5X1k57JBJ4H7H1QjX7CnJYAB1I19r/trVZTMcSms7/kLNZ8RaU4Nt2agcwZzv82Hfx6Q7/TOLU7agAKeFfc8cA=="],

-    "@ai-sdk/openai/@ai-sdk/provider-utils": ["@ai-sdk/provider-utils@4.0.27", "", { "dependencies": { "@ai-sdk/provider": "3.0.10", "@standard-schema/spec": "^1.1.0", "eventsource-parser": "^3.0.8" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-ubkAJ+xODouwtmN1tYlvTPphH1hPOBfZaEQe8U7skGvFAnIRs9PPpsq57bC2+Ky/MB4yzhd6YOsxTAx9sGpazw=="],
+    "@ai-sdk/openai/@ai-sdk/provider-utils": ["@ai-sdk/provider-utils@4.0.38", "", { "dependencies": { "@ai-sdk/provider": "3.0.14", "@standard-schema/spec": "^1.1.0", "eventsource-parser": "^3.0.8" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-/HHGmtKllqjg1OLc023v9w9kK3laW7Z6TzfZukYQWCsGBbzB9p60zTvvpXFVcs44NZBVXL3viOa1HRKUbeee8g=="],

    "@ai-sdk/perplexity/@ai-sdk/provider-utils": ["@ai-sdk/provider-utils@4.0.21", "", { "dependencies": { "@ai-sdk/provider": "3.0.8", "@standard-schema/spec": "^1.1.0", "eventsource-parser": "^3.0.6" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-MtFUYI1/8mgDvRmaBDjbLJPFFrMG777AvSgyIFQtZHIMzm88R/12vYBBpnk7pfiWLFE1DSZzY4WDYzGbKAcmiw=="],

@@ -6340,10 +6340,6 @@

    "@solidjs/start/vite": ["vite@7.1.10", "", { "dependencies": { "esbuild": "^0.25.0", "fdir": "^6.5.0", "picomatch": "^4.0.3", "postcss": "^8.5.6", "rollup": "^4.43.0", "tinyglobby": "^0.2.15" }, "optionalDependencies": { "fsevents": "~2.3.3" }, "peerDependencies": { "@types/node": "^20.19.0 || >=22.12.0", "jiti": ">=1.21.0", "less": "^4.0.0", "lightningcss": "^1.21.0", "sass": "^1.70.0", "sass-embedded": "^1.70.0", "stylus": ">=0.54.8", "sugarss": "^5.0.0", "terser": "^5.16.0", "tsx": "^4.8.1", "yaml": "^2.4.2" }, "optionalPeers": ["@types/node", "jiti", "less", "lightningcss", "sass", "sass-embedded", "stylus", "sugarss", "terser", "tsx", "yaml"], "bin": { "vite": "bin/vite.js" } }, "sha512-CmuvUBzVJ/e3HGxhg6cYk88NGgTnBoOo7ogtfJJ0fefUWAxN/WDSUa50o+oVBxuIhO8FoEZW0j2eW7sfjs5EtA=="],

-    "@standard-community/standard-json/effect": ["effect@4.0.0-rc.111", "", { "dependencies": { "@standard-schema/spec": "^1.1.0", "fast-check": "^4.9.0", "msgpackr": "^2.0.5" } }, "sha512-ASd5L58EIR0CUNueZNKKjSsyOCd+2alxOAIaTcHaqkJkPsaYSsw5Cg/cfANk5K4Jr2YsX756xvX11shzqsreWA=="],
-
-    "@standard-community/standard-openapi/effect": ["effect@4.0.0-rc.111", "", { "dependencies": { "@standard-schema/spec": "^1.1.0", "fast-check": "^4.9.0", "msgpackr": "^2.0.5" } }, "sha512-ASd5L58EIR0CUNueZNKKjSsyOCd+2alxOAIaTcHaqkJkPsaYSsw5Cg/cfANk5K4Jr2YsX756xvX11shzqsreWA=="],
-
    "@tailwindcss/node/lightningcss": ["lightningcss@1.32.0", "", { "dependencies": { "detect-libc": "^2.0.3" }, "optionalDependencies": { "lightningcss-android-arm64": "1.32.0", "lightningcss-darwin-arm64": "1.32.0", "lightningcss-darwin-x64": "1.32.0", "lightningcss-freebsd-x64": "1.32.0", "lightningcss-linux-arm-gnueabihf": "1.32.0", "lightningcss-linux-arm64-gnu": "1.32.0", "lightningcss-linux-arm64-musl": "1.32.0", "lightningcss-linux-x64-gnu": "1.32.0", "lightningcss-linux-x64-musl": "1.32.0", "lightningcss-win32-arm64-msvc": "1.32.0", "lightningcss-win32-x64-msvc": "1.32.0" } }, "sha512-NXYBzinNrblfraPGyrbPoD19C1h9lfI/1mzgWYvXUTe414Gz/X1FD2XBZSZM7rRTrMA8JL3OtAaGifrIKhQ5yQ=="],

    "@tailwindcss/oxide-wasm32-wasi/@emnapi/core": ["@emnapi/core@1.11.3", "", { "dependencies": { "@emnapi/wasi-threads": "1.2.3", "tslib": "^2.4.0" }, "bundled": true }, "sha512-zLpS5asjEb7lq8jYLq37N6XKaE41DIexlY1rF/z4/tIl3wo13Sqm28fRyfIsKZD+NZ8mM5RoKkpW/rBcuoSZSg=="],
@@ -6386,8 +6382,6 @@

    "ai/@ai-sdk/provider-utils": ["@ai-sdk/provider-utils@5.0.27", "", { "dependencies": { "@ai-sdk/provider": "4.0.7", "@standard-schema/spec": "^1.1.0", "@workflow/serde": "4.1.0", "eventsource-parser": "^3.0.8", "undici": "^7.28.0" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-EzAn4pdgG5g0xXtH6lE2zyNmfjDQIDjATkfqzuidEI35g++hh4+07vnjzkT/RmGmIClPZiRj/Q2GMPV2V7mkHw=="],

-    "ai-gateway-provider/@ai-sdk/openai": ["@ai-sdk/openai@3.0.84", "", { "dependencies": { "@ai-sdk/provider": "3.0.14", "@ai-sdk/provider-utils": "4.0.38" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-cmgbeJL0bbY0yTJH4/AdmP5E7MjWRL9G8UdhIi0JlV/So03o82ORJofW8OzwCZPTORVQblFbpZXYGDcUd9NdUQ=="],
-
    "ansi-align/string-width": ["string-width@4.2.3", "", { "dependencies": { "emoji-regex": "^8.0.0", "is-fullwidth-code-point": "^3.0.0", "strip-ansi": "^6.0.1" } }, "sha512-wKyQRQpjJ0sIp62ErSZdGsjMJWsap5oRNihHhu6G7JVO/9jIB6UyevL+tXuOqrng8j/cxKTWyWUwvSTriiZz/g=="],

    "anymatch/picomatch": ["picomatch@2.3.2", "", {}, "sha512-V7+vQEJ06Z+c5tSye8S+nHUfI51xoXIXjHQ99cQtKUkQqqO1kO/KCJUfZXuB47h/YBlDhah2H3hdUGXn8ie0oA=="],
@@ -6922,8 +6916,6 @@

    "@brendonovich/vite-plugin-opencode/@opencode-ai/client/@opencode-ai/schema": ["@opencode-ai/schema@0.0.0-beta-18050", "", { "dependencies": { "@standard-schema/spec": "1.1.0", "effect": "4.0.0-rc.111" } }, "sha512-/D6VXaWlytTXR3IOiMLIKuPcfp7FQNUzRPm9z3K7UBFd1Bw4q/WZksaf5RVcBGz+0YRxYMc1V4D7MFlceSgtyg=="],

-    "@brendonovich/vite-plugin-opencode/@opencode-ai/client/effect": ["effect@4.0.0-rc.111", "", { "dependencies": { "@standard-schema/spec": "^1.1.0", "fast-check": "^4.9.0", "msgpackr": "^2.0.5" } }, "sha512-ASd5L58EIR0CUNueZNKKjSsyOCd+2alxOAIaTcHaqkJkPsaYSsw5Cg/cfANk5K4Jr2YsX756xvX11shzqsreWA=="],
-
    "@bruits/satteri-wasm32-wasi/@emnapi/core/@emnapi/wasi-threads": ["@emnapi/wasi-threads@1.2.2", "", { "dependencies": { "tslib": "^2.4.0" } }, "sha512-c95qOXkHdydNKhscBTebqEC1CVAZpyqOfVfBzQ1qgzyl3gfeldUjIggDbIZgDKsHLgnsM+igH7TJ/eAasaVuMA=="],

    "@electron/asar/minimatch/brace-expansion": ["brace-expansion@1.1.18", "", { "dependencies": { "balanced-match": "^1.0.0", "concat-map": "0.0.1" } }, "sha512-Edep/X9fGqVNmzKBVsDYIOtD+z1tuezV70LBjdCst9Tqu76lsnvRiZ6oTic1n+/BIwX6QDGAO94PN4N2SADvtw=="],
@@ -7250,10 +7242,6 @@

    "@vitest/expect/@vitest/utils/@vitest/pretty-format": ["@vitest/pretty-format@3.2.4", "", { "dependencies": { "tinyrainbow": "^2.0.0" } }, "sha512-IVNZik8IVRJRTr9fxlitMKeJeXFFFN0JaB9PHPGQ8NKQbGpfjlTx9zO4RefN8gp7eqjNy8nyK3NZmBzOPeIxtA=="],

-    "ai-gateway-provider/@ai-sdk/openai/@ai-sdk/provider": ["@ai-sdk/provider@3.0.14", "", { "dependencies": { "json-schema": "^0.4.0" } }, "sha512-5X1k57JBJ4H7H1QjX7CnJYAB1I19r/trVZTMcSms7/kLNZ8RaU4Nt2agcwZzv82Hfx6Q7/TOLU7agAKeFfc8cA=="],
-
-    "ai-gateway-provider/@ai-sdk/openai/@ai-sdk/provider-utils": ["@ai-sdk/provider-utils@4.0.38", "", { "dependencies": { "@ai-sdk/provider": "3.0.14", "@standard-schema/spec": "^1.1.0", "eventsource-parser": "^3.0.8" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-/HHGmtKllqjg1OLc023v9w9kK3laW7Z6TzfZukYQWCsGBbzB9p60zTvvpXFVcs44NZBVXL3viOa1HRKUbeee8g=="],
-
    "ai/@ai-sdk/provider-utils/undici": ["undici@7.29.0", "", {}, "sha512-IDxfleLmmbSskfWSUATiN1nfn2rDuvnMOqb5CWR92iIfojA0Ud+ulOAAEQ57LPr9rWmsreUyf5lwyao+7GNNVw=="],

    "ansi-align/string-width/emoji-regex": ["emoji-regex@8.0.0", "", {}, "sha512-MSjYzcWNOA0ewAHpz0MxpYFvwg6yjy1NG3xteoqz644VCo/RPgnr1/GGt+ic3iJTzQ8Eu3TdM14SawnVUmGE6A=="],
@@ -235,6 +235,7 @@ const bucketNew = new sst.cloudflare.Bucket("ZenDataNew")
 const DISCORD_INCIDENT_WEBHOOK_URL = new sst.Secret("DISCORD_INCIDENT_WEBHOOK_URL")
 const AWS_SES_ACCESS_KEY_ID = new sst.Secret("AWS_SES_ACCESS_KEY_ID")
 const AWS_SES_SECRET_ACCESS_KEY = new sst.Secret("AWS_SES_SECRET_ACCESS_KEY")
+const ENTERPRISE_SALES_INBOX_EMAIL = new sst.Secret("ENTERPRISE_SALES_INBOX_EMAIL")

 const SALESFORCE_CLIENT_ID = new sst.Secret("SALESFORCE_CLIENT_ID")
 const SALESFORCE_CLIENT_SECRET = new sst.Secret("SALESFORCE_CLIENT_SECRET")
@@ -263,6 +264,7 @@ new sst.cloudflare.x.SolidStart("Console", {
    EMAILOCTOPUS_API_KEY,
    AWS_SES_ACCESS_KEY_ID,
    AWS_SES_SECRET_ACCESS_KEY,
+    ENTERPRISE_SALES_INBOX_EMAIL,
    SALESFORCE_CLIENT_ID,
    SALESFORCE_CLIENT_SECRET,
    SALESFORCE_INSTANCE_URL,
@@ -1,8 +1,8 @@
 {
  "nodeModules": {
-    "x86_64-linux": "sha256-I2sHU46Bdp1DmW/ygNKKxF6U53oNIeVeAS8TmMdtvzw=",
-    "aarch64-linux": "sha256-aQUEBmfHhaR/YdsRyw0u6hsEo0TEmXHBMyiGnOFRCeI=",
-    "aarch64-darwin": "sha256-40Y3HYSjs8S0M3SFNUrU8lMYaHy7vpmiW+7b6Y9pKzs=",
-    "x86_64-darwin": "sha256-7iGegI/1pXcR5W0UBX0wAWAga+F34fhj9lbAHotVwEo="
+    "x86_64-linux": "sha256-T1JUv8gUrXavDI1HdVGqNRal4y2Bh5QavpR4DV+bJdA=",
+    "aarch64-linux": "sha256-wlbWdEUVWAafgxaZis0Q+SAas8k6hu3s0C39CnMZ2gw=",
+    "aarch64-darwin": "sha256-74rRue8KQWKzgMMb540beqMzp1DUShlPte6qklYCVTs=",
+    "x86_64-darwin": "sha256-ihr3DoLd/4Lw6WbyXzSQXMP4hrVGk/UiRkO2aoQn8vg="
  }
 }
@@ -5,9 +5,9 @@
  "version": "0.0.0",
  "private": true,
  "type": "module",
-  "packageManager": "bun@1.4.0",
+  "packageManager": "bun@1.4.2",
  "scripts": {
-    "dev": "bun run --cwd packages/cli --conditions=browser src/index.ts",
+    "dev": "bun run --cwd packages/cli src/index.ts",
    "dev:live": "sh -c 'OPENCODE_TUI_CHANNEL=dev OPENCODE_PASSWORD=\"$(opencode2 service get password)\" exec bun run dev \"$@\" --server \"$(opencode2 service status)\"' --",
    "dev:desktop": "bun --cwd packages/desktop dev",
    "dev:web": "bun --cwd packages/app dev",
@@ -43,7 +43,7 @@
      "@effect/platform-node-shared": "4.0.0-rc.112",
      "@effect/sql-sqlite-bun": "4.0.0-rc.112",
      "@npmcli/arborist": "9.4.0",
-      "@types/bun": "1.3.13",
+      "@types/bun": "1.4.0",
      "@types/cross-spawn": "6.0.6",
      "@octokit/rest": "22.0.0",
      "@hono/standard-validator": "0.2.0",
@@ -146,10 +146,6 @@
    "esbuild",
    "node-pty",
    "protobufjs",
-    "tree-sitter",
-    "tree-sitter-bash",
-    "tree-sitter-powershell",
-    "web-tree-sitter",
    "electron"
  ],
  "overrides": {
@@ -159,6 +155,7 @@
    "@effect/platform-node-shared": "catalog:",
    "@types/bun": "catalog:",
    "@types/node": "catalog:",
+    "bun-types": "1.4.2",
    "effect": "catalog:",
    "solid-js": "catalog:"
  },
@@ -81,7 +81,7 @@ export const route = Route.make({

 Route defaults are request-shaping defaults such as `headers`, `limits`, `generation`, `providerOptions`, and `http`. Endpoint host/query belongs on the route endpoint. Selected `LanguageModel` values carry only model id, provider id, and the configured route value. Model capability/catalog metadata lives outside this package; protocol support is enforced by request lowering and typed `AIError`s.

-The four-axis decomposition is the reason DeepSeek, TogetherAI, Cerebras, Baseten, Fireworks, and DeepInfra all reuse `OpenAIChat.protocol` verbatim — each provider deployment is a 5-15 line `Route.make(...)` call instead of a 300-400 line route clone. Bug fixes in one protocol propagate to every consumer of that protocol in a single commit.
+The four-axis decomposition is the reason DeepSeek, TogetherAI, Cerebras, Baseten, Fireworks, and DeepInfra all reuse `OpenAIChat.protocol` verbatim — each provider owns a small `Route.make(...)` composition instead of a protocol clone. Bug fixes in one protocol propagate to every consumer of that protocol in a single commit.

 When a provider supports multiple physical transports, selection remains execution policy below its semantic route. `OpenResponsesChannel.transport(...)` owns the provider-neutral Responses WebSocket concept: it prepares one final request, executes HTTP by default, strips WebSocket-disallowed fields, and passes a generic channel exchange to a per-call `WebSocketChannelExecutor` when supplied. Provider-specific Responses routes opt in with handshake and connection-age policy. `Route.streamPrepared` owns decoding and acknowledges channel completion only after successful full consumption.

@@ -115,6 +115,7 @@ Keep provider facades small and explicit:
 - Prefer `apiKey` as provider-specific sugar and `auth` as the explicit override; keep them mutually exclusive in provider option types with `ProviderAuthOption`.
 - Resolve `apiKey` → `Auth` with `AuthOptions.bearer(options, "<PROVIDER>_API_KEY")` (it honors an explicit `auth` override and falls back to `Auth.config(envVar)` so missing keys surface a typed `Authentication` error rather than a runtime crash).
 - Use separate top-level facades for products with different required setup, such as `CloudflareAIGateway` and `CloudflareWorkersAI`.
+- Give every named provider its own file and top-level export. Keep its endpoint, auth defaults, and route setup in that file. Compose shared protocols directly; do not nest named provider presets under generic compatible facades or keep their endpoints in a shared provider profile registry.

 `Provider.make(...)` remains available for simple static provider definitions, but new built-in providers should prefer plain configured facades unless a helper removes real duplication without adding runtime behavior.

@@ -247,6 +247,8 @@ it does not repair or truncate them.

 For explicit compaction, script a `CompactionResponse` through `push`, `always`, or `serve`. Its `replacement` contains the next context window, including retained user messages. The client returns that result and usage directly, with the same lazy request recording and gates. Generation and compaction reject fixtures for the wrong operation instead of converting between response shapes.

+For `compact(request, { mechanism: "trigger" })`, script a `CompactionCheckpointResponse` instead. It carries `checkpoint`, `responseID`, and optional `usage`. Endpoint and trigger calls reject each other's fixtures; both share the same queue, gates, lazy recording, and fallback controls.
+
 The published legacy `Service`, `layer`, `clientLayer`, and module-level controls remain available as adapters
 over the same implementation, including the legacy live `requests` array. New tests should use `Test` and
 `testLayer`.
@@ -259,7 +261,7 @@ This is different from prompt caching, server-side history storage, or truncatio

 ### Explicit compaction

-`LLMClient.compact(request)` is the caller-controlled operation for OpenAI, Azure, and xAI Responses. It performs exactly one HTTP call to `/responses/compact`, using the selected route's endpoint, credentials, query, and HTTP middleware. It returns a `CompactionResponse` with `replacement: Message[]` and optional `usage`, not a normal generation response.
+`LLMClient.compact(request)` (equivalently, `{ mechanism: "endpoint" }`) is the caller-controlled operation for OpenAI, Azure, and xAI Responses. It performs exactly one HTTP call to `/responses/compact`, using the selected route's endpoint, credentials, query, and HTTP middleware. It returns a `CompactionResponse` with `replacement: Message[]` and optional `usage`, not a normal generation response. This mechanism does not accept a WebSocket executor.

 Prefer this operation, where supported, when the application owns compaction policy and durable context updates.

@@ -279,6 +281,41 @@ Generation-only body overlays such as `stream` and `store` are not sent to the c

 The input must still fit the model's context window. Explicit compaction is not an overflow-recovery operation. Anthropic does not expose this operation in this package; its in-band compaction remains available below. Compatible routes do not inherit an explicit compact endpoint simply because they use a Responses protocol.

+### Streamed checkpoint compaction
+
+OpenAI Responses also exposes a separate, explicitly selected mechanism:
+
+```ts
+const result =
+  yield *
+  LLMClient.compact(request, {
+    mechanism: "trigger",
+    webSocket, // Optional: without it, the request uses HTTP/SSE.
+  })
+
+result.checkpoint // Successful encrypted CompactionPart.
+result.responseID
+result.usage
+```
+
+This appends a native `compaction_trigger` control item to the full input and sends a normal Responses request. It follows the [Codex V2 request shape](https://github.com/openai/codex/blob/728cb12/codex-rs/core/src/compact_remote_v2_attempt.rs), with tools and instructions retained, `stream: true`, `store: false`, and parallel tool calls enabled. It removes normal-answer text/output-format controls, forced tool choices, output-token/tool-call limits, and automatic `context_management`. Body overlays cannot replace `input` or supply `previous_response_id`/`conversation`; the complete canonical history is required for safe stateless replay. Session/cache identifiers, auth, headers, query parameters, service tier, and supported prompt-cache settings are preserved.
+
+Only a successful `response.completed` with a response ID and exactly one logical encrypted checkpoint succeeds. Repeated item events are correlated by ID/output slot, including ID-less checkpoints. Other output is ignored, not returned as assistant text or dispatched as tools. Failed, incomplete, malformed, and interrupted responses return errors rather than partial checkpoints.
+
+The result is **not a replacement window**. The caller selects retained history, combines it with `result.checkpoint`, and durably installs it before continuing. The operation does not choose a retention budget, prune messages, or modify the original request.
+
+The supplied WebSocket executor can reuse a compatible append baseline for the compaction request. On completion the protocol supplies no continuation checkpoint, clearing the old baseline so the next generation sends the newly installed window in full. Validation occurs before transport completion is acknowledged. There is no operation-level retry or fallback to `/responses/compact`; existing safe transport fallback may use SSE, with full history and no connection-local response ID.
+
+Trigger support is separate from endpoint support. Only the OpenAI Responses route advertises it; Azure, xAI, Chat, and compatible Responses routes do not inherit it. Untyped calls still fail before sending: missing route capabilities return `UnsupportedOperation`, while unknown mechanism names and invalid inputs return `InvalidRequest`. Dynamic callers must narrow for the selected mechanism:
+
+```ts
+if (LLMClient.canCompact(request, { mechanism: "trigger" })) {
+  const result = yield * LLMClient.compact(request, { mechanism: "trigger" })
+}
+```
+
+This capability describes protocol implementation, **not universal availability on OpenAI API deployments**. The host application owns subscription/deployment eligibility, OAuth, endpoint selection, and deployment-specific headers. Local protocol/socket tests do not establish live provider support.
+
 ### Advanced: in-band compaction

 `providerOptions.contextManagement` lets the provider decide when to compact during an ordinary `generate` or `stream` call. This is an advanced option for callers that own persistence and recovery: persist the complete assistant message, including its checkpoint, before continuing. Enabling the option does not provide durable checkpoint storage, interruption recovery, or model-switch policy. Keep the prior context until a successful checkpoint has been persisted.
@@ -424,7 +461,18 @@ const gateway = CloudflareAIGateway.configure({
 }).model("workers-ai/@cf/meta/llama-3.1-8b-instruct")
 ```

-Included providers: OpenAI, Anthropic, Google (Gemini), Google Vertex Gemini and Anthropic, Amazon Bedrock, Azure OpenAI, Cloudflare AI Gateway, Cloudflare Workers AI, GitHub Copilot, OpenRouter, xAI, Z.ai, plus generic OpenAI-compatible Chat and Responses entrypoints and an Anthropic Messages-compatible entrypoint.
+Included LLM providers: OpenAI, Anthropic, Google (Gemini), Google Vertex, Amazon Bedrock, Azure OpenAI, Baseten, Cerebras, Cloudflare AI Gateway, Cloudflare Workers AI, DeepInfra, DeepSeek, Fireworks, Groq, Mistral, OpenRouter, TogetherAI, and xAI. Z.ai currently exposes image generation. Generic Chat Completions, Responses, and Anthropic Messages-compatible entrypoints support custom endpoints.
+
+Each named provider owns its module, endpoint, authentication, and route setup. Providers with the same wire format compose the shared protocol directly:
+
+```ts
+import { DeepSeek, Fireworks } from "@opencode-ai/ai/providers"
+
+const deepseek = DeepSeek.configure({ apiKey }).model("deepseek-chat")
+const fireworks = Fireworks.configure({ apiKey }).model("accounts/fireworks/models/my-model")
+```
+
+The former `OpenAICompatible.baseten`, `.cerebras`, `.deepinfra`, `.deepseek`, `.fireworks`, `.groq`, and `.togetherai` presets are replaced by the top-level `Baseten`, `Cerebras`, `DeepInfra`, `DeepSeek`, `Fireworks`, `Groq`, and `TogetherAI` exports. Use `CloudflareAIGateway` and `CloudflareWorkersAI` directly; each has its own module. `OpenAICompatible` configures generic endpoints with an explicit `baseURL`.

 ### Package-like entrypoints

@@ -482,7 +530,15 @@ model("claude-sonnet-4-6", { project: "my-project", location: "global" })

 Provider facades such as `OpenAI.configure(...).responses(...)` remain the direct application API. Package-like entrypoints are the self-similar loading contract used when a catalog selects behavior by export path.

-Other provider exports listed above remain direct facades until they explicitly implement the package-like contract. Exporting a provider facade does not implicitly make it a catalog-loadable provider package.
+Every named LLM provider listed above also exports `model(modelID, settings)` from its own package entrypoint. The extracted providers are available at:
+
+- `@opencode-ai/ai/providers/baseten`
+- `@opencode-ai/ai/providers/deepseek`
+- `@opencode-ai/ai/providers/fireworks`
+- `@opencode-ai/ai/providers/cloudflare-ai-gateway`
+- `@opencode-ai/ai/providers/cloudflare-workers-ai`
+
+Core resolves known compatible catalog providers through their dedicated entrypoints, including the `fireworks-ai` catalog ID through `providers/fireworks`.

 ## Provider options & HTTP overlays

@@ -28,6 +28,7 @@
    "typescript": "catalog:"
  },
  "dependencies": {
+    "@aws-sdk/credential-providers": "3.1057.0",
    "@smithy/eventstream-codec": "4.2.14",
    "@smithy/util-utf8": "4.2.2",
    "@opencode-ai/schema": "workspace:*",
@@ -7,7 +7,8 @@ import { AwsV4Signer } from "aws4fetch"
 import { Config, ConfigProvider, Effect, FileSystem, PlatformError, Redacted } from "effect"
 import { FetchHttpClient, HttpClient, HttpClientRequest, type HttpClientResponse } from "effect/unstable/http"
 import * as ProviderShared from "../src/protocols/shared"
-import * as Cloudflare from "../src/providers/cloudflare"
+import { CloudflareAIGateway } from "../src/providers/cloudflare-ai-gateway.js"
+import { CloudflareWorkersAI } from "../src/providers/cloudflare-workers-ai.js"

 type Provider = {
  readonly id: string
@@ -120,11 +121,11 @@ const PROVIDERS: ReadonlyArray<Provider> = [
    ],
    validate: (env) =>
      validateChat({
-        url: `${Cloudflare.aiGatewayBaseURL({
+        url: `${CloudflareAIGateway.baseURL({
          accountId: env.CLOUDFLARE_ACCOUNT_ID,
          gatewayId: env.CLOUDFLARE_GATEWAY_ID || undefined,
        })}/chat/completions`,
-        token: Redacted.make(envValue(env, Cloudflare.aiGatewayAuthEnvVars)),
+        token: Redacted.make(envValue(env, CloudflareAIGateway.authEnvVars)),
        tokenHeader: "cf-aig-authorization",
        model: "workers-ai/@cf/meta/llama-3.1-8b-instruct",
      }),
@@ -140,8 +141,8 @@ const PROVIDERS: ReadonlyArray<Provider> = [
    ],
    validate: (env) =>
      validateChat({
-        url: `${Cloudflare.workersAIBaseURL({ accountId: env.CLOUDFLARE_ACCOUNT_ID })}/chat/completions`,
-        token: Redacted.make(envValue(env, Cloudflare.workersAIAuthEnvVars)),
+        url: `${CloudflareWorkersAI.baseURL({ accountId: env.CLOUDFLARE_ACCOUNT_ID })}/chat/completions`,
+        token: Redacted.make(envValue(env, CloudflareWorkersAI.authEnvVars)),
        model: "@cf/meta/llama-3.1-8b-instruct",
      }),
  },
@@ -11,7 +11,7 @@
 // Manual `cache: CacheHint` placements on individual parts are preserved and
 // count against the four-breakpoint budget; auto only fills remaining slots.
 import { CacheHint, type CachePolicy, type CachePolicyObject } from "./schema/options.js"
-import { LLMRequest, Message, ToolDefinition, type ContentPart } from "./schema/messages.js"
+import { LLMRequest, Message, ToolDefinition, type ContentPart, type ToolEntry } from "./schema/messages.js"

 const AUTO: CachePolicyObject = {
  tools: true,
@@ -50,18 +50,24 @@ interface Budget {
  remaining: number
 }

-const markLastTool = (
-  tools: ReadonlyArray<ToolDefinition>,
-  hint: CacheHint,
-  budget: Budget,
-): ReadonlyArray<ToolDefinition> => {
-  if (tools.length === 0) return tools
-  const last = tools.length - 1
-  if (tools[last]!.cache || budget.remaining === 0) return tools
+const markLastTool = (tools: ReadonlyArray<ToolEntry>, hint: CacheHint, budget: Budget): ReadonlyArray<ToolEntry> => {
+  const target = tools.at(-1)
+  if (target === undefined) return tools
+  if (target.type === "namespace") {
+    const nested = markLastTool(target.tools, hint, budget)
+    return nested === target.tools ? tools : [...tools.slice(0, -1), { ...target, tools: nested }]
+  }
+  if (target.cache || budget.remaining === 0) return tools
  budget.remaining -= 1
-  return tools.map((tool, i) => (i === last ? new ToolDefinition({ ...tool, cache: hint }) : tool))
+  return [...tools.slice(0, -1), new ToolDefinition({ ...target, cache: hint })]
 }

+const countToolHints = (tools: ReadonlyArray<ToolEntry>): number =>
+  tools.reduce(
+    (count, tool) => count + (tool.type === "tool" ? (tool.cache === undefined ? 0 : 1) : countToolHints(tool.tools)),
+    0,
+  )
+
 const markSystemBoundaries = (system: LLMRequest["system"], hint: CacheHint, budget: Budget): LLMRequest["system"] => {
  if (system.length === 0) return system
  let changed = false
@@ -122,7 +128,7 @@ const markMessages = (
 }

 const countHints = (request: LLMRequest) =>
-  request.tools.reduce((count, tool) => count + (tool.cache === undefined ? 0 : 1), 0) +
+  countToolHints(request.tools) +
  request.system.reduce((count, part) => count + (part.cache === undefined ? 0 : 1), 0) +
  request.messages.reduce(
    (count, message) =>
@@ -12,9 +12,10 @@ import {
  LanguageModel,
  SystemPart,
  ToolChoice,
-  ToolDefinition,
+  ToolEntry,
  type ContentPart,
  type LanguageModelProviderOptions,
+  type ToolEntryInput,
 } from "./schema/index.js"
 import { make as makeTool, toDefinitions, type ToolSchema } from "./tool.js"

@@ -27,7 +28,7 @@ export type RequestInput<SelectedLanguageModel extends LanguageModel = LanguageM
  readonly system?: string | SystemPart | ReadonlyArray<SystemPart>
  readonly prompt?: string | ContentPart | ReadonlyArray<ContentPart>
  readonly messages?: ReadonlyArray<Message | Message.Input>
-  readonly tools?: ReadonlyArray<ToolDefinition.Input>
+  readonly tools?: ReadonlyArray<ToolEntryInput>
  readonly toolChoice?: ToolChoice.Input
  readonly generation?: GenerationOptions.Input
  readonly providerOptions?: NoInfer<LanguageModelProviderOptions<SelectedLanguageModel>>
@@ -56,7 +57,7 @@ export const request = <const SelectedLanguageModel extends LanguageModel>(
    ...rest,
    system: SystemPart.content(requestSystem),
    messages: [...(messages?.map(Message.make) ?? []), ...(prompt === undefined ? [] : [Message.user(prompt)])],
-    tools: tools?.map(ToolDefinition.make) ?? [],
+    tools: tools?.map(ToolEntry.make) ?? [],
    toolChoice: requestToolChoice ? ToolChoice.make(requestToolChoice) : undefined,
    generation: requestGeneration === undefined ? undefined : GenerationOptions.make(requestGeneration),
    providerOptions: requestProviderOptions,
@@ -1067,10 +1067,11 @@ const fromRequest = Effect.fn("AnthropicMessages.fromRequest")(function* (reques
  // messages. Tools live highest in the cache hierarchy, so when callers
  // over-mark we keep their tool hints and shed the message-tail ones first.
  const breakpoints = Cache.newBreakpoints(ANTHROPIC_BREAKPOINT_CAP)
+  const flattened = ProviderShared.flattenToolRequest(request)
  const tools =
-    request.tools.length === 0
+    flattened.tools.length === 0
      ? undefined
-      : request.tools.map((tool) =>
+      : flattened.tools.map((tool) =>
          lowerTool(
            breakpoints,
            tool,
@@ -1088,7 +1089,7 @@ const fromRequest = Effect.fn("AnthropicMessages.fromRequest")(function* (reques
          text: part.text,
          cache_control: cacheControl(breakpoints, part.cache),
        }))
-  const messages = yield* lowerMessages(request, breakpoints)
+  const messages = yield* lowerMessages(flattened.request, breakpoints)
  if (breakpoints.dropped > 0) {
    yield* Effect.logWarning(
      `Anthropic Messages: dropped ${breakpoints.dropped} cache breakpoint(s); the API allows at most ${ANTHROPIC_BREAKPOINT_CAP} per request.`,
@@ -415,10 +415,7 @@ const lowerMessages = Effect.fn("BedrockConverse.lowerMessages")(function* (

 // System prompts share the cache-point convention: emit the text block, then
 // optionally a positional `cachePoint` marker.
-const lowerSystem = (
-  breakpoints: BedrockCache.Breakpoints,
-  system: ReadonlyArray<LLMRequest["system"][number]>,
-) => {
+const lowerSystem = (breakpoints: BedrockCache.Breakpoints, system: ReadonlyArray<LLMRequest["system"][number]>) => {
  const content = system
    .filter((part) => part.text.length > 0)
    .flatMap((part) => textWithCache(breakpoints, part.text, part.cache))
@@ -427,21 +424,22 @@ const lowerSystem = (

 const fromRequest = Effect.fn("BedrockConverse.fromRequest")(function* (request: LLMRequest) {
  const toolChoice = request.toolChoice ? yield* lowerToolChoice(request.toolChoice) : undefined
+  const flattened = ProviderShared.flattenToolRequest(request)
  const generation = request.generation
  // Bedrock-Claude shares Anthropic's 4-breakpoint cap. Spend the budget in
  // tools → system → messages order to favour the highest-impact prefixes.
-  const breakpoints = BedrockCache.breakpoints()
+  const breakpoints = BedrockCache.breakpoints(request.model.id)
  const toolConfig = (() => {
-    if (request.tools.length === 0) return undefined
+    if (flattened.tools.length === 0) return undefined
    return {
-      tools: lowerTools(request.model.compatibility?.toolSchema, breakpoints, request.tools),
+      tools: lowerTools(request.model.compatibility?.toolSchema, breakpoints, flattened.tools),
      // Converse has no native "none". Keep definitions stable for prompt
      // caching and omit only the unsupported choice.
      toolChoice,
    }
  })()
  const system = lowerSystem(breakpoints, request.system)
-  const messages = yield* lowerMessages(request, breakpoints)
+  const messages = yield* lowerMessages(flattened.request, breakpoints)
  if (breakpoints.dropped > 0) {
    yield* Effect.logWarning(
      `Bedrock Converse: dropped ${breakpoints.dropped} cache breakpoint(s); the API allows at most ${BedrockCache.BEDROCK_BREAKPOINT_CAP} per request.`,
@@ -508,7 +506,6 @@ const mapUsage = (usage: BedrockUsageSchema | undefined, providerMetadataKey: st
 interface ParserState {
  readonly providerMetadataKey: string
  readonly tools: ToolStream.State<number>
-  readonly finishedTools: ReadonlySet<number>
  // Bedrock splits the finish into `messageStop` (carries `stopReason`) and
  // `metadata` (carries usage). Hold both in state so `onHalt` can emit exactly
  // one finish after both chunks have had a chance to arrive.
@@ -620,16 +617,14 @@ const step = (state: ParserState, event: BedrockEvent) =>
    }

    if (event.contentBlockDelta?.delta?.toolUse) {
-      const index = event.contentBlockDelta.contentBlockIndex
-      if (state.finishedTools.has(index)) return [state, []] as const
-      const result = ToolStream.appendExisting(
-        ADAPTER,
+      // A delta for a block that is not open, whether it already stopped or never
+      // started, has nothing to attach to and is dropped.
+      const result = ToolStream.append(
        state.tools,
-        index,
+        event.contentBlockDelta.contentBlockIndex,
        event.contentBlockDelta.delta.toolUse.input,
-        "Bedrock Converse tool delta is missing its tool call",
      )
-      if (ToolStream.isError(result)) return yield* result
+      if (!result) return [state, []] as const
      const events: LLMEvent[] = []
      const lifecycle = result.events.length ? Lifecycle.stepStart(state.lifecycle, events) : state.lifecycle
      events.push(...result.events)
@@ -668,7 +663,6 @@ const step = (state: ParserState, event: BedrockEvent) =>
            state.hasToolCalls,
          lifecycle,
          tools: result.tools,
-          finishedTools: resultEvents.length > 0 ? new Set([...state.finishedTools, index]) : state.finishedTools,
          reasoningSignatures: Object.fromEntries(
            Object.entries(state.reasoningSignatures).filter(([key]) => key !== String(index)),
          ),
@@ -765,7 +759,6 @@ export const protocol = Protocol.make({
    initial: (request) => ({
      providerMetadataKey: request.model.route.providerMetadataKey ?? String(request.model.provider),
      tools: ToolStream.empty<number>(),
-      finishedTools: new Set<number>(),
      finishReason: undefined,
      usage: undefined,
      hasToolCalls: false,
@@ -465,7 +465,8 @@ function mapSafetySettings(value: unknown) {
 }

 const fromRequest = Effect.fn("Gemini.fromRequest")(function* (request: LLMRequest) {
-  const hasTools = request.tools.length > 0
+  const flattened = ProviderShared.flattenToolRequest(request)
+  const hasTools = flattened.tools.length > 0
  const generation = request.generation
  const options = resolveOptions(request)
  const toolSchemaCompatibility = request.model.compatibility?.toolSchema
@@ -483,7 +484,7 @@ const fromRequest = Effect.fn("Gemini.fromRequest")(function* (request: LLMReque

  return {
    cachedContent: options.cachedContent,
-    contents: yield* lowerMessages(request),
+    contents: yield* lowerMessages(flattened.request),
    safetySettings: options.safetySettings,
    serviceTier: options.serviceTier,
    systemInstruction:
@@ -491,7 +492,7 @@ const fromRequest = Effect.fn("Gemini.fromRequest")(function* (request: LLMReque
    tools: hasTools
      ? [
          {
-            functionDeclarations: request.tools.map((tool) =>
+            functionDeclarations: flattened.tools.map((tool) =>
              lowerTool(tool, ToolSchemaProjection.modelCompatibility(tool.inputSchema, toolSchemaCompatibility)),
            ),
          },
@@ -638,6 +639,14 @@ const step = (state: ParserState, event: GeminiEvent) => {
      : state.usage,
  }
  const candidate = event.candidates?.[0]
+  if (candidate?.finishReason && mapFinishReason(candidate.finishReason, state.hasToolCalls) === "error")
+    return Effect.fail(
+      ProviderShared.eventError(
+        state.route,
+        `Gemini stopped with ${candidate.finishReason}`,
+        ProviderShared.encodeJson(event),
+      ),
+    )
  if (!candidate?.content)
    return Effect.succeed([
      { ...nextState, finishReason: candidate?.finishReason ?? nextState.finishReason },
@@ -9,6 +9,8 @@ import {
  AIError,
  InvalidProviderOutputError,
  LLMEvent,
+  ProviderInternalError,
+  UnknownProviderError,
  Usage,
  type FinishReasonDetails,
  type LLMRequest,
@@ -414,10 +416,11 @@ export const fromRequest = Effect.fn("MistralChat.fromRequest")(function* (reque
        tool: (name) => ({ type: "function" as const, function: { name } }),
      })
    : undefined
+  const flattened = ProviderShared.flattenToolRequest(request)
  return {
    model: request.model.id,
-    messages: yield* lowerMessages(request),
-    tools: request.tools.length > 0 ? request.tools.map(lowerTool) : undefined,
+    messages: yield* lowerMessages(flattened.request),
+    tools: flattened.tools.length > 0 ? flattened.tools.map(lowerTool) : undefined,
    tool_choice: toolChoice,
    stream: true as const,
    max_tokens: request.generation?.maxTokens,
@@ -699,6 +702,16 @@ const step = Effect.fn("MistralChat.step")(function* (state: ParserState, event:
    normalized: mapFinishReason(choice.finish_reason),
    raw: choice.finish_reason,
  }
+  if (finishReason.normalized === "error") {
+    const details = {
+      message: `Mistral Chat stopped with ${finishReason.raw}`,
+      body: ProviderShared.encodeJson(event),
+    }
+    return yield* new AIError({
+      reason:
+        finishReason.raw === "network_error" ? new ProviderInternalError(details) : new UnknownProviderError(details),
+    })
+  }
  const incomplete = finishReason.normalized === "length" || finishReason.normalized === "content-filter"
  if (!incomplete && Object.keys(withTools.pendingTools).length > 0)
    return yield* ProviderShared.eventError(
@@ -113,8 +113,8 @@ const driver = (options: Options, body: string): WebSocketChannelDriver => {
          responseID = created
          return { type: "frame", frame }
        }
-        // Keepalives carry no response state and may arrive before response.created.
-        if (event.type === "keepalive") return { type: "frame", frame }
+        // Keepalives and provider notifications carry no response state and may precede response.created.
+        if (!event.type.startsWith("response.")) return { type: "frame", frame }
        if (!responseID)
          return yield* ProviderShared.eventError(
            options.id,
@@ -42,6 +42,7 @@ const canonical = (value: unknown): string => {
  if (Array.isArray(value)) return `[${value.map(canonical).join(",")}]`
  if (!ProviderShared.isRecord(value)) return ProviderShared.encodeJson(value)
  return `{${Object.keys(value)
+    .filter((key) => value[key] !== undefined)
    .sort()
    .map((key) => `${ProviderShared.encodeJson(key)}:${canonical(value[key])}`)
    .join(",")}}`
@@ -57,7 +58,12 @@ const comparable = (value: unknown) => {
  if (value.type === "message" && value.role === "assistant")
    return {
      role: "assistant",
-      content: value.content,
+      // Annotations and logprobs describe the response, not the text replayed in model input.
+      content: Array.isArray(value.content)
+        ? value.content.map((part) =>
+            ProviderShared.isRecord(part) && part.type === "output_text" ? { type: part.type, text: part.text } : part,
+          )
+        : value.content,
      ...(value.phase === undefined ? {} : { phase: value.phase }),
    }
  if (value.type === "function_call")
@@ -121,7 +127,7 @@ const rejected = (

 export const driver = (input: DriverInput): WebSocketChannelDriver => {
  const { previous_response_id: _previousResponseID, ...request } = input.request
-  let output: unknown[] = []
+  let output: OpenResponses.StreamItem[] = []
  return {
    create: (checkpoint) =>
      Effect.sync(() => {
@@ -149,6 +155,12 @@ export const driver = (input: DriverInput): WebSocketChannelDriver => {
          if (rejection === "websocket_connection_limit_reached") return rejected(observation, "rotate-and-retry-full")
        }
        if (observation.type !== "completed") return observation
+        // A trigger installs a different context window. Clear the append baseline, retaining the socket.
+        if (
+          Array.isArray(request.input) &&
+          request.input.some((item) => ProviderShared.isRecord(item) && item.type === "compaction_trigger")
+        )
+          return observation
        const responseID = event.response?.id
        if (!responseID || responseID.trim().length === 0) return observation
        return {
@@ -159,7 +171,14 @@ export const driver = (input: DriverInput): WebSocketChannelDriver => {
              version: VERSION,
              responseID,
              request,
-              output: event.response?.output ? [...event.response.output] : output.slice(),
+              // Completion can re-encrypt reasoning. Callers replay the item already emitted by output_item.done.
+              output: event.response?.output
+                ? event.response.output.map((item) =>
+                    item.type === "reasoning" && item.id !== undefined
+                      ? (output.find((done) => done.type === item.type && done.id === item.id) ?? item)
+                      : item,
+                  )
+                : output.slice(),
            } satisfies CheckpointValue,
          },
        }
@@ -189,6 +189,7 @@ export const InputItem = Schema.Union([
    id: Schema.optionalKey(Schema.String),
    call_id: Schema.String,
    name: Schema.String,
+    namespace: Schema.optional(Schema.String),
    arguments: Schema.String,
  }),
  Schema.Struct({
@@ -315,6 +316,7 @@ export const StreamItem = Schema.StructWithRest(
    id: Schema.optional(Schema.String),
    call_id: Schema.optional(Schema.String),
    name: Schema.optional(Schema.String),
+    namespace: Schema.optional(Schema.String),
    arguments: Schema.optional(Schema.String),
    encrypted_content: optionalNull(Schema.String),
  }),
@@ -433,7 +435,6 @@ export interface ParserState {
 type ReasoningSummaryStatus = "active" | "can-conclude" | "concluded"

 interface ReasoningStreamItem {
-  readonly open: boolean
  readonly encryptedContent: string | null | undefined
  // Keyed by the wire protocol's numeric `summary_index`. JS object keys coerce to
  // strings, but typing the map as `Record<number, ...>` documents intent
@@ -489,6 +490,7 @@ const lowerToolCall = (part: ToolCallPart, providerMetadataKey: string): OpenRes
    ...(id === undefined ? {} : { id }),
    call_id: part.id,
    name: part.name,
+    namespace: part.namespace,
    arguments: ProviderShared.encodeJson(part.input),
  }
 }
@@ -808,14 +810,15 @@ export const fromRequestWithAdapter = Effect.fn("OpenResponses.fromRequestWithAd
  request: LLMRequest,
  adapter: ProviderAdapter,
 ) {
+  const projected = ProviderShared.flattenToolRequest(request)
  const toolSchemaCompatibility = request.model.compatibility?.toolSchema
  return {
-    ...(yield* lowerConversation(request, adapter)),
+    ...(yield* lowerConversation(projected.request, adapter)),
    ...lowerGeneration(request),
    tools:
-      request.tools.length === 0
+      projected.tools.length === 0
        ? undefined
-        : yield* Effect.forEach(request.tools, (tool) =>
+        : yield* Effect.forEach(projected.tools, (tool) =>
            lowerTool(
              adapter.name,
              tool,
@@ -920,7 +923,7 @@ const joinReasoningText = (parts: ReadonlyArray<string | undefined>) => {
  return parts.filter((part) => part !== undefined).join("\n\n")
 }

-const outputItemID = (state: ParserState, event: Event) =>
+const outputItemID = (state: Pick<ParserState, "outputItems">, event: Event) =>
  event.output_index === undefined ? event.item_id : (state.outputItems[event.output_index] ?? event.item_id)

 const ITEM_ID_PREFIX: Readonly<Record<string, string>> = {
@@ -932,7 +935,11 @@ const ITEM_ID_PREFIX: Readonly<Record<string, string>> = {

 // An item without an id adopts the id already open in its output slot,
 // otherwise it gets a locally minted one.
-const resolveItem = (state: ParserState, item: StreamItem, index: number | undefined): OutputItem => ({
+const resolveItem = (
+  state: Pick<ParserState, "outputItems">,
+  item: StreamItem,
+  index: number | undefined,
+): OutputItem => ({
  ...item,
  id:
    item.id ??
@@ -942,7 +949,7 @@ const resolveItem = (state: ParserState, item: StreamItem, index: number | undef

 // Registered output slots are authoritative for `item_id` routing, and items
 // are resolved here so everything downstream can rely on `item.id`.
-export const normalize = (state: ParserState, input: Event): NormalizedEvent => ({
+export const normalize = (state: Pick<ParserState, "outputItems">, input: Event): NormalizedEvent => ({
  ...input,
  item_id: input.item_id === undefined ? undefined : outputItemID(state, input),
  item: input.item ? resolveItem(state, input.item, input.output_index) : input.item,
@@ -950,7 +957,7 @@ export const normalize = (state: ParserState, input: Event): NormalizedEvent =>

 const startReasoningSummaryPart = (state: ParserState, itemID: string, index: number): StepResult => {
  const item = state.reasoningItems[itemID]
-  if (!item?.open || index === 0 || item.summaryParts[index] !== undefined) return [state, NO_EVENTS]
+  if (!item || index === 0 || item.summaryParts[index] !== undefined) return [state, NO_EVENTS]

  const events: LLMEvent[] = []
  const lifecycle = Object.entries(item.summaryParts)
@@ -990,7 +997,7 @@ const startReasoningSummaryPart = (state: ParserState, itemID: string, index: nu

 export const onReasoningDelta = (state: ParserState, event: Event, itemID: string): StepResult => {
  const item = state.reasoningItems[itemID]
-  if (!event.delta || !item?.open) return [state, NO_EVENTS]
+  if (!event.delta || !item) return [state, NO_EVENTS]
  const index = event.summary_index ?? 0
  if (item.summaryParts[index] === "concluded") return [state, NO_EVENTS]
  const [started, emitted] = startReasoningSummaryPart(state, itemID, index)
@@ -1015,7 +1022,7 @@ export const onReasoningDelta = (state: ParserState, event: Event, itemID: strin
 // as a single delta unless that summary index already streamed one.
 export const onReasoningDone = (state: ParserState, event: Event, itemID: string): StepResult => {
  const item = state.reasoningItems[itemID]
-  if (!item?.open || typeof event.text !== "string") return [state, NO_EVENTS]
+  if (!item || typeof event.text !== "string") return [state, NO_EVENTS]
  const index = event.summary_index ?? 0
  if (item.deltaIndexes.has(index)) return [state, NO_EVENTS]
  return onReasoningDelta(state, { ...event, delta: event.text }, itemID)
@@ -1074,7 +1081,6 @@ const onOutputItemAdded = (state: ParserState, event: NormalizedEvent): StepResu
        reasoningItems: {
          ...state.reasoningItems,
          [item.id]: {
-            open: true,
            encryptedContent: item.encrypted_content,
            summaryParts: { 0: "active" },
            deltaIndexes: new Set(),
@@ -1096,11 +1102,20 @@ const onOutputItemAdded = (state: ParserState, event: NormalizedEvent): StepResu
      tools: ToolStream.start(state.tools, item.id, {
        id: item.call_id,
        name: item.name ?? "",
+        namespace: item.namespace,
        input: item.arguments ?? "",
        providerMetadata: metadata,
      }),
    },
-    [...events, LLMEvent.toolInputStart({ id: item.call_id, name: item.name ?? "", providerMetadata: metadata })],
+    [
+      ...events,
+      LLMEvent.toolInputStart({
+        id: item.call_id,
+        name: item.name ?? "",
+        namespace: item.namespace,
+        providerMetadata: metadata,
+      }),
+    ],
  ]
 }

@@ -1112,7 +1127,7 @@ const onReasoningSummaryPartAdded = (state: ParserState, event: Event): StepResu
 const onReasoningSummaryPartDone = (state: ParserState, event: Event): StepResult => {
  if (event.item_id === undefined || event.summary_index === undefined) return [state, NO_EVENTS]
  const item = state.reasoningItems[event.item_id]
-  if (!item?.open) return [state, NO_EVENTS]
+  if (!item) return [state, NO_EVENTS]
  if (item.summaryParts[event.summary_index] !== "active") return [state, NO_EVENTS]
  return [
    {
@@ -1219,7 +1234,12 @@ const onOutputItemDone = Effect.fn("OpenResponses.onOutputItemDone")(function* (
    const registered = state.tools[item.id] !== undefined
    const tools = registered
      ? state.tools
-      : ToolStream.start(state.tools, item.id, { id: item.call_id, name: item.name, providerMetadata: metadata })
+      : ToolStream.start(state.tools, item.id, {
+          id: item.call_id,
+          name: item.name,
+          namespace: item.namespace,
+          providerMetadata: metadata,
+        })
    const result =
      item.arguments === undefined
        ? yield* ToolStream.finish(state.id, tools, item.id)
@@ -1230,7 +1250,15 @@ const onOutputItemDone = Effect.fn("OpenResponses.onOutputItemDone")(function* (
    const resultEvents =
      registered || finished.length === 0
        ? finished
-        : [LLMEvent.toolInputStart({ id: item.call_id, name: item.name, providerMetadata: metadata }), ...finished]
+        : [
+            LLMEvent.toolInputStart({
+              id: item.call_id,
+              name: item.name,
+              namespace: item.namespace,
+              providerMetadata: metadata,
+            }),
+            ...finished,
+          ]
    const lifecycle = resultEvents.length ? Lifecycle.stepStart(state.lifecycle, events) : state.lifecycle
    events.push(...resultEvents)
    return [
@@ -1247,7 +1275,6 @@ const onOutputItemDone = Effect.fn("OpenResponses.onOutputItemDone")(function* (
  }

  if (item.type === "reasoning") {
-    if (state.reasoningItems[item.id]?.open === false) return [state, NO_EVENTS] satisfies StepResult
    const metadata = reasoningMetadata(state, item)
    const summaryParts: ReadonlyArray<unknown> = Array.isArray(item.summary) ? item.summary : []
    const summary: Array<string | undefined> = []
@@ -1274,53 +1301,14 @@ const onOutputItemDone = Effect.fn("OpenResponses.onOutputItemDone")(function* (
        const finalText = fragments.length === 1 ? itemText : summary[Number(index)]
        lifecycle = Lifecycle.reasoningEnd(lifecycle, events, `${item.id}:${index}`, metadata, finalText || undefined)
      }
-      return [
-        {
-          ...state,
-          lifecycle,
-          reasoningItems: {
-            ...state.reasoningItems,
-            [item.id]: {
-              ...reasoningItem,
-              open: false,
-              encryptedContent: item.encrypted_content ?? reasoningItem.encryptedContent,
-            },
-          },
-        },
-        events,
-      ] satisfies StepResult
+      const reasoningItems = { ...state.reasoningItems }
+      delete reasoningItems[item.id]
+      return [{ ...state, lifecycle, reasoningItems }, events] satisfies StepResult
    }
-    if (!state.lifecycle.reasoning.has(item.id)) {
-      const lifecycle = Lifecycle.stepStart(state.lifecycle, events)
-      events.push(LLMEvent.reasoningStart({ id: item.id, providerMetadata: metadata }))
-      events.push(
-        LLMEvent.reasoningEnd({
-          id: item.id,
-          providerMetadata: metadata,
-          text: itemText,
-        }),
-      )
-      return [
-        {
-          ...state,
-          lifecycle,
-          reasoningItems: {
-            ...state.reasoningItems,
-            [item.id]: {
-              open: false,
-              encryptedContent: item.encrypted_content,
-              summaryParts: { 0: "concluded" },
-              deltaIndexes: new Set(),
-            },
-          },
-        },
-        events,
-      ] satisfies StepResult
-    }
-    return [
-      { ...state, lifecycle: Lifecycle.reasoningEnd(state.lifecycle, events, item.id, metadata) },
-      events,
-    ] satisfies StepResult
+    const lifecycle = Lifecycle.stepStart(state.lifecycle, events)
+    events.push(LLMEvent.reasoningStart({ id: item.id, providerMetadata: metadata }))
+    events.push(LLMEvent.reasoningEnd({ id: item.id, providerMetadata: metadata, text: itemText }))
+    return [{ ...state, lifecycle }, events] satisfies StepResult
  }

  return [state, NO_EVENTS] satisfies StepResult
@@ -736,6 +736,7 @@ export const fromRequest = Effect.fn("OpenAIChat.fromRequest")(function* (
    )
  const generation = request.generation
  const toolSchemaCompatibility = request.model.compatibility?.toolSchema
+  const flattened = ProviderShared.flattenToolRequest(request)
  const provider = String(request.model.provider)
  const baseURL = request.model.route.endpoint.baseURL
  const detectedMaxTokensField = detectMaxTokensField(provider, baseURL)
@@ -748,16 +749,16 @@ export const fromRequest = Effect.fn("OpenAIChat.fromRequest")(function* (
  const zaiToolStream =
    request.model.compatibility?.zaiToolStream ?? detectZaiToolStream(provider, baseURL, request.model.id)
  const hasHistory = hasToolHistory(request.messages)
-  const hasActiveTools = request.tools.length > 0
+  const hasActiveTools = flattened.tools.length > 0
  return {
    model: request.model.id,
-    messages: yield* lowerMessages(request, options),
+    messages: yield* lowerMessages(flattened.request, options),
    tools:
-      request.tools.length === 0
+      flattened.tools.length === 0
        ? hasHistory
          ? []
          : undefined
-        : request.tools.map((tool) =>
+        : flattened.tools.map((tool) =>
            lowerTool(
              tool,
              ToolSchemaProjection.modelCompatibility(tool.inputSchema, toolSchemaCompatibility),
@@ -5,7 +5,7 @@ import { Auth } from "../route/auth.js"
 import { Endpoint } from "../route/endpoint.js"
 import { Protocol } from "../route/protocol.js"
 import { HttpTransport } from "../route/transport/index.js"
-import type { LLMRequest, JsonSchema, ToolDefinition } from "../schema/index.js"
+import { LLMRequest, mergeJsonRecords, type JsonSchema, type ToolDefinition, type ToolEntry } from "../schema/index.js"
 import { OpenResponses } from "./open-responses.js"
 import { JsonObject, optionalArray, optionalNull, ProviderShared } from "./shared.js"
 import { OpenAIImage } from "./utils/openai-image.js"
@@ -13,6 +13,7 @@ import { ResponsesHostedTools } from "./utils/responses-hosted-tools.js"
 import { ToolSchemaProjection } from "./utils/tool-schema.js"
 import { OpenResponsesChannel } from "./open-responses-channel.js"
 import { ResponsesCompaction } from "./utils/responses-compaction.js"
+import { ResponsesCheckpoint } from "./utils/responses-checkpoint.js"

 const ADAPTER = "openai-responses"
 const NAME = "OpenAI Responses"
@@ -75,7 +76,18 @@ const OpenAIResponsesHostedToolItem = Schema.Union([
  ),
 ])

-const OpenAIResponsesTools = Schema.Union([OpenResponses.Tool, OpenAIResponsesImageGenerationTool])
+const OpenAIResponsesNamespace = Schema.Struct({
+  type: Schema.tag("namespace"),
+  name: Schema.String,
+  description: Schema.String,
+  tools: Schema.Array(OpenResponses.Tool),
+})
+
+const OpenAIResponsesTools = Schema.Union([
+  OpenResponses.Tool,
+  OpenAIResponsesNamespace,
+  OpenAIResponsesImageGenerationTool,
+])

 const OpenAIResponsesToolChoice = Schema.Union([
  OpenResponses.ToolChoice,
@@ -103,6 +115,18 @@ const OpenAIResponsesBody = Schema.Struct({
 })
 export type OpenAIResponsesBody = Schema.Schema.Type<typeof OpenAIResponsesBody>

+/** Request control, never conversation content. */
+export const CompactionTrigger = Schema.Struct({ type: Schema.Literal("compaction_trigger") })
+const CheckpointBody = Schema.Struct({
+  ...OpenAIResponsesBody.fields,
+  input: Schema.Array(Schema.Union([OpenResponses.InputItem, OpenAIResponsesHostedToolItem, CompactionTrigger])),
+  store: Schema.Literal(false),
+  prompt_cache_retention: optionalNull(Schema.String),
+  prompt_cache_options: optionalNull(
+    Schema.Struct({ mode: Schema.optional(Schema.String), ttl: Schema.optional(Schema.String) }),
+  ),
+})
+
 const adapter = {
  id: ADAPTER,
  name: NAME,
@@ -128,13 +152,33 @@ const lowerTool = Effect.fn("OpenAIResponses.lowerTool")(function* (tool: ToolDe
  return yield* OpenResponses.lowerTool(NAME, tool, inputSchema)
 })

-const lowerToolChoice = (toolChoice: NonNullable<LLMRequest["toolChoice"]>, tools: ReadonlyArray<ToolDefinition>) =>
+// Native namespaces hold only function tools, so deeper levels flatten into
+// the leaf names the same way non-native protocols flatten the whole tree.
+const lowerToolEntry = Effect.fn("OpenAIResponses.lowerToolEntry")(function* (
+  tool: ToolEntry,
+  compatibility: Parameters<typeof ToolSchemaProjection.modelCompatibility>[1],
+) {
+  if (tool.type === "tool")
+    return yield* lowerTool(tool, ToolSchemaProjection.modelCompatibility(tool.inputSchema, compatibility))
+  // OpenAI requires a namespace description; fall back to a generic one so a
+  // missing description never blocks the request.
+  return {
+    type: "namespace" as const,
+    name: tool.name,
+    description: tool.description ?? `Tools in the ${tool.name} namespace.`,
+    tools: yield* Effect.forEach(ProviderShared.flattenTools(tool.tools), (leaf) =>
+      OpenResponses.lowerTool(NAME, leaf, ToolSchemaProjection.modelCompatibility(leaf.inputSchema, compatibility)),
+    ),
+  }
+})
+
+const lowerToolChoice = (toolChoice: NonNullable<LLMRequest["toolChoice"]>, tools: ReadonlyArray<ToolEntry>) =>
  ProviderShared.matchToolChoice(NAME, toolChoice, {
    auto: () => "auto" as const,
    none: () => "none" as const,
    required: () => "required" as const,
    tool: (name) =>
-      tools.some((tool) => tool.name === name && nativeImageTool(tool) !== undefined)
+      tools.some((tool) => tool.type === "tool" && tool.name === name && nativeImageTool(tool) !== undefined)
        ? ({ type: "image_generation" } as const)
        : { type: "function" as const, name },
  })
@@ -153,15 +197,42 @@ const fromRequest = Effect.fn("OpenAIResponses.fromRequest")(function* (request:
    tools:
      request.tools.length === 0
        ? undefined
-        : yield* Effect.forEach(request.tools, (tool) =>
-            lowerTool(tool, ToolSchemaProjection.modelCompatibility(tool.inputSchema, toolSchemaCompatibility)),
-          ),
+        : yield* Effect.forEach(request.tools, (tool) => lowerToolEntry(tool, toolSchemaCompatibility)),
    tool_choice:
      OpenResponses.allowedToolChoice(request) ??
      (request.toolChoice ? yield* lowerToolChoice(request.toolChoice, request.tools) : undefined),
  })
 })

+const checkpointBody = {
+  schema: CheckpointBody,
+  from: Effect.fn("OpenAIResponses.checkpointBody")(function* (request: LLMRequest) {
+    const native = yield* fromRequest(LLMRequest.update(request, { toolChoice: undefined }))
+    const overlay = request.http?.body
+    // Complete history is required for stateless replay and SSE recovery. Raw input overrides bypass that contract.
+    if (
+      overlay?.input !== undefined ||
+      overlay?.previous_response_id !== undefined ||
+      overlay?.conversation !== undefined
+    )
+      return yield* ProviderShared.invalidRequest(
+        "Trigger compaction requires complete canonical history, not an input or continuation override",
+      )
+    return yield* ProviderShared.validateWith(Schema.decodeUnknownEffect(CheckpointBody))({
+      ...mergeJsonRecords(native, overlay),
+      input: [...native.input, { type: "compaction_trigger" }],
+      stream: true,
+      store: false,
+      parallel_tool_calls: true,
+      tool_choice: undefined,
+      context_management: undefined,
+      text: undefined,
+      max_output_tokens: undefined,
+      max_tool_calls: undefined,
+    })
+  }),
+}
+
 const hostedToolResult = Effect.fn("OpenAIResponses.hostedToolResult")(function* (item: ResponsesHostedTools.Item) {
  const isError = item.error !== undefined && item.error !== null
  if (item.type === "image_generation_call" && item.result) {
@@ -239,7 +310,7 @@ export const transport = channelTransport({
 })

 export const route = Route.make({
-  compact: ResponsesCompaction.make(adapter),
+  compact: { endpoint: ResponsesCompaction.make(adapter), trigger: ResponsesCheckpoint.make(checkpointBody) },
  id: ADAPTER,
  provider: "openai",
  providerMetadataKey: "openai",
@@ -9,11 +9,14 @@ import {
  UnsupportedOperationError,
  AIError,
  HttpContext,
+  LLMRequest,
+  Message,
+  ToolDefinition,
  type ContentPart,
-  type LLMRequest,
  type MediaPart,
  type ProviderID,
  type TextPart,
+  type ToolEntry,
  type ToolResultPart,
 } from "../schema/index.js"
 import { isRecord } from "../utils/record.js"
@@ -46,6 +49,7 @@ export const promptCacheKey = (request: LLMRequest): string | undefined => {
 export interface ToolAccumulator {
  readonly id: string
  readonly name: string
+  readonly namespace?: string
  readonly input: string
 }

@@ -279,6 +283,38 @@ export const unsupportedOperation = (input: {
    }),
  })

+/**
+ * Lower namespaces to flat definitions for protocols without a native
+ * namespace construct. Leaf names join their namespace path with `_` because
+ * `.` is not broadly accepted in provider tool names.
+ */
+export const flattenTools = (tools: ReadonlyArray<ToolEntry>, path: ReadonlyArray<string> = []) => {
+  const flat = tools.flatMap((tool): ReadonlyArray<ToolDefinition> => {
+    if (tool.type === "namespace") return flattenTools(tool.tools, [...path, tool.name])
+    if (path.length === 0) return [tool]
+    return [new ToolDefinition({ ...tool, name: [...path, tool.name].join("_") })]
+  })
+  return [...new Map(flat.map((tool) => [tool.name, tool])).values()]
+}
+
+export const flattenToolRequest = (request: LLMRequest) => {
+  const messages = request.messages.map((message) => {
+    const content = message.content.map((part) => {
+      if ((part.type !== "tool-call" && part.type !== "tool-result") || part.namespace === undefined) return part
+      return { ...part, name: `${part.namespace}_${part.name}`, namespace: undefined }
+    })
+    return content.every((part, index) => part === message.content[index])
+      ? message
+      : new Message({ ...message, content })
+  })
+  return {
+    tools: flattenTools(request.tools),
+    request: messages.every((message, index) => message === request.messages[index])
+      ? request
+      : LLMRequest.update(request, { messages }),
+  }
+}
+
 export const imageResponse = Effect.fn("ProviderShared.imageResponse")(function* (
  route: string,
  name: string,
@@ -2,13 +2,12 @@ import { AwsV4Signer } from "aws4fetch"
 import { Effect } from "effect"
 import { Headers } from "effect/unstable/http"
 import { Auth, type AuthInput } from "../../route/auth.js"
+import { AIError, AuthenticationError } from "../../schema/index.js"
 import { ProviderShared } from "../shared.js"

 /**
 * AWS credentials for SigV4 signing. Bedrock also supports Bearer API key auth,
- * which provider facades configure as route auth instead of SigV4. STS-vended
- * credentials should be refreshed by the consumer (rebuild the model) before
- * they expire; the route does not refresh.
+ * which provider facades configure as route auth instead of SigV4.
 */
 export interface Credentials {
  readonly region: string
@@ -17,6 +16,44 @@ export interface Credentials {
  readonly sessionToken?: string
 }

+/** Static credentials or an effect resolved before every request. */
+export type CredentialSource = Credentials | Effect.Effect<Credentials, AIError>
+
+export interface DefaultChainOptions {
+  readonly region: string
+  /** Shared config profile passed to the AWS default chain. */
+  readonly profile?: string
+}
+
+/**
+ * Resolve credentials through the AWS default provider chain: environment
+ * variables, shared config and SSO caches, web identity tokens, process
+ * credentials, and container or instance metadata. A fresh chain runs on every
+ * request so credentials rotated on disk without an expiration (for example
+ * shared-config keys rewritten by a corporate SSO tool) are always re-read;
+ * the SDK's own memoization would otherwise pin them for the process lifetime.
+ */
+export const defaultChain = (options: DefaultChainOptions): Effect.Effect<Credentials, AIError> =>
+  Effect.tryPromise({
+    try: async () => {
+      const { fromNodeProviderChain } = await import("@aws-sdk/credential-providers")
+      const identity = await fromNodeProviderChain(options.profile === undefined ? {} : { profile: options.profile })()
+      return {
+        region: options.region,
+        accessKeyId: identity.accessKeyId,
+        secretAccessKey: identity.secretAccessKey,
+        ...(identity.sessionToken === undefined ? {} : { sessionToken: identity.sessionToken }),
+      }
+    },
+    catch: (error) =>
+      new AIError({
+        reason: new AuthenticationError({
+          message: `AWS default credential chain failed: ${ProviderShared.errorText(error)}`,
+          cause: error,
+        }),
+      }),
+  })
+
 const signRequest = (input: {
  readonly url: string
  readonly body: string
@@ -48,16 +85,17 @@ const signRequest = (input: {

 /** Sign the exact JSON bytes with SigV4 using credentials configured on the route. */
 export const sigV4 = (
-  credentials: Credentials | undefined,
+  source: CredentialSource | undefined,
  options: { readonly service?: string; readonly name?: string } = {},
 ) =>
  Auth.custom((input: AuthInput) => {
    return Effect.gen(function* () {
-      if (!credentials) {
+      if (!source) {
        return yield* ProviderShared.invalidRequest(
          `${options.name ?? "Bedrock Converse"} requires either route bearer auth or AWS credentials configured on the route`,
        )
      }
+      const credentials = Effect.isEffect(source) ? yield* source : source
      const headersForSigning = Headers.set(input.headers, "content-type", "application/json")
      const signed = yield* signRequest({
        url: input.url,
@@ -74,4 +112,35 @@ export const sigV4 = (
 /** Bedrock route auth defaults to SigV4 and expects credentials from route configuration. */
 export const auth = sigV4(undefined)

+export const resolveRegion = (input: {
+  readonly region?: string
+  readonly credentials?: { readonly region: string }
+}) =>
+  input.region ?? input.credentials?.region ?? process.env.AWS_REGION ?? process.env.AWS_DEFAULT_REGION ?? "us-east-1"
+
+export interface ResolveAuthInput {
+  readonly apiKey?: string
+  readonly credentials?: Credentials
+  readonly profile?: string
+}
+
+export interface ResolveAuthOptions {
+  readonly service?: string
+  readonly name?: string
+  /** `sigv4` ignores an ambient `AWS_BEARER_TOKEN_BEDROCK`; `bearer` is validated by the caller. */
+  readonly mode?: "bearer" | "sigv4"
+}
+
+/**
+ * Bearer wins over SigV4 and explicit static credentials win over the default
+ * chain, matching the AWS SDK's own precedence for `AWS_BEARER_TOKEN_BEDROCK`.
+ * The region is applied to the SigV4 scope so it always matches the endpoint host.
+ */
+export const resolveAuth = (input: ResolveAuthInput, region: string, options: ResolveAuthOptions = {}) => {
+  const apiKey = options.mode === "sigv4" ? undefined : (input.apiKey ?? process.env.AWS_BEARER_TOKEN_BEDROCK)
+  if (apiKey !== undefined) return Auth.bearer(apiKey)
+  if (input.credentials !== undefined) return sigV4({ ...input.credentials, region }, options)
+  return sigV4(defaultChain({ region, profile: input.profile }), options)
+}
+
 export * as BedrockAuth from "./bedrock-auth.js"
@@ -1,6 +1,6 @@
 import { Schema } from "effect"
 import type { CacheHint } from "../../schema/index.js"
-import { newBreakpoints, ttlBucket, type Breakpoints } from "./cache.js"
+import { newBreakpoints, ttlBucket } from "./cache.js"

 // Bedrock cache markers are positional: emit a `cachePoint` block immediately
 // after the content the caller wants treated as a cacheable prefix. Bedrock
@@ -13,24 +13,46 @@ export const CachePointBlock = Schema.Struct({
 })
 export type CachePointBlock = Schema.Schema.Type<typeof CachePointBlock>

-// Callers pass a shared counter through every `block()` call site so the
-// four-breakpoint budget is respected across `system`, `messages`, and `tools`.
+const LEGACY_CLAUDE = ["anthropic.claude-instant", "anthropic.claude-v1", "anthropic.claude-v2", "anthropic.claude-3-"]
+
+// These legacy Claude releases support explicit caching, but only for five minutes.
+const CLAUDE_5M = [
+  "anthropic.claude-3-5-sonnet-20241022-v2:0",
+  "anthropic.claude-3-5-haiku-20241022-v1:0",
+  "anthropic.claude-3-7-sonnet-20250219-v1:0",
+  "anthropic.claude-sonnet-4-20250514-v1:0",
+  "anthropic.claude-opus-4-20250514-v1:0",
+  "anthropic.claude-opus-4-1-20250805-v1:0",
+]
+
+// Callers share the four-breakpoint budget across system, messages, and tools.
 export const BEDROCK_BREAKPOINT_CAP = 4

-export type { Breakpoints } from "./cache.js"
-export const breakpoints = () => newBreakpoints(BEDROCK_BREAKPOINT_CAP)
+export const breakpoints = (modelID: string) => {
+  // Substring matching also handles regional prefixes and model-bearing ARNs.
+  const short = CLAUDE_5M.some((id) => modelID.includes(id))
+  return {
+    ...newBreakpoints(BEDROCK_BREAKPOINT_CAP),
+    // Assume modern Claude releases retain caching support; older generations need an explicit exception.
+    // Other model families use implicit caching where available.
+    supported: modelID.includes("anthropic.claude-") && (short || !LEGACY_CLAUDE.some((id) => modelID.includes(id))),
+    ttl1h: !short,
+  }
+}
+export type Breakpoints = ReturnType<typeof breakpoints>

 const DEFAULT_5M: CachePointBlock = { cachePoint: { type: "default" } }
 const DEFAULT_1H: CachePointBlock = { cachePoint: { type: "default", ttl: "1h" } }

 export const block = (breakpoints: Breakpoints, cache: CacheHint | undefined): CachePointBlock | undefined => {
+  if (!breakpoints.supported) return undefined
  if (cache?.type !== "ephemeral" && cache?.type !== "persistent") return undefined
  if (breakpoints.remaining <= 0) {
    breakpoints.dropped += 1
    return undefined
  }
  breakpoints.remaining -= 1
-  return ttlBucket(cache.ttlSeconds) === "1h" ? DEFAULT_1H : DEFAULT_5M
+  return breakpoints.ttl1h && ttlBucket(cache.ttlSeconds) === "1h" ? DEFAULT_1H : DEFAULT_5M
 }

 export * as BedrockCache from "./bedrock-cache.js"
@@ -0,0 +1,122 @@
+import { Effect, Schema, Stream } from "effect"
+import { Route, type RouteBody, type TriggerCompactOperation } from "../../route/client.js"
+import { Protocol } from "../../route/protocol.js"
+import { CompactionCheckpointResponse, HttpOptions, LLMEvent, LLMRequest } from "../../schema/index.js"
+import { OpenResponses } from "../open-responses.js"
+import { ProviderShared } from "../shared.js"
+
+interface State {
+  readonly parser: Pick<OpenResponses.ParserState, "id" | "provider" | "outputItems">
+  readonly checkpoints: Readonly<Record<string, CompactionCheckpointResponse["checkpoint"]>>
+  readonly responseID?: string
+}
+
+const onOutputItem = Effect.fn("ResponsesCheckpoint.onOutputItem")(function* (
+  state: State,
+  input: OpenResponses.Event,
+) {
+  const event = OpenResponses.normalize(state.parser, input)
+  const item = event.item
+  if (!item) return state
+  const parser =
+    event.output_index === undefined || state.parser.outputItems[event.output_index] === item.id
+      ? state.parser
+      : { ...state.parser, outputItems: { ...state.parser.outputItems, [event.output_index]: item.id } }
+  const next = parser === state.parser ? state : { ...state, parser }
+  if (event.type === "response.output_item.added" || item.type !== "compaction") return next
+  if (
+    event.output_index !== undefined &&
+    Object.entries(state.parser.outputItems).some(
+      ([index, id]) => id === item.id && Number(index) !== event.output_index,
+    )
+  )
+    return yield* ProviderShared.eventError(parser.id, "Compaction checkpoint appeared in multiple output slots")
+  if (!item.encrypted_content)
+    return yield* ProviderShared.eventError(parser.id, "Compaction output is missing its encrypted content")
+  const previous = state.checkpoints[item.id]
+  if (previous && previous.encrypted !== item.encrypted_content)
+    return yield* ProviderShared.eventError(parser.id, "Compaction output changed after completion")
+  if (previous) return next
+  return {
+    ...next,
+    checkpoints: {
+      ...state.checkpoints,
+      [item.id]: { type: "compaction", provider: parser.provider, id: item.id, encrypted: item.encrypted_content },
+    },
+  } satisfies State
+})
+
+/** Collect a trigger response before acknowledging transport completion. No generation output escapes. */
+export const make = <Body>(body: RouteBody<Body>): TriggerCompactOperation =>
+  Effect.fn("ResponsesCheckpoint.execute")(function* (request, executor, options) {
+    const source = request.model.route
+    let result: CompactionCheckpointResponse | undefined
+    // Route registries erase the frame type. The codec validates that boundary before parsing.
+    const event: Schema.Codec<OpenResponses.Event, unknown> = OpenResponses.protocol.stream.event
+    const protocol = Protocol.make({
+      id: source.protocol,
+      body,
+      stream: {
+        event,
+        initial: (request: LLMRequest): State => ({
+          parser: { id: source.id, provider: request.model.provider, outputItems: {} },
+          checkpoints: {},
+        }),
+        terminal: OpenResponses.terminal,
+        step: Effect.fn("ResponsesCheckpoint.step")(function* (state: State, event: OpenResponses.Event) {
+          if (event.response?.id && state.responseID && event.response.id !== state.responseID)
+            return yield* ProviderShared.eventError(source.id, "Compaction response ID changed during execution")
+          if (event.type === "response.created") return [{ ...state, responseID: event.response?.id }, []] as const
+          if (event.type === "error" || event.type === "response.failed")
+            return yield* OpenResponses.providerFailure(event, "Compaction request failed")
+          if (event.type === "response.incomplete")
+            return yield* ProviderShared.eventError(source.id, "Compaction response was incomplete")
+          if (event.type === "response.output_item.added" || event.type === "response.output_item.done")
+            return [yield* onOutputItem(state, event), []] as const
+          if (event.type !== "response.completed") return [state, []] as const
+          const responseID = event.response?.id
+          if (!responseID?.trim())
+            return yield* ProviderShared.eventError(source.id, "Compaction response is missing its response ID")
+          if (event.response?.status !== undefined && event.response.status !== "completed")
+            return yield* ProviderShared.eventError(source.id, "Compaction response did not complete successfully")
+          let next = state
+          for (const [index, item] of (event.response?.output ?? []).entries()) {
+            next = yield* onOutputItem(next, { type: "response.output_item.done", output_index: index, item })
+          }
+          const checkpoints = Object.values(next.checkpoints)
+          const checkpoint = checkpoints[0]
+          if (checkpoints.length !== 1 || !checkpoint)
+            return yield* ProviderShared.eventError(
+              source.id,
+              "Compaction response must contain exactly one checkpoint",
+            )
+          result = new CompactionCheckpointResponse({
+            checkpoint,
+            responseID,
+            usage: OpenResponses.mapUsage(event.response?.usage, OpenResponses.metadataKey(request.model)),
+          })
+          return [next, [LLMEvent.finish({ reason: { normalized: "stop" } })]] as const
+        }),
+      },
+    })
+    const route = Route.make({
+      id: source.id,
+      provider: source.provider,
+      providerMetadataKey: source.providerMetadataKey,
+      protocol,
+      endpoint: source.endpoint,
+      auth: source.auth,
+      transport: source.transport,
+    })
+    const native = yield* body.from(request)
+    // The body builder already applied and validated overlays. Do not let transport reapply them.
+    const preparedRequest = LLMRequest.update(request, {
+      http: request.http === undefined ? undefined : new HttpOptions({ ...request.http, body: undefined }),
+    })
+    const prepared = yield* route.prepareTransport(native, preparedRequest, options)
+    yield* route.streamPrepared(prepared, preparedRequest, { http: executor }, options).pipe(Stream.runDrain)
+    if (!result) return yield* ProviderShared.eventError(source.id, "Compaction response ended without a checkpoint")
+    return result
+  })
+
+export * as ResponsesCheckpoint from "./responses-checkpoint.js"
@@ -55,6 +55,7 @@ const inputStart = (tool: PendingTool) =>
  LLMEvent.toolInputStart({
    id: tool.id,
    name: tool.name,
+    namespace: tool.namespace,
    providerExecuted: tool.providerExecuted ? true : undefined,
    providerMetadata: tool.providerMetadata,
  })
@@ -63,6 +64,7 @@ const inputDelta = (tool: PendingTool, text: string) =>
  LLMEvent.toolInputDelta({
    id: tool.id,
    name: tool.name,
+    namespace: tool.namespace,
    text,
    input: Option.getOrElse(parsePartialInput(tool.input), () => ({})),
  })
@@ -85,6 +87,7 @@ const toolCall = (route: string, tool: PendingTool, inputOverride?: string) => {
        LLMEvent.toolCall({
          id: tool.id,
          name: tool.name,
+          namespace: tool.namespace,
          input,
          providerExecuted: tool.providerExecuted ? true : undefined,
          providerMetadata: tool.providerMetadata,
@@ -94,7 +97,12 @@ const toolCall = (route: string, tool: PendingTool, inputOverride?: string) => {
 }

 const finishEvents = (tool: PendingTool, event: ToolCall): ReadonlyArray<LLMEvent> => [
-  LLMEvent.toolInputEnd({ id: tool.id, name: tool.name, providerMetadata: tool.providerMetadata }),
+  LLMEvent.toolInputEnd({
+    id: tool.id,
+    name: tool.name,
+    namespace: tool.namespace,
+    providerMetadata: tool.providerMetadata,
+  }),
  event,
 ]

@@ -150,6 +158,7 @@ export const appendOrStart = <K extends StreamKey>(
  const tool = {
    id,
    name,
+    namespace: current?.namespace,
    input: `${current?.input ?? ""}${delta.text}`,
    providerExecuted: current?.providerExecuted,
    providerMetadata: current?.providerMetadata,
@@ -159,6 +168,17 @@ export const appendOrStart = <K extends StreamKey>(
  return appendTool(tools, key, tool, delta.text)
 }

+/**
+ * Append argument text to a started tool. Returns `undefined` when no tool is
+ * open under `key`, for protocols that ignore deltas without a matching block.
+ */
+export const append = <K extends StreamKey>(tools: State<K>, key: K, text: string): AppendOutcome<K> | undefined => {
+  const current = tools[key]
+  if (!current) return undefined
+  if (text.length === 0) return { tools, tool: current, events: [] }
+  return appendTool(tools, key, { ...current, input: `${current.input}${text}` }, text)
+}
+
 /**
 * Append argument text to a tool that must already have been started. This keeps
 * protocols honest when their stream grammar promises a start event before any
@@ -170,12 +190,7 @@ export const appendExisting = <K extends StreamKey>(
  key: K,
  text: string,
  missingToolMessage: string,
-): AppendOutcome<K> | AIError => {
-  const current = tools[key]
-  if (!current) return eventError(route, missingToolMessage)
-  if (text.length === 0) return { tools, tool: current, events: [] }
-  return appendTool(tools, key, { ...current, input: `${current.input}${text}` }, text)
-}
+): AppendOutcome<K> | AIError => append(tools, key, text) ?? eventError(route, missingToolMessage)

 /**
 * Finalize one pending tool call: parse the accumulated raw JSON, remove it
@@ -1,5 +1,5 @@
 import type { LanguageModel, ProviderOptions } from "./schema/index.js"
-import type { CompactOperation } from "./route/client.js"
+import type { CompactionOperations } from "./route/client.js"

 export interface Settings extends Readonly<Record<string, unknown>> {
  readonly baseURL?: string
@@ -10,7 +10,7 @@ export interface Settings extends Readonly<Record<string, unknown>> {
 export interface Definition<
  ProviderSettings extends Settings = Settings,
  Options extends ProviderOptions = ProviderOptions,
-  Compact extends CompactOperation | undefined = CompactOperation | undefined,
+  Compact extends CompactionOperations | undefined = CompactionOperations | undefined,
 > {
  readonly model: (modelID: string, settings: ProviderSettings) => LanguageModel<Options, Compact>
 }
@@ -1,4 +1,3 @@
-import { Auth } from "../route/auth.js"
 import { Route, type RouteDefaultsInput } from "../route/client.js"
 import type { ProviderPackage } from "../provider-package.js"
 import { OpenAIChat } from "../protocols/openai-chat.js"
@@ -10,9 +9,15 @@ import { withOpenAIOptions, type OpenAIProviderOptionsInput } from "./openai-opt
 export const id = ProviderID.make("amazon-bedrock")

 export type Config = RouteDefaultsInput & {
+  /** Bedrock API key. Falls back to `AWS_BEARER_TOKEN_BEDROCK`; bearer auth takes precedence over SigV4. */
  readonly apiKey?: string
+  /** `sigv4` ignores `apiKey` fallbacks from the environment; `bearer` requires a token. */
+  readonly auth?: "bearer" | "sigv4"
  readonly baseURL?: string
+  /** Static SigV4 credentials. When omitted the AWS default credential chain resolves them per request. */
  readonly credentials?: Credentials
+  /** Shared config profile for the default credential chain. */
+  readonly profile?: string
  readonly region?: string
  readonly providerOptions?: OpenAIProviderOptionsInput
 }
@@ -22,6 +27,7 @@ export interface Settings extends ProviderPackage.Settings {
  readonly auth?: "bearer" | "sigv4"
  readonly baseURL?: string
  readonly credentials?: Credentials
+  readonly profile?: string
  readonly region?: string
  readonly topP?: number
  readonly providerOptions?: OpenAIProviderOptionsInput
@@ -47,23 +53,35 @@ const chatRoute = OpenAIChat.route.with({
 export const routes = [responsesRoute, chatRoute]

 const configuredRoute = <Body, Prepared>(route: Route<Body, Prepared>, input: Config) => {
-  const region = input.region ?? input.credentials?.region ?? "us-east-1"
-  const credentials = input.credentials === undefined ? undefined : { ...input.credentials, region }
+  const region = BedrockAuth.resolveRegion(input)
  return route.with({
    endpoint: { baseURL: input.baseURL ?? `https://bedrock-mantle.${region}.api.aws/v1` },
-    auth:
-      input.apiKey === undefined
-        ? BedrockAuth.sigV4(credentials, { service: "bedrock-mantle", name: "Bedrock Mantle" })
-        : Auth.bearer(input.apiKey),
+    auth: BedrockAuth.resolveAuth(input, region, {
+      service: "bedrock-mantle",
+      name: "Bedrock Mantle",
+      mode: input.auth,
+    }),
  })
 }

 const defaults = (input: Config) => {
-  const { apiKey: _, baseURL: _baseURL, credentials: _credentials, region: _region, ...rest } = input
+  const {
+    apiKey: _,
+    auth: _auth,
+    baseURL: _baseURL,
+    credentials: _credentials,
+    profile: _profile,
+    region: _region,
+    ...rest
+  } = input
  return rest
 }

 export const configure = (input: Config = {}) => {
+  if (input.auth === "bearer" && input.apiKey === undefined && process.env.AWS_BEARER_TOKEN_BEDROCK === undefined)
+    throw new Error("Amazon Bedrock Mantle bearer auth requires apiKey")
+  if (input.auth === "sigv4" && input.apiKey !== undefined)
+    throw new Error("Amazon Bedrock Mantle SigV4 auth does not accept apiKey")
  const configuredResponsesRoute = configuredRoute(responsesRoute, input)
  const configuredChatRoute = configuredRoute(chatRoute, input)
  const modelDefaults = defaults(input)
@@ -87,29 +105,26 @@ export const configure = (input: Config = {}) => {

 export const provider = configure()

-const config = (settings: Settings): Config => {
-  if (settings.auth === "bearer" && settings.apiKey === undefined)
-    throw new Error("Amazon Bedrock Mantle bearer auth requires apiKey")
-  if (settings.auth === "sigv4" && settings.apiKey !== undefined)
-    throw new Error("Amazon Bedrock Mantle SigV4 auth does not accept apiKey")
-  return {
-    apiKey: settings.auth === "sigv4" ? undefined : settings.apiKey,
+const fromSettings = (settings: Settings) =>
+  configure({
+    apiKey: settings.apiKey,
+    auth: settings.auth,
    baseURL: settings.baseURL,
    credentials: settings.credentials,
    generation: settings.topP === undefined ? undefined : { topP: settings.topP },
    headers: settings.headers === undefined ? undefined : { ...settings.headers },
    http: settings.body === undefined ? undefined : { body: { ...settings.body } },
+    profile: settings.profile,
    providerOptions: settings.providerOptions,
    region: settings.region,
-  }
-}
+  })

 export const chatModel: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (
  modelID,
  settings,
-) => configure(config(settings)).chat(modelID)
+) => fromSettings(settings).chat(modelID)
 export const responsesModel: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (
  modelID,
  settings,
-) => configure(config(settings)).responses(modelID)
+) => fromSettings(settings).responses(modelID)
 export const model = responsesModel
@@ -1,17 +1,23 @@
 import type { RouteDefaultsInput } from "../route/client.js"
-import { Auth } from "../route/auth.js"
 import type { ProviderPackage } from "../provider-package.js"
 import { ProviderID, type ModelID } from "../schema/index.js"
 import * as BedrockConverse from "../protocols/bedrock-converse.js"
 import type { BedrockCredentials } from "../protocols/bedrock-converse.js"
+import { BedrockAuth } from "../protocols/utils/bedrock-auth.js"

 export const id = ProviderID.make("amazon-bedrock")

 export type Config = RouteDefaultsInput & {
+  /** Bedrock API key. Falls back to `AWS_BEARER_TOKEN_BEDROCK`; bearer auth takes precedence over SigV4. */
  readonly apiKey?: string
+  /** `sigv4` ignores `apiKey` fallbacks from the environment; `bearer` requires a token. */
+  readonly auth?: "bearer" | "sigv4"
  readonly headers?: Record<string, string>
+  /** Static SigV4 credentials. When omitted the AWS default credential chain resolves them per request. */
  readonly credentials?: BedrockCredentials
-  /** AWS region. Defaults to `us-east-1` when neither this nor `credentials.region` is set. */
+  /** Shared config profile for the default credential chain. */
+  readonly profile?: string
+  /** AWS region. Falls back to `credentials.region`, `AWS_REGION`, `AWS_DEFAULT_REGION`, then `us-east-1`. */
  readonly region?: string
  /** Override the computed `https://bedrock-runtime.<region>.amazonaws.com` URL. */
  readonly baseURL?: string
@@ -22,6 +28,7 @@ export interface Settings extends ProviderPackage.Settings {
  readonly auth?: "bearer" | "sigv4"
  readonly baseURL?: string
  readonly credentials?: BedrockCredentials
+  readonly profile?: string
  readonly region?: string
  readonly topP?: number
 }
@@ -30,14 +37,17 @@ export const routes = [BedrockConverse.route]
 const bedrockBaseURL = (region: string) => `https://bedrock-runtime.${region}.amazonaws.com`

 const configuredRoute = (input: Config) => {
-  const { apiKey, credentials, region, baseURL, ...rest } = input
-  const resolvedRegion = region ?? credentials?.region ?? "us-east-1"
+  const { apiKey, auth, credentials, profile, region, baseURL, ...rest } = input
+  if (auth === "bearer" && apiKey === undefined && process.env.AWS_BEARER_TOKEN_BEDROCK === undefined)
+    throw new Error("Amazon Bedrock bearer auth requires apiKey")
+  if (auth === "sigv4" && apiKey !== undefined) throw new Error("Amazon Bedrock SigV4 auth does not accept apiKey")
+  const resolvedRegion = BedrockAuth.resolveRegion(input)
  return BedrockConverse.route.with({
    ...rest,
    provider: id,
    providerMetadataKey: "bedrock",
    endpoint: { baseURL: baseURL ?? bedrockBaseURL(resolvedRegion) },
-    auth: apiKey === undefined ? BedrockConverse.sigV4Auth(credentials) : Auth.bearer(apiKey),
+    auth: BedrockAuth.resolveAuth({ apiKey, credentials, profile }, resolvedRegion, { mode: auth }),
  })
 }

@@ -51,18 +61,15 @@ export const configure = (input: Config = {}) => {
 }

 export const provider = configure()
-export const model: ProviderPackage.Definition<Settings>["model"] = (modelID, settings) => {
-  if (settings.auth === "bearer" && settings.apiKey === undefined)
-    throw new Error("Amazon Bedrock bearer auth requires apiKey")
-  if (settings.auth === "sigv4" && settings.apiKey !== undefined)
-    throw new Error("Amazon Bedrock SigV4 auth does not accept apiKey")
-  return configure({
-    apiKey: settings.auth === "sigv4" ? undefined : settings.apiKey,
+export const model: ProviderPackage.Definition<Settings>["model"] = (modelID, settings) =>
+  configure({
+    apiKey: settings.apiKey,
+    auth: settings.auth,
    baseURL: settings.baseURL,
    credentials: settings.credentials,
    generation: settings.topP === undefined ? undefined : { topP: settings.topP },
    headers: settings.headers === undefined ? undefined : { ...settings.headers },
    http: settings.body === undefined ? undefined : { body: { ...settings.body } },
+    profile: settings.profile,
    region: settings.region,
  }).model(modelID)
-}
@@ -1,7 +1,7 @@
 import { Headers } from "effect/unstable/http"
 import { Auth } from "../route/auth.js"
 import { type AtLeastOne, type ProviderAuthOption } from "../route/auth-options.js"
-import type { Route, RouteDefaultsInput, CompactOperation } from "../route/client.js"
+import type { Route, RouteDefaultsInput, CompactionOperations } from "../route/client.js"
 import type { ProviderPackage } from "../provider-package.js"
 import { ProviderID, type ModelID } from "../schema/index.js"
 import * as OpenAIChat from "../protocols/openai-chat.js"
@@ -39,6 +39,7 @@ export type Settings = ProviderPackage.Settings &
 const resourceBaseURL = (resourceName: string) => `https://${resourceName.trim()}.openai.azure.com/openai`

 const responsesRoute = OpenAIResponses.route.with({
+  compact: { endpoint: OpenAIResponses.route.compact.endpoint },
  id: "azure-openai-responses",
  provider: id,
  auth: routeAuth,
@@ -102,7 +103,7 @@ const auth = (input: Config) => {
  )
 }

-const configuredRoute = <Body, Prepared, Compact extends CompactOperation | undefined>(
+const configuredRoute = <Body, Prepared, Compact extends CompactionOperations | undefined>(
  route: Route<Body, Prepared, Compact>,
  input: Config,
  modelID: string | ModelID,
@@ -168,7 +169,7 @@ const config = (settings: Settings): Config => {
 export const responsesModel: ProviderPackage.Definition<
  Settings,
  OpenAIProviderOptionsInput,
-  CompactOperation
+  typeof responsesRoute.compact
 >["model"] = (modelID, settings) => configure(config(settings)).responses(modelID)
 export const chatModel: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (
  modelID,
@@ -0,0 +1,60 @@
+import type { ProviderPackage } from "../provider-package.js"
+import { OpenAIChat } from "../protocols/openai-chat.js"
+import { AuthOptions, type ProviderAuthOption } from "../route/auth-options.js"
+import { Route, type RouteDefaultsInput } from "../route/client.js"
+import { Endpoint } from "../route/endpoint.js"
+import { ProviderID, type ModelID } from "../schema/index.js"
+import type { OpenAIProviderOptionsInput } from "./openai-options.js"
+
+export const id = ProviderID.make("baseten")
+const baseURL = "https://inference.baseten.co/v1"
+
+export type LanguageModelOptions = Omit<RouteDefaultsInput, "providerOptions"> &
+  ProviderAuthOption<"optional"> & {
+    readonly baseURL?: string
+    readonly providerOptions?: OpenAIProviderOptionsInput
+  }
+
+export interface Settings extends ProviderPackage.Settings {
+  readonly apiKey?: string
+  readonly baseURL?: string
+  readonly providerOptions?: OpenAIProviderOptionsInput
+}
+
+export const route = Route.make({
+  id: "baseten-chat",
+  provider: id,
+  providerMetadataKey: "baseten",
+  protocol: OpenAIChat.protocol,
+  endpoint: Endpoint.path("/chat/completions", { baseURL }),
+  framing: OpenAIChat.framing,
+})
+
+export const routes = [route]
+
+export const configure = (input: LanguageModelOptions = {}) => {
+  const { apiKey: _apiKey, auth: _auth, baseURL: endpoint, ...defaults } = input
+  const configured = route.with({
+    ...defaults,
+    endpoint: { baseURL: endpoint ?? baseURL },
+    auth: AuthOptions.bearer(input, "BASETEN_API_KEY"),
+  })
+  return {
+    id,
+    model: (modelID: string | ModelID) => configured.model<OpenAIProviderOptionsInput>({ id: modelID }),
+    configure,
+  }
+}
+
+export const provider = configure()
+
+export const model: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (modelID, settings) =>
+  configure({
+    apiKey: settings.apiKey,
+    baseURL: settings.baseURL,
+    headers: settings.headers === undefined ? undefined : { ...settings.headers },
+    http: settings.body === undefined ? undefined : { body: { ...settings.body } },
+    providerOptions: settings.providerOptions,
+  }).model(modelID)
+
+export * as Baseten from "./baseten.js"
@@ -1,12 +1,13 @@
 import type { ProviderPackage } from "../provider-package.js"
-import { OpenAICompatibleChat } from "../protocols/openai-compatible-chat.js"
+import { OpenAIChat } from "../protocols/openai-chat.js"
 import { AuthOptions, type ProviderAuthOption } from "../route/auth-options.js"
-import type { RouteDefaultsInput } from "../route/client.js"
+import { Route, type RouteDefaultsInput } from "../route/client.js"
+import { Endpoint } from "../route/endpoint.js"
 import { ProviderID, type ModelID } from "../schema/index.js"
-import { profiles } from "./openai-compatible-profile.js"
 import type { OpenAIProviderOptionsInput } from "./openai-options.js"

 export const id = ProviderID.make("cerebras")
+const baseURL = "https://api.cerebras.ai/v1"

 export type LanguageModelOptions = Omit<RouteDefaultsInput, "providerOptions"> &
  ProviderAuthOption<"optional"> & {
@@ -20,19 +21,22 @@ export interface Settings extends ProviderPackage.Settings {
  readonly providerOptions?: OpenAIProviderOptionsInput
 }

-export const route = OpenAICompatibleChat.route.with({
+export const route = Route.make({
  id: "cerebras-chat",
  provider: id,
-  endpoint: { baseURL: profiles.cerebras.baseURL },
+  providerMetadataKey: "cerebras",
+  protocol: OpenAIChat.protocol,
+  endpoint: Endpoint.path("/chat/completions", { baseURL }),
+  framing: OpenAIChat.framing,
 })

 export const routes = [route]

 export const configure = (input: LanguageModelOptions = {}) => {
-  const { apiKey: _apiKey, auth: _auth, baseURL, ...defaults } = input
+  const { apiKey: _apiKey, auth: _auth, baseURL: endpoint, ...defaults } = input
  const configured = route.with({
    ...defaults,
-    endpoint: { baseURL: baseURL ?? profiles.cerebras.baseURL },
+    endpoint: { baseURL: endpoint ?? baseURL },
    auth: AuthOptions.bearer(input, "CEREBRAS_API_KEY"),
  })
  return {
@@ -0,0 +1,98 @@
+import type { Config, Redacted } from "effect"
+import type { ProviderPackage } from "../provider-package.js"
+import { OpenAIChat } from "../protocols/openai-chat.js"
+import { Auth } from "../route/auth.js"
+import type { AtLeastOne, ProviderAuthOption } from "../route/auth-options.js"
+import { Route, type RouteDefaultsInput } from "../route/client.js"
+import { Endpoint } from "../route/endpoint.js"
+import { ProviderID, type ModelID } from "../schema/index.js"
+import type { OpenAIProviderOptionsInput } from "./openai-options.js"
+
+export const id = ProviderID.make("cloudflare-ai-gateway")
+export const authEnvVars = ["CLOUDFLARE_API_TOKEN", "CF_AIG_TOKEN"] as const
+
+type GatewayURL = AtLeastOne<{
+  readonly accountId: string
+  readonly baseURL: string
+}> & {
+  readonly gatewayId?: string
+}
+
+export type LanguageModelOptions = GatewayURL &
+  Omit<RouteDefaultsInput, "providerOptions"> &
+  ProviderAuthOption<"optional"> & {
+    /** Cloudflare AI Gateway authentication token. Sent as `cf-aig-authorization`. */
+    readonly gatewayApiKey?: string | Redacted.Redacted | Config.Config<string | Redacted.Redacted>
+    readonly providerOptions?: OpenAIProviderOptionsInput
+  }
+
+export type Settings = ProviderPackage.Settings &
+  GatewayURL & {
+    readonly apiKey?: string
+    readonly gatewayApiKey?: string
+    readonly providerOptions?: OpenAIProviderOptionsInput
+  }
+
+export const baseURL = (input: GatewayURL) => {
+  if (input.baseURL) return input.baseURL
+  if (!input.accountId) throw new Error("CloudflareAIGateway.configure requires accountId unless baseURL is supplied")
+  return `https://gateway.ai.cloudflare.com/v1/${encodeURIComponent(input.accountId)}/${encodeURIComponent(input.gatewayId?.trim() || "default")}/compat`
+}
+
+const auth = (input: LanguageModelOptions) => {
+  if ("auth" in input && input.auth) return input.auth
+  const gateway = Auth.optional(input.gatewayApiKey, "gatewayApiKey")
+    .orElse(Auth.config(authEnvVars[0]))
+    .orElse(Auth.config(authEnvVars[1]))
+    .pipe(Auth.bearerHeader("cf-aig-authorization"))
+  if (!("apiKey" in input) || input.apiKey === undefined) return gateway
+  if (input.gatewayApiKey === undefined) return Auth.bearer(input.apiKey)
+  return Auth.bearerHeader("cf-aig-authorization", input.gatewayApiKey).andThen(Auth.bearer(input.apiKey))
+}
+
+export const route = Route.make({
+  id: "cloudflare-ai-gateway",
+  provider: id,
+  providerMetadataKey: "cloudflare-ai-gateway",
+  protocol: OpenAIChat.protocol,
+  endpoint: Endpoint.path("/chat/completions"),
+  framing: OpenAIChat.framing,
+})
+
+export const routes = [route]
+
+export const configure = (input: LanguageModelOptions) => {
+  const {
+    accountId: _accountId,
+    gatewayId: _gatewayId,
+    apiKey: _apiKey,
+    gatewayApiKey: _gatewayApiKey,
+    baseURL: _baseURL,
+    auth: _auth,
+    ...defaults
+  } = input
+  const configured = route.with({
+    ...defaults,
+    endpoint: { baseURL: baseURL(input) },
+    auth: auth(input),
+  })
+  return {
+    id,
+    model: (modelID: string | ModelID) => configured.model<OpenAIProviderOptionsInput>({ id: modelID }),
+    configure,
+  }
+}
+
+export const provider = { id, configure }
+
+export const model: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (modelID, settings) =>
+  configure({
+    apiKey: settings.apiKey,
+    gatewayApiKey: settings.gatewayApiKey,
+    baseURL: baseURL(settings),
+    headers: settings.headers === undefined ? undefined : { ...settings.headers },
+    http: settings.body === undefined ? undefined : { body: { ...settings.body } },
+    providerOptions: settings.providerOptions,
+  }).model(modelID)
+
+export * as CloudflareAIGateway from "./cloudflare-ai-gateway.js"
@@ -0,0 +1,71 @@
+import type { ProviderPackage } from "../provider-package.js"
+import { OpenAIChat } from "../protocols/openai-chat.js"
+import { AuthOptions, type AtLeastOne, type ProviderAuthOption } from "../route/auth-options.js"
+import { Route, type RouteDefaultsInput } from "../route/client.js"
+import { Endpoint } from "../route/endpoint.js"
+import { ProviderID, type ModelID } from "../schema/index.js"
+import type { OpenAIProviderOptionsInput } from "./openai-options.js"
+
+export const id = ProviderID.make("cloudflare-workers-ai")
+export const authEnvVars = ["CLOUDFLARE_API_KEY", "CLOUDFLARE_WORKERS_AI_TOKEN"] as const
+
+type WorkersAIURL = AtLeastOne<{
+  readonly accountId: string
+  readonly baseURL: string
+}>
+
+export type LanguageModelOptions = WorkersAIURL &
+  Omit<RouteDefaultsInput, "providerOptions"> &
+  ProviderAuthOption<"optional"> & {
+    readonly providerOptions?: OpenAIProviderOptionsInput
+  }
+
+export type Settings = ProviderPackage.Settings &
+  WorkersAIURL & {
+    readonly apiKey?: string
+    readonly providerOptions?: OpenAIProviderOptionsInput
+  }
+
+export const baseURL = (input: WorkersAIURL) => {
+  if (input.baseURL) return input.baseURL
+  if (!input.accountId) throw new Error("CloudflareWorkersAI.configure requires accountId unless baseURL is supplied")
+  return `https://api.cloudflare.com/client/v4/accounts/${encodeURIComponent(input.accountId)}/ai/v1`
+}
+
+export const route = Route.make({
+  id: "cloudflare-workers-ai",
+  provider: id,
+  providerMetadataKey: "cloudflare-workers-ai",
+  protocol: OpenAIChat.protocol,
+  endpoint: Endpoint.path("/chat/completions"),
+  framing: OpenAIChat.framing,
+})
+
+export const routes = [route]
+
+export const configure = (input: LanguageModelOptions) => {
+  const { accountId: _accountId, apiKey: _apiKey, auth: _auth, baseURL: _baseURL, ...defaults } = input
+  const configured = route.with({
+    ...defaults,
+    endpoint: { baseURL: baseURL(input) },
+    auth: AuthOptions.bearer(input, authEnvVars),
+  })
+  return {
+    id,
+    model: (modelID: string | ModelID) => configured.model<OpenAIProviderOptionsInput>({ id: modelID }),
+    configure,
+  }
+}
+
+export const provider = { id, configure }
+
+export const model: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (modelID, settings) =>
+  configure({
+    apiKey: settings.apiKey,
+    baseURL: baseURL(settings),
+    headers: settings.headers === undefined ? undefined : { ...settings.headers },
+    http: settings.body === undefined ? undefined : { body: { ...settings.body } },
+    providerOptions: settings.providerOptions,
+  }).model(modelID)
+
+export * as CloudflareWorkersAI from "./cloudflare-workers-ai.js"
@@ -1,133 +0,0 @@
-import type { Config, Redacted } from "effect"
-import * as OpenAICompatibleChat from "../protocols/openai-compatible-chat.js"
-import { Auth } from "../route/auth.js"
-import { AuthOptions, type AtLeastOne, type ProviderAuthOption } from "../route/auth-options.js"
-import type { RouteDefaultsInput } from "../route/client.js"
-import { ProviderID, type ModelID } from "../schema/index.js"
-import type { OpenAIProviderOptionsInput } from "./openai-options.js"
-
-export const aiGatewayID = ProviderID.make("cloudflare-ai-gateway")
-export const workersAIID = ProviderID.make("cloudflare-workers-ai")
-export const aiGatewayAuthEnvVars = ["CLOUDFLARE_API_TOKEN", "CF_AIG_TOKEN"] as const
-export const workersAIAuthEnvVars = ["CLOUDFLARE_API_KEY", "CLOUDFLARE_WORKERS_AI_TOKEN"] as const
-
-type CloudflareSecret = string | Redacted.Redacted | Config.Config<string | Redacted.Redacted>
-
-type GatewayURL = AtLeastOne<{
-  readonly accountId: string
-  readonly baseURL: string
-}> & {
-  readonly gatewayId?: string
-}
-
-export type AIGatewayOptions = GatewayURL &
-  Omit<RouteDefaultsInput, "providerOptions"> &
-  ProviderAuthOption<"optional"> & {
-    /** Cloudflare AI Gateway authentication token. Sent as `cf-aig-authorization`. */
-    readonly gatewayApiKey?: CloudflareSecret
-    readonly providerOptions?: OpenAIProviderOptionsInput
-  }
-
-type WorkersAIURL = AtLeastOne<{
-  readonly accountId: string
-  readonly baseURL: string
-}>
-
-export type WorkersAIOptions = WorkersAIURL &
-  Omit<RouteDefaultsInput, "providerOptions"> &
-  ProviderAuthOption<"optional"> & {
-    readonly providerOptions?: OpenAIProviderOptionsInput
-  }
-
-export const aiGatewayBaseURL = (input: GatewayURL) => {
-  if (input.baseURL) return input.baseURL
-  if (!input.accountId) throw new Error("CloudflareAIGateway.configure requires accountId unless baseURL is supplied")
-  return `https://gateway.ai.cloudflare.com/v1/${encodeURIComponent(input.accountId)}/${encodeURIComponent(input.gatewayId?.trim() || "default")}/compat`
-}
-
-const aiGatewayAuth = (input: AIGatewayOptions) => {
-  if ("auth" in input && input.auth) return input.auth
-  const gateway = Auth.optional(input.gatewayApiKey, "gatewayApiKey")
-    .orElse(Auth.config("CLOUDFLARE_API_TOKEN"))
-    .orElse(Auth.config("CF_AIG_TOKEN"))
-    .pipe(Auth.bearerHeader("cf-aig-authorization"))
-  if (!("apiKey" in input) || input.apiKey === undefined) return gateway
-  if (input.gatewayApiKey === undefined) return Auth.bearer(input.apiKey)
-  return Auth.bearerHeader("cf-aig-authorization", input.gatewayApiKey).andThen(Auth.bearer(input.apiKey))
-}
-
-export const workersAIBaseURL = (input: WorkersAIURL) => {
-  if (input.baseURL) return input.baseURL
-  if (!input.accountId) throw new Error("CloudflareWorkersAI.configure requires accountId unless baseURL is supplied")
-  return `https://api.cloudflare.com/client/v4/accounts/${encodeURIComponent(input.accountId)}/ai/v1`
-}
-
-const workersAIAuth = (input: WorkersAIOptions) => {
-  return AuthOptions.bearer(input, workersAIAuthEnvVars)
-}
-
-export const aiGatewayRoute = OpenAICompatibleChat.route.with({
-  id: "cloudflare-ai-gateway",
-  provider: aiGatewayID,
-})
-
-export const workersAIRoute = OpenAICompatibleChat.route.with({
-  id: "cloudflare-workers-ai",
-  provider: workersAIID,
-})
-
-export const routes = [aiGatewayRoute, workersAIRoute]
-
-const aiGatewayDefaults = (options: AIGatewayOptions) => {
-  const {
-    accountId: _accountId,
-    gatewayId: _gatewayId,
-    apiKey: _apiKey,
-    gatewayApiKey: _gatewayApiKey,
-    baseURL: _baseURL,
-    auth: _auth,
-    ...rest
-  } = options
-  return rest
-}
-
-const workersAIDefaults = (options: WorkersAIOptions) => {
-  const { accountId: _accountId, apiKey: _apiKey, auth: _auth, baseURL: _baseURL, ...rest } = options
-  return rest
-}
-
-const configureAIGateway = (options: AIGatewayOptions) => {
-  const route = aiGatewayRoute.with({
-    ...aiGatewayDefaults(options),
-    endpoint: { baseURL: aiGatewayBaseURL(options) },
-    auth: aiGatewayAuth(options),
-  })
-  return {
-    id: aiGatewayID,
-    model: (modelID: string | ModelID) => route.model<OpenAIProviderOptionsInput>({ id: modelID }),
-    configure: configureAIGateway,
-  }
-}
-
-const configureWorkersAI = (options: WorkersAIOptions) => {
-  const route = workersAIRoute.with({
-    ...workersAIDefaults(options),
-    endpoint: { baseURL: workersAIBaseURL(options) },
-    auth: workersAIAuth(options),
-  })
-  return {
-    id: workersAIID,
-    model: (modelID: string | ModelID) => route.model<OpenAIProviderOptionsInput>({ id: modelID }),
-    configure: configureWorkersAI,
-  }
-}
-
-export const CloudflareAIGateway = {
-  id: aiGatewayID,
-  configure: configureAIGateway,
-}
-
-export const CloudflareWorkersAI = {
-  id: workersAIID,
-  configure: configureWorkersAI,
-}
@@ -1,12 +1,13 @@
 import type { ProviderPackage } from "../provider-package.js"
-import { OpenAICompatibleChat } from "../protocols/openai-compatible-chat.js"
+import { OpenAIChat } from "../protocols/openai-chat.js"
 import { AuthOptions, type ProviderAuthOption } from "../route/auth-options.js"
-import type { RouteDefaultsInput } from "../route/client.js"
+import { Route, type RouteDefaultsInput } from "../route/client.js"
+import { Endpoint } from "../route/endpoint.js"
 import { ProviderID, type ModelID } from "../schema/index.js"
-import { profiles } from "./openai-compatible-profile.js"
 import type { OpenAIProviderOptionsInput } from "./openai-options.js"

 export const id = ProviderID.make("deepinfra")
+const baseURL = "https://api.deepinfra.com/v1/openai"

 export type LanguageModelOptions = Omit<RouteDefaultsInput, "providerOptions"> &
  ProviderAuthOption<"optional"> & {
@@ -20,21 +21,24 @@ export interface Settings extends ProviderPackage.Settings {
  readonly providerOptions?: OpenAIProviderOptionsInput
 }

-export const route = OpenAICompatibleChat.route.with({
+export const route = Route.make({
  id: "deepinfra-chat",
  provider: id,
-  endpoint: { baseURL: profiles.deepinfra.baseURL },
+  providerMetadataKey: "deepinfra",
+  protocol: OpenAIChat.protocol,
+  endpoint: Endpoint.path("/chat/completions", { baseURL }),
+  framing: OpenAIChat.framing,
 })

 export const routes = [route]

 export const configure = (input: LanguageModelOptions = {}) => {
-  const { apiKey: _apiKey, auth: _auth, baseURL, ...defaults } = input
-  const root = baseURL?.replace(/\/+$/, "")
+  const { apiKey: _apiKey, auth: _auth, baseURL: endpoint, ...defaults } = input
+  const root = endpoint?.replace(/\/+$/, "")
  const configured = route.with({
    ...defaults,
    endpoint: {
-      baseURL: root === undefined ? profiles.deepinfra.baseURL : root.endsWith("/openai") ? root : `${root}/openai`,
+      baseURL: root === undefined ? baseURL : root.endsWith("/openai") ? root : `${root}/openai`,
    },
    auth: AuthOptions.bearer(input, "DEEPINFRA_API_KEY"),
  })
@@ -0,0 +1,64 @@
+import type { ProviderPackage } from "../provider-package.js"
+import { OpenAIChat } from "../protocols/openai-chat.js"
+import { AuthOptions, type ProviderAuthOption } from "../route/auth-options.js"
+import { Route, type RouteDefaultsInput } from "../route/client.js"
+import { Endpoint } from "../route/endpoint.js"
+import { ProviderID, type ModelID } from "../schema/index.js"
+import type { OpenAIProviderOptionsInput } from "./openai-options.js"
+
+export const id = ProviderID.make("deepseek")
+const baseURL = "https://api.deepseek.com/v1"
+
+export type LanguageModelOptions = Omit<RouteDefaultsInput, "providerOptions"> &
+  ProviderAuthOption<"optional"> & {
+    readonly baseURL?: string
+    readonly providerOptions?: OpenAIProviderOptionsInput
+  }
+
+export interface Settings extends ProviderPackage.Settings {
+  readonly apiKey?: string
+  readonly baseURL?: string
+  readonly providerOptions?: OpenAIProviderOptionsInput
+}
+
+export const route = Route.make({
+  id: "deepseek-chat",
+  provider: id,
+  providerMetadataKey: "deepseek",
+  protocol: OpenAIChat.protocol,
+  endpoint: Endpoint.path("/chat/completions", { baseURL }),
+  framing: OpenAIChat.framing,
+})
+
+export const routes = [route]
+
+export const configure = (input: LanguageModelOptions = {}) => {
+  const { apiKey: _apiKey, auth: _auth, baseURL: endpoint, ...defaults } = input
+  const configured = route.with({
+    ...defaults,
+    endpoint: { baseURL: endpoint ?? baseURL },
+    auth: AuthOptions.bearer(input, "DEEPSEEK_API_KEY"),
+  })
+  return {
+    id,
+    model: (modelID: string | ModelID) =>
+      configured.model<OpenAIProviderOptionsInput>({
+        id: modelID,
+        compatibility: { maxTokensField: "max_tokens", supportsStore: false },
+      }),
+    configure,
+  }
+}
+
+export const provider = configure()
+
+export const model: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (modelID, settings) =>
+  configure({
+    apiKey: settings.apiKey,
+    baseURL: settings.baseURL,
+    headers: settings.headers === undefined ? undefined : { ...settings.headers },
+    http: settings.body === undefined ? undefined : { body: { ...settings.body } },
+    providerOptions: settings.providerOptions,
+  }).model(modelID)
+
+export * as DeepSeek from "./deepseek.js"
@@ -0,0 +1,60 @@
+import type { ProviderPackage } from "../provider-package.js"
+import { OpenAIChat } from "../protocols/openai-chat.js"
+import { AuthOptions, type ProviderAuthOption } from "../route/auth-options.js"
+import { Route, type RouteDefaultsInput } from "../route/client.js"
+import { Endpoint } from "../route/endpoint.js"
+import { ProviderID, type ModelID } from "../schema/index.js"
+import type { OpenAIProviderOptionsInput } from "./openai-options.js"
+
+export const id = ProviderID.make("fireworks")
+const baseURL = "https://api.fireworks.ai/inference/v1"
+
+export type LanguageModelOptions = Omit<RouteDefaultsInput, "providerOptions"> &
+  ProviderAuthOption<"optional"> & {
+    readonly baseURL?: string
+    readonly providerOptions?: OpenAIProviderOptionsInput
+  }
+
+export interface Settings extends ProviderPackage.Settings {
+  readonly apiKey?: string
+  readonly baseURL?: string
+  readonly providerOptions?: OpenAIProviderOptionsInput
+}
+
+export const route = Route.make({
+  id: "fireworks-chat",
+  provider: id,
+  providerMetadataKey: "fireworks",
+  protocol: OpenAIChat.protocol,
+  endpoint: Endpoint.path("/chat/completions", { baseURL }),
+  framing: OpenAIChat.framing,
+})
+
+export const routes = [route]
+
+export const configure = (input: LanguageModelOptions = {}) => {
+  const { apiKey: _apiKey, auth: _auth, baseURL: endpoint, ...defaults } = input
+  const configured = route.with({
+    ...defaults,
+    endpoint: { baseURL: endpoint ?? baseURL },
+    auth: AuthOptions.bearer(input, "FIREWORKS_API_KEY"),
+  })
+  return {
+    id,
+    model: (modelID: string | ModelID) => configured.model<OpenAIProviderOptionsInput>({ id: modelID }),
+    configure,
+  }
+}
+
+export const provider = configure()
+
+export const model: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (modelID, settings) =>
+  configure({
+    apiKey: settings.apiKey,
+    baseURL: settings.baseURL,
+    headers: settings.headers === undefined ? undefined : { ...settings.headers },
+    http: settings.body === undefined ? undefined : { body: { ...settings.body } },
+    providerOptions: settings.providerOptions,
+  }).model(modelID)
+
+export * as Fireworks from "./fireworks.js"
@@ -1,6 +1,7 @@
 import type { ProviderPackage } from "../provider-package.js"
-import { OpenAICompatibleChat } from "../protocols/openai-compatible-chat.js"
-import type { RouteDefaultsInput } from "../route/client.js"
+import { OpenAIChat } from "../protocols/openai-chat.js"
+import { Route, type RouteDefaultsInput } from "../route/client.js"
+import { Endpoint } from "../route/endpoint.js"
 import { ProviderID, type ModelID } from "../schema/index.js"
 import { GoogleVertexShared } from "./google-vertex-shared.js"
 import type { OpenAIProviderOptionsInput } from "./openai-options.js"
@@ -24,10 +25,13 @@ export interface Settings extends ProviderPackage.Settings {
  readonly providerOptions?: OpenAIProviderOptionsInput
 }

-const route = OpenAICompatibleChat.route.with({
+const route = Route.make({
  id: "google-vertex-chat",
  provider: id,
  providerMetadataKey: "vertex",
+  protocol: OpenAIChat.protocol,
+  endpoint: Endpoint.path("/chat/completions"),
+  framing: OpenAIChat.framing,
 })

 export const routes = [route]
@@ -1,6 +1,7 @@
 import type { ProviderPackage } from "../provider-package.js"
-import { OpenAICompatibleResponses } from "../protocols/openai-compatible-responses.js"
-import type { RouteDefaultsInput } from "../route/client.js"
+import { OpenResponses } from "../protocols/open-responses.js"
+import { Route, type RouteDefaultsInput } from "../route/client.js"
+import { Endpoint } from "../route/endpoint.js"
 import { ProviderID, type ModelID } from "../schema/index.js"
 import { GoogleVertexShared } from "./google-vertex-shared.js"
 import type { OpenResponsesProviderOptionsInput } from "./open-responses-options.js"
@@ -24,11 +25,14 @@ export interface Settings extends ProviderPackage.Settings {
  readonly providerOptions?: OpenResponsesProviderOptionsInput
 }

-const route = OpenAICompatibleResponses.route.with({
+const route = Route.make({
  id: "google-vertex-responses",
  provider: id,
  providerMetadataKey: "vertex",
-  providerOptions: { store: false },
+  protocol: OpenResponses.protocol,
+  endpoint: Endpoint.path(OpenResponses.PATH),
+  transport: OpenResponses.httpTransport,
+  defaults: { providerOptions: { store: false, include: ["reasoning.encrypted_content"] } },
 })

 export const routes = [route]
@@ -7,10 +7,10 @@ import { Route, type RouteDefaultsInput } from "../route/client.js"
 import { Endpoint } from "../route/endpoint.js"
 import { Protocol } from "../route/protocol.js"
 import { ProviderID, type ModelID, type LLMRequest } from "../schema/index.js"
-import { profiles } from "./openai-compatible-profile.js"
 import type { OpenAIProviderOptionsInput } from "./openai-options.js"

 export const id = ProviderID.make("groq")
+const baseURL = "https://api.groq.com/openai/v1"

 export type ProviderOptions = Pick<OpenAIProviderOptionsInput, "reasoningEffort"> & {
  /** Controls visible reasoning on GPT-OSS; other models always use parsed reasoning. */
@@ -73,15 +73,15 @@ export const route = Route.make({
  provider: id,
  providerMetadataKey: "openai",
  protocol,
-  endpoint: Endpoint.path("/chat/completions", { baseURL: profiles.groq.baseURL }),
+  endpoint: Endpoint.path("/chat/completions", { baseURL }),
  framing: OpenAIChat.framing,
 })

 export const configure = (input: LanguageModelOptions = {}) => {
-  const { apiKey: _apiKey, auth: _auth, baseURL, ...defaults } = input
+  const { apiKey: _apiKey, auth: _auth, baseURL: endpoint, ...defaults } = input
  const configured = route.with({
    ...defaults,
-    endpoint: { baseURL: baseURL ?? profiles.groq.baseURL },
+    endpoint: { baseURL: endpoint ?? baseURL },
    auth: AuthOptions.bearer(input, "GROQ_API_KEY"),
  })
  return {
@@ -3,10 +3,13 @@ export * as AnthropicCompatible from "./anthropic-compatible.js"
 export * as AmazonBedrock from "./amazon-bedrock.js"
 export * as AmazonBedrockMantle from "./amazon-bedrock-mantle.js"
 export * as Azure from "./azure.js"
+export * as Baseten from "./baseten.js"
 export * as Cerebras from "./cerebras.js"
-export * as Cloudflare from "./cloudflare.js"
-export { CloudflareAIGateway, CloudflareWorkersAI } from "./cloudflare.js"
+export * as CloudflareAIGateway from "./cloudflare-ai-gateway.js"
+export * as CloudflareWorkersAI from "./cloudflare-workers-ai.js"
 export * as DeepInfra from "./deepinfra.js"
+export * as DeepSeek from "./deepseek.js"
+export * as Fireworks from "./fireworks.js"
 export * as Google from "./google.js"
 export * as GoogleVertex from "./google-vertex.js"
 export * as GoogleVertexChat from "./google-vertex-chat.js"
@@ -1,20 +0,0 @@
-export interface OpenAICompatibleProfile {
-  readonly provider: string
-  readonly baseURL: string
-}
-
-export const profiles = {
-  baseten: { provider: "baseten", baseURL: "https://inference.baseten.co/v1" },
-  cerebras: { provider: "cerebras", baseURL: "https://api.cerebras.ai/v1" },
-  deepinfra: { provider: "deepinfra", baseURL: "https://api.deepinfra.com/v1/openai" },
-  deepseek: { provider: "deepseek", baseURL: "https://api.deepseek.com/v1" },
-  fireworks: { provider: "fireworks", baseURL: "https://api.fireworks.ai/inference/v1" },
-  groq: { provider: "groq", baseURL: "https://api.groq.com/openai/v1" },
-  openrouter: { provider: "openrouter", baseURL: "https://openrouter.ai/api/v1" },
-  togetherai: { provider: "togetherai", baseURL: "https://api.together.xyz/v1" },
-  xai: { provider: "xai", baseURL: "https://api.x.ai/v1" },
-} as const satisfies Record<string, OpenAICompatibleProfile>
-
-export const byProvider: Record<string, OpenAICompatibleProfile> = Object.fromEntries(
-  Object.values(profiles).map((profile) => [profile.provider, profile]),
-)
@@ -1,9 +1,8 @@
 import { ProviderID, type ModelID } from "../schema/index.js"
-import * as OpenAICompatibleChat from "../protocols/openai-compatible-chat.js"
+import { OpenAICompatibleChat } from "../protocols/openai-compatible-chat.js"
 import type { RouteDefaultsInput } from "../route/client.js"
 import { AuthOptions, type ProviderAuthOption } from "../route/auth-options.js"
 import type { ProviderPackage } from "../provider-package.js"
-import { profiles, type OpenAICompatibleProfile } from "./openai-compatible-profile.js"
 import type { OpenAIProviderOptionsInput } from "./openai-options.js"

 export const id = ProviderID.make("openai-compatible")
@@ -22,12 +21,6 @@ export interface Settings extends ProviderPackage.Settings {
  readonly providerOptions?: OpenAIProviderOptionsInput
 }

-export type FamilyModelOptions = Omit<RouteDefaultsInput, "providerOptions"> &
-  ProviderAuthOption<"optional"> & {
-    readonly baseURL?: string
-    readonly providerOptions?: OpenAIProviderOptionsInput
-  }
-
 export const routes = [OpenAICompatibleChat.route]

 export const configure = (input: GenericModelOptions) => {
@@ -47,22 +40,6 @@ export const configure = (input: GenericModelOptions) => {
  }
 }

-const define = (profile: OpenAICompatibleProfile) => {
-  const configureProfile = (input: FamilyModelOptions = {}) => {
-    const facade = configure({
-      ...input,
-      baseURL: input.baseURL ?? profile.baseURL,
-      provider: profile.provider,
-    })
-    return {
-      id: ProviderID.make(profile.provider),
-      model: facade.model,
-      configure: configureProfile,
-    }
-  }
-  return configureProfile()
-}
-
 export const provider = {
  id,
  configure,
@@ -78,10 +55,4 @@ export const model: ProviderPackage.Definition<Settings, OpenAIProviderOptionsIn
    providerOptions: settings.providerOptions,
  }).model(modelID)

-export const baseten = define(profiles.baseten)
-export const cerebras = define(profiles.cerebras)
-export const deepinfra = define(profiles.deepinfra)
-export const deepseek = define(profiles.deepseek)
-export const fireworks = define(profiles.fireworks)
-export const groq = define(profiles.groq)
-export const togetherai = define(profiles.togetherai)
+export * as OpenAICompatible from "./openai-compatible.js"
@@ -1,5 +1,5 @@
 import { AuthOptions, type ProviderAuthOption } from "../route/auth-options.js"
-import type { Route, RouteDefaultsInput, CompactOperation } from "../route/client.js"
+import type { Route, RouteDefaultsInput, CompactionOperations } from "../route/client.js"
 import type { ProviderPackage } from "../provider-package.js"
 import { HttpOptions, ProviderID, ToolDefinition, mergeHttpOptions, type ModelID } from "../schema/index.js"
 import * as OpenAIChat from "../protocols/openai-chat.js"
@@ -73,7 +73,7 @@ const defaults = (input: Config) => {
  return rest
 }

-const configuredRoute = <Body, Prepared, Compact extends CompactOperation | undefined>(
+const configuredRoute = <Body, Prepared, Compact extends CompactionOperations | undefined>(
  route: Route<Body, Prepared, Compact>,
  input: Config,
 ) =>
@@ -132,10 +132,11 @@ const config = (settings: Settings): Config => {
  }
 }

-export const model: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput, CompactOperation>["model"] = (
-  modelID,
-  settings,
-) => {
+export const model: ProviderPackage.Definition<
+  Settings,
+  OpenAIProviderOptionsInput,
+  typeof OpenAIResponses.route.compact
+>["model"] = (modelID, settings) => {
  return configure(config(settings)).responses(modelID)
 }

@@ -5,13 +5,12 @@ import { Protocol } from "../route/protocol.js"
 import { AuthOptions, type ProviderAuthOption } from "../route/auth-options.js"
 import { ProviderID, type CacheHint, type ModelID } from "../schema/index.js"
 import type { ProviderPackage } from "../provider-package.js"
-import * as OpenAICompatibleProfiles from "./openai-compatible-profile.js"
-import * as OpenAIChat from "../protocols/openai-chat.js"
+import { OpenAIChat } from "../protocols/openai-chat.js"
 import { newBreakpoints, ttlBucket } from "../protocols/utils/cache.js"
 import { isRecord } from "../protocols/shared.js"

-export const profile = OpenAICompatibleProfiles.profiles.openrouter
-export const id = ProviderID.make(profile.provider)
+export const id = ProviderID.make("openrouter")
+const baseURL = "https://openrouter.ai/api/v1"
 const ADAPTER = "openrouter"

 type OpenRouterString<Known extends string> = Known | (string & {})
@@ -162,20 +161,20 @@ const bodyOptions = (input: unknown) => {

 export const route = Route.make({
  id: ADAPTER,
-  provider: profile.provider,
+  provider: id,
  providerMetadataKey: "openrouter",
  protocol,
-  endpoint: Endpoint.path("/chat/completions", { baseURL: profile.baseURL }),
+  endpoint: Endpoint.path("/chat/completions", { baseURL }),
  framing: OpenAIChat.framing,
 })

 export const routes = [route]

 const configuredRoute = (input: LanguageModelOptions) => {
-  const { apiKey: _, auth: _auth, baseURL, ...rest } = input
+  const { apiKey: _, auth: _auth, baseURL: endpoint, ...rest } = input
  return route.with({
    ...rest,
-    endpoint: { baseURL: baseURL ?? profile.baseURL },
+    endpoint: { baseURL: endpoint ?? baseURL },
    auth: AuthOptions.bearer(input, "OPENROUTER_API_KEY"),
  })
 }
@@ -1,12 +1,13 @@
 import type { ProviderPackage } from "../provider-package.js"
-import { OpenAICompatibleChat } from "../protocols/openai-compatible-chat.js"
+import { OpenAIChat } from "../protocols/openai-chat.js"
 import { AuthOptions, type ProviderAuthOption } from "../route/auth-options.js"
-import type { RouteDefaultsInput } from "../route/client.js"
+import { Route, type RouteDefaultsInput } from "../route/client.js"
+import { Endpoint } from "../route/endpoint.js"
 import { ProviderID, type ModelID } from "../schema/index.js"
-import { profiles } from "./openai-compatible-profile.js"
 import type { OpenAIProviderOptionsInput } from "./openai-options.js"

 export const id = ProviderID.make("togetherai")
+const baseURL = "https://api.together.xyz/v1"

 export type LanguageModelOptions = Omit<RouteDefaultsInput, "providerOptions"> &
  ProviderAuthOption<"optional"> & {
@@ -20,19 +21,22 @@ export interface Settings extends ProviderPackage.Settings {
  readonly providerOptions?: OpenAIProviderOptionsInput
 }

-export const route = OpenAICompatibleChat.route.with({
+export const route = Route.make({
  id: "togetherai-chat",
  provider: id,
-  endpoint: { baseURL: profiles.togetherai.baseURL },
+  providerMetadataKey: "togetherai",
+  protocol: OpenAIChat.protocol,
+  endpoint: Endpoint.path("/chat/completions", { baseURL }),
+  framing: OpenAIChat.framing,
 })

 export const routes = [route]

 export const configure = (input: LanguageModelOptions = {}) => {
-  const { apiKey: _apiKey, auth: _auth, baseURL, ...defaults } = input
+  const { apiKey: _apiKey, auth: _auth, baseURL: endpoint, ...defaults } = input
  const configured = route.with({
    ...defaults,
-    endpoint: { baseURL: baseURL ?? profiles.togetherai.baseURL },
+    endpoint: { baseURL: endpoint ?? baseURL },
    auth: AuthOptions.bearer(input, ["TOGETHER_API_KEY", "TOGETHER_AI_API_KEY"]),
  })
  return {
@@ -1,10 +1,8 @@
 import { AuthOptions, type ProviderAuthOption } from "../route/auth-options.js"
-import { Route, type RouteDefaultsInput, type CompactOperation } from "../route/client.js"
+import { Route, type RouteDefaultsInput } from "../route/client.js"
 import { Endpoint } from "../route/endpoint.js"
 import { HttpOptions, ProviderID, type ModelID } from "../schema/index.js"
-import * as OpenAICompatibleProfiles from "./openai-compatible-profile.js"
-import * as OpenAICompatibleChat from "../protocols/openai-compatible-chat.js"
-import * as OpenAIChat from "../protocols/openai-chat.js"
+import { OpenAIChat } from "../protocols/openai-chat.js"
 import { OpenResponsesChannel } from "../protocols/open-responses-channel.js"
 import { XAIResponses } from "../protocols/xai-responses.js"
 import { XAIImages } from "../protocols/xai-images.js"
@@ -12,6 +10,7 @@ import type { OpenAIOptionsInput } from "./openai-options.js"
 import type { ProviderPackage } from "../provider-package.js"

 export const id = ProviderID.make("xai")
+const baseURL = "https://api.x.ai/v1"

 export type XAIProviderOptionsInput = OpenAIOptionsInput & { readonly contextManagement?: never }

@@ -32,12 +31,12 @@ export type { XAIImageOptions } from "../protocols/xai-images.js"
 const RESPONSES_WEBSOCKET_ROTATE_AFTER_MS = 24 * 60 * 1000

 const responsesRoute = Route.make({
-  compact: XAIResponses.compact,
+  compact: { endpoint: XAIResponses.compact },
  id: "openai-responses",
  provider: id,
  providerMetadataKey: "xai",
  protocol: XAIResponses.protocol,
-  endpoint: Endpoint.path("/responses", { baseURL: OpenAICompatibleProfiles.profiles.xai.baseURL }),
+  endpoint: Endpoint.path("/responses", { baseURL }),
  transport: OpenResponsesChannel.transport({
    id: "openai-responses",
    name: "xAI Responses",
@@ -51,8 +50,8 @@ const chatRoute = Route.make({
  provider: id,
  providerMetadataKey: "xai",
  protocol: OpenAIChat.protocol,
-  endpoint: Endpoint.path("/chat/completions", { baseURL: OpenAICompatibleProfiles.profiles.xai.baseURL }),
-  transport: OpenAICompatibleChat.route.transport,
+  endpoint: Endpoint.path("/chat/completions", { baseURL }),
+  framing: OpenAIChat.framing,
  headers: ({ request }): Record<string, string> =>
    request.promptCacheKey ? { "x-grok-conv-id": request.promptCacheKey } : {},
 })
@@ -62,19 +61,19 @@ export const routes = [responsesRoute, chatRoute]
 const auth = (options: ProviderAuthOption<"optional">) => AuthOptions.bearer(options, "XAI_API_KEY")

 const configuredResponsesRoute = (input: LanguageModelOptions) => {
-  const { apiKey: _, auth: _auth, baseURL, ...rest } = input
+  const { apiKey: _, auth: _auth, baseURL: endpoint, ...rest } = input
  return responsesRoute.with({
    ...rest,
-    endpoint: { baseURL: baseURL ?? OpenAICompatibleProfiles.profiles.xai.baseURL },
+    endpoint: { baseURL: endpoint ?? baseURL },
    auth: auth(input),
  })
 }

 const configuredChatRoute = (input: LanguageModelOptions) => {
-  const { apiKey: _, auth: _auth, baseURL, ...rest } = input
+  const { apiKey: _, auth: _auth, baseURL: endpoint, ...rest } = input
  return chatRoute.with({
    ...rest,
-    endpoint: { baseURL: baseURL ?? OpenAICompatibleProfiles.profiles.xai.baseURL },
+    endpoint: { baseURL: endpoint ?? baseURL },
    auth: auth(input),
  })
 }
@@ -88,7 +87,7 @@ export const configure = (input: LanguageModelOptions = {}) => {
    XAIImages.model({
      id: modelID,
      auth: auth(input),
-      baseURL: input.baseURL ?? OpenAICompatibleProfiles.profiles.xai.baseURL,
+      baseURL: input.baseURL ?? baseURL,
      headers: input.headers,
      http: input.http === undefined ? undefined : HttpOptions.make(input.http),
    })
@@ -103,10 +102,11 @@ export const configure = (input: LanguageModelOptions = {}) => {
 }

 export const provider = configure()
-export const model: ProviderPackage.Definition<Settings, XAIProviderOptionsInput, CompactOperation>["model"] = (
-  modelID,
-  settings,
-) =>
+export const model: ProviderPackage.Definition<
+  Settings,
+  XAIProviderOptionsInput,
+  typeof responsesRoute.compact
+>["model"] = (modelID, settings) =>
  configure({
    apiKey: settings.apiKey,
    baseURL: settings.baseURL,
@@ -14,6 +14,7 @@ import type { ProtocolID, ProviderOptions } from "../schema/index.js"
 import {
  AIError,
  CompactionResponse,
+  CompactionCheckpointResponse,
  AIErrorReason,
  GenerationOptions,
  HttpOptions,
@@ -38,7 +39,7 @@ export interface RouteBody<Body> {
 export interface Route<
  Body,
  Prepared = unknown,
-  Compact extends CompactOperation | undefined = CompactOperation | undefined,
+  Compact extends CompactionOperations | undefined = CompactionOperations | undefined,
 > {
  readonly compact: Compact
  readonly id: string
@@ -53,7 +54,15 @@ export interface Route<
  readonly transport: Transport<Body, Prepared, unknown>
  readonly defaults: RouteDefaults
  readonly body: RouteBody<Body>
-  readonly with: (patch: RoutePatch<Body, Prepared>) => Route<Body, Prepared, Compact>
+  readonly with: {
+    <Next extends CompactionOperations | undefined>(
+      patch: RoutePatch<Body, Prepared> & { readonly compact: Next },
+    ): Route<Body, Prepared, Next>
+    (
+      patch: Omit<RoutePatch<Body, Prepared>, "compact"> & { readonly compact?: undefined },
+    ): Route<Body, Prepared, Compact>
+    (patch: RoutePatch<Body, Prepared>): Route<Body, Prepared>
+  }
  readonly model: <Options extends ProviderOptions = ProviderOptions>(
    input: RouteMappedLanguageModelInput,
  ) => LanguageModel<Options, Compact>
@@ -74,7 +83,7 @@ export interface Route<
 // Normal call sites use `OpenAIChat.route`; callers only need body types
 // when preparing a request with a protocol-specific type assertion.
 // oxlint-disable-next-line typescript-eslint/no-explicit-any
-export type AnyRoute<Compact extends CompactOperation | undefined = CompactOperation | undefined> = Route<
+export type AnyRoute<Compact extends CompactionOperations | undefined = CompactionOperations | undefined> = Route<
  any,
  any,
  Compact
@@ -101,6 +110,7 @@ export interface RouteDefaultsInput {
 }

 export interface RoutePatch<Body, Prepared> extends RouteDefaultsInput {
+  readonly compact?: CompactionOperations
  readonly id?: string
  readonly provider?: string | ProviderID
  readonly providerMetadataKey?: string
@@ -111,7 +121,7 @@ export interface RoutePatch<Body, Prepared> extends RouteDefaultsInput {

 type RouteMappedLanguageModelInput = RouteLanguageModelInput | RouteRoutedLanguageModelInput

-const makeRouteLanguageModel = <Options extends ProviderOptions, Compact extends CompactOperation | undefined>(
+const makeRouteLanguageModel = <Options extends ProviderOptions, Compact extends CompactionOperations | undefined>(
  route: AnyRoute<Compact>,
  mapped: RouteMappedLanguageModelInput,
 ) => {
@@ -162,10 +172,7 @@ export const httpOptions = (input: HttpOptionsInput | undefined) => {
 }

 export interface Interface {
-  readonly compact: (
-    request: CompactionRequest,
-    options?: Pick<StreamOptions, "http">,
-  ) => Effect.Effect<CompactionResponse, AIError>
+  readonly compact: CompactMethod
  readonly stream: StreamMethod
  readonly generate: GenerateMethod
 }
@@ -189,12 +196,64 @@ export type CompactOperation = (
  options?: Pick<StreamOptions, "http">,
 ) => Effect.Effect<CompactionResponse, AIError>

-export type CompactionRequest = LLMRequest & {
-  readonly model: LanguageModel<ProviderOptions, CompactOperation>
+export type TriggerCompactOperation = (
+  request: LLMRequest,
+  executor: RequestExecutor.Interface,
+  options: TriggerCompactOptions,
+) => Effect.Effect<CompactionCheckpointResponse, AIError>
+
+/** Protocol capabilities, not deployment/model eligibility. */
+export interface CompactionOperations {
+  readonly endpoint?: CompactOperation
+  readonly trigger?: TriggerCompactOperation
 }

-export const canCompact = (request: LLMRequest): request is CompactionRequest =>
-  request.model.route.compact !== undefined
+export interface EndpointCompactOptions extends Pick<StreamOptions, "http"> {
+  readonly mechanism?: "endpoint"
+  readonly webSocket?: never
+}
+
+export interface TriggerCompactOptions extends StreamOptions {
+  readonly mechanism: "trigger"
+}
+
+// Keep the required route shape explicit: the schema class's self type erases its model parameter in assignability.
+export type CompactionRequest = LLMRequest & {
+  readonly model: LanguageModel<ProviderOptions, { readonly endpoint: CompactOperation }>
+}
+export type CheckpointRequest = LLMRequest & {
+  readonly model: LanguageModel<ProviderOptions, { readonly trigger: TriggerCompactOperation }>
+}
+
+export interface CompactMethod<R = never> {
+  (request: CheckpointRequest, options: TriggerCompactOptions): Effect.Effect<CompactionCheckpointResponse, AIError, R>
+  (request: CompactionRequest, options?: EndpointCompactOptions): Effect.Effect<CompactionResponse, AIError, R>
+}
+
+export function canCompact(
+  request: LLMRequest,
+  options?: { readonly mechanism?: "endpoint" },
+): request is CompactionRequest
+export function canCompact(
+  request: LLMRequest,
+  options: { readonly mechanism: "trigger" },
+): request is CheckpointRequest
+export function canCompact(request: LLMRequest, options?: { readonly mechanism?: string }) {
+  if (options?.mechanism === "trigger") return request.model.route.compact?.trigger !== undefined
+  if (options?.mechanism !== undefined && options.mechanism !== "endpoint") return false
+  return request.model.route.compact?.endpoint !== undefined
+}
+
+const unsupportedCompaction = (request: LLMRequest, mechanism: string | undefined) => {
+  if (mechanism !== undefined && mechanism !== "endpoint" && mechanism !== "trigger")
+    return ProviderShared.invalidRequest(`Unknown compaction mechanism: ${mechanism}`)
+  return ProviderShared.unsupportedOperation({
+    operation: mechanism === "trigger" ? "compact.trigger" : "compact",
+    provider: request.model.provider,
+    route: request.model.route.id,
+    message: `${request.model.provider}/${request.model.route.id} does not support ${mechanism === "trigger" ? "trigger" : "explicit"} compaction`,
+  })
+}

 export class Service extends Context.Service<Service, Interface>()("@opencode/LLMClient") {}

@@ -216,7 +275,7 @@ const resolveRequestOptions = (request: LLMRequest) => {
 }

 export interface MakeInput<Body, Frame, Event, State> {
-  readonly compact?: CompactOperation
+  readonly compact?: CompactionOperations
  /** Route id used in diagnostics and prepared request metadata. */
  readonly id: string
  /** Provider identity for route-owned model construction. */
@@ -238,7 +297,7 @@ export interface MakeInput<Body, Frame, Event, State> {
 }

 export interface MakeTransportInput<Body, Prepared, Frame, Event, State> {
-  readonly compact?: CompactOperation
+  readonly compact?: CompactionOperations
  /** Route id used in diagnostics and prepared request metadata. */
  readonly id: string
  /** Provider identity for route-owned model construction. */
@@ -326,9 +385,10 @@ function makeFromTransport<Body, Prepared, Frame, Event, State>(
      defaults: routeInput.defaults ?? {},
      body: protocol.body,
      with: (patch: RoutePatch<Body, Prepared>) => {
-        const { id, provider, providerMetadataKey, auth, transport, endpoint, ...defaults } = patch
+        const { compact, id, provider, providerMetadataKey, auth, transport, endpoint, ...defaults } = patch
        return build({
          ...routeInput,
+          compact: "compact" in patch ? compact : routeInput.compact,
          id: id ?? routeInput.id,
          provider: provider ?? routeInput.provider,
          providerMetadataKey:
@@ -343,7 +403,7 @@ function makeFromTransport<Body, Prepared, Frame, Event, State>(
        })
      },
      model: <Options extends ProviderOptions = ProviderOptions>(input: RouteMappedLanguageModelInput) =>
-        makeRouteLanguageModel<Options, CompactOperation | undefined>(route, input),
+        makeRouteLanguageModel<Options, CompactionOperations | undefined>(route, input),
      prepareTransport: (body, request, options) =>
        routeInput.transport.prepare({
          body,
@@ -440,12 +500,12 @@ function makeFromTransport<Body, Prepared, Frame, Event, State>(
  return build({ ...input, defaults: mergeRouteDefaults(undefined, input.defaults ?? {}) })
 }

-export function make<Body, Prepared, Frame, Event, State>(
-  input: MakeTransportInput<Body, Prepared, Frame, Event, State> & { readonly compact: CompactOperation },
-): Route<Body, Prepared, CompactOperation>
-export function make<Body, Frame, Event, State>(
-  input: MakeInput<Body, Frame, Event, State> & { readonly compact: CompactOperation },
-): Route<Body, HttpTransport.HttpPrepared<Frame>, CompactOperation>
+export function make<Body, Prepared, Frame, Event, State, Compact extends CompactionOperations>(
+  input: MakeTransportInput<Body, Prepared, Frame, Event, State> & { readonly compact: Compact },
+): Route<Body, Prepared, Compact>
+export function make<Body, Frame, Event, State, Compact extends CompactionOperations>(
+  input: MakeInput<Body, Frame, Event, State> & { readonly compact: Compact },
+): Route<Body, HttpTransport.HttpPrepared<Frame>, Compact>
 export function make<Body, Prepared, Frame, Event, State>(
  input: MakeTransportInput<Body, Prepared, Frame, Event, State>,
 ): Route<Body, Prepared>
@@ -487,10 +547,14 @@ export function make<Body, Prepared, Frame, Event, State>(
 }

 const prepareRequest = (request: LLMRequest) => {
-  const original = applyCachePolicy(resolveRequestOptions(request))
+  const original = resolveRequestOptions(request)
  const sanitized = LLMRequest.update(original, sanitizeSurrogates({ ...LLMRequest.input(original), model: undefined }))
-  const tools = [...new Map(sanitized.tools.map((tool) => [tool.name, tool])).values()]
-  const resolved = tools.length === sanitized.tools.length ? sanitized : LLMRequest.update(sanitized, { tools })
+  // Deduplicate per sibling level; a tool and a namespace may share a name.
+  const dedupe = (tools: LLMRequest["tools"]): LLMRequest["tools"] =>
+    [...new Map(tools.map((tool) => [`${tool.type}:${tool.name}`, tool])).values()].map((tool) =>
+      tool.type === "tool" ? tool : { ...tool, tools: dedupe(tool.tools) },
+    )
+  const resolved = applyCachePolicy(LLMRequest.update(sanitized, { tools: dedupe(sanitized.tools) }))
  const headers = resolved.model.route.headers?.({ request: resolved })
  return headers === undefined
    ? resolved
@@ -557,14 +621,23 @@ export function generate(request: LLMRequest, options?: StreamOptions): Effect.E
  })
 }

-export const compact = (
+export function compact(
+  request: CheckpointRequest,
+  options: TriggerCompactOptions,
+): Effect.Effect<CompactionCheckpointResponse, AIError, Service>
+export function compact(
  request: CompactionRequest,
-  options?: Pick<StreamOptions, "http">,
-): Effect.Effect<CompactionResponse, AIError, Service> =>
-  Effect.gen(function* () {
+  options?: EndpointCompactOptions,
+): Effect.Effect<CompactionResponse, AIError, Service>
+export function compact(request: LLMRequest, options?: EndpointCompactOptions | TriggerCompactOptions) {
+  return Effect.gen(function* () {
    const client = yield* Service
-    return yield* client.compact(request, options)
+    if (options?.mechanism === "trigger" && canCompact(request, options)) return yield* client.compact(request, options)
+    if ((options?.mechanism === undefined || options.mechanism === "endpoint") && canCompact(request))
+      return yield* client.compact(request, options)
+    return yield* unsupportedCompaction(request, options?.mechanism)
  })
+}

 export const streamRequest = (request: LLMRequest, options?: StreamOptions) =>
  Stream.unwrap(
@@ -578,21 +651,27 @@ export const layer: Layer.Layer<Service, never, RequestExecutor.Service> = Layer
  Effect.gen(function* () {
    const executor = yield* RequestExecutor.Service
    const stream = streamRequestWith({ http: executor })
+    function compact(
+      request: CompactionRequest,
+      options?: EndpointCompactOptions,
+    ): Effect.Effect<CompactionResponse, AIError>
+    function compact(
+      request: CheckpointRequest,
+      options: TriggerCompactOptions,
+    ): Effect.Effect<CompactionCheckpointResponse, AIError>
+    function compact(request: LLMRequest, options?: EndpointCompactOptions | TriggerCompactOptions) {
+      return Effect.suspend((): Effect.Effect<CompactionResponse | CompactionCheckpointResponse, AIError> => {
+        if (options?.mechanism === "trigger" && canCompact(request, options))
+          return request.model.route.compact.trigger(prepareRequest(request), executor, options)
+        if ((options?.mechanism === undefined || options.mechanism === "endpoint") && canCompact(request))
+          return request.model.route.compact.endpoint(prepareRequest(request), executor, options)
+        return unsupportedCompaction(request, options?.mechanism)
+      })
+    }
    return Service.of({
      stream,
      generate: generateWith(stream),
-      compact: (request, options) =>
-        Effect.suspend(() => {
-          const operation = request.model.route.compact
-          if (!operation)
-            return ProviderShared.unsupportedOperation({
-              operation: "compact",
-              provider: request.model.provider,
-              route: request.model.route.id,
-              message: `${request.model.provider}/${request.model.route.id} does not support explicit compaction`,
-            })
-          return operation(prepareRequest(request), executor, options)
-        }),
+      compact,
    })
  }),
 )
@@ -9,6 +9,12 @@ export type {
  Interface as LLMClientShape,
  Service as LLMClientService,
  StreamOptions,
+  CompactMethod,
+  CompactionOperations,
+  CompactionRequest,
+  CheckpointRequest,
+  EndpointCompactOptions,
+  TriggerCompactOptions,
 } from "./client.js"
 export * from "./executor.js"
 export { Auth } from "./auth.js"
@@ -97,6 +97,21 @@ export class CompactionResponse extends Schema.Class<CompactionResponse>("LLM.Co
  usage: Schema.optional(Usage),
 }) {}

+/** A checkpoint only; retained history and replacement-window construction belong to the caller. */
+export class CompactionCheckpointResponse extends Schema.Class<CompactionCheckpointResponse>(
+  "LLM.CompactionCheckpointResponse",
+)({
+  checkpoint: CompactionPart.pipe(
+    Schema.refine(
+      (part): part is CompactionPart & { readonly encrypted: string; readonly text?: never } =>
+        part.encrypted !== undefined && part.encrypted.length > 0,
+      { message: "A checkpoint response requires encrypted compaction content" },
+    ),
+  ),
+  responseID: Schema.String.check(Schema.isPattern(/\S/)),
+  usage: Schema.optional(Usage),
+}) {}
+
 export const StepStart = Schema.Struct({
  type: Schema.tag("step-start"),
  index: Schema.Number,
@@ -155,6 +170,7 @@ export const ToolInputStart = Schema.Struct({
  type: Schema.tag("tool-input-start"),
  id: ToolCallID,
  name: Schema.String,
+  namespace: Schema.optional(Schema.String),
  providerExecuted: Schema.optional(Schema.Boolean),
  providerMetadata: Schema.optional(ProviderMetadata),
 }).annotate({ identifier: "LLM.Event.ToolInputStart" })
@@ -164,6 +180,7 @@ export const ToolInputDelta = Schema.Struct({
  type: Schema.tag("tool-input-delta"),
  id: ToolCallID,
  name: Schema.String,
+  namespace: Schema.optional(Schema.String),
  text: Schema.String,
  /** Best-effort parse of all input fragments received through this delta. */
  input: Schema.optional(Schema.Unknown),
@@ -174,6 +191,7 @@ export const ToolInputEnd = Schema.Struct({
  type: Schema.tag("tool-input-end"),
  id: ToolCallID,
  name: Schema.String,
+  namespace: Schema.optional(Schema.String),
  providerMetadata: Schema.optional(ProviderMetadata),
 }).annotate({ identifier: "LLM.Event.ToolInputEnd" })
 export type ToolInputEnd = Schema.Schema.Type<typeof ToolInputEnd>
@@ -183,6 +201,7 @@ export const ToolInputError = Schema.Struct({
  type: Schema.tag("tool-input-error"),
  id: ToolCallID,
  name: Schema.String,
+  namespace: Schema.optional(Schema.String),
  raw: Schema.String,
 }).annotate({ identifier: "LLM.Event.ToolInputError" })
 export type ToolInputError = Schema.Schema.Type<typeof ToolInputError>
@@ -191,6 +210,7 @@ export const ToolCall = Schema.Struct({
  type: Schema.tag("tool-call"),
  id: ToolCallID,
  name: Schema.String,
+  namespace: Schema.optional(Schema.String),
  input: Schema.Unknown,
  providerExecuted: Schema.optional(Schema.Boolean),
  providerMetadata: Schema.optional(ProviderMetadata),
@@ -201,6 +221,7 @@ export const ToolResult = Schema.Struct({
  type: Schema.tag("tool-result"),
  id: ToolCallID,
  name: Schema.String,
+  namespace: Schema.optional(Schema.String),
  result: ToolResultValue,
  output: Schema.optional(ToolOutput),
  providerExecuted: Schema.optional(Schema.Boolean),
@@ -212,6 +233,7 @@ export const ToolError = Schema.Struct({
  type: Schema.tag("tool-error"),
  id: ToolCallID,
  name: Schema.String,
+  namespace: Schema.optional(Schema.String),
  message: Schema.String,
  error: Schema.optional(Schema.Defect()),
  providerMetadata: Schema.optional(ProviderMetadata),
@@ -385,6 +407,7 @@ interface ContentAssembly {

 interface ToolInputAssembly {
  readonly name: string
+  readonly namespace?: string
  readonly text: string
  readonly providerMetadata?: ProviderMetadata
 }
@@ -522,12 +545,17 @@ const reduceToolInputStart = (state: ResponseState, event: ToolInputStart): Resp
  ...state,
  toolInputs: {
    ...state.toolInputs,
-    [event.id]: { name: event.name, text: "", providerMetadata: event.providerMetadata },
+    [event.id]: {
+      name: event.name,
+      namespace: event.namespace,
+      text: "",
+      providerMetadata: event.providerMetadata,
+    },
  },
 })

 const reduceToolInputDelta = (state: ResponseState, event: ToolInputDelta): ResponseState => {
-  const current = state.toolInputs[event.id] ?? { name: event.name, text: "" }
+  const current = state.toolInputs[event.id] ?? { name: event.name, namespace: event.namespace, text: "" }
  return {
    ...state,
    toolInputs: { ...state.toolInputs, [event.id]: { ...current, text: current.text + event.text } },
@@ -535,7 +563,7 @@ const reduceToolInputDelta = (state: ResponseState, event: ToolInputDelta): Resp
 }

 const reduceToolInputEnd = (state: ResponseState, event: ToolInputEnd): ResponseState => {
-  const current = state.toolInputs[event.id] ?? { name: event.name, text: "" }
+  const current = state.toolInputs[event.id] ?? { name: event.name, namespace: event.namespace, text: "" }
  return {
    ...state,
    toolInputs: {
@@ -543,6 +571,7 @@ const reduceToolInputEnd = (state: ResponseState, event: ToolInputEnd): Response
      [event.id]: {
        ...current,
        name: event.name,
+        namespace: event.namespace,
        providerMetadata: event.providerMetadata ?? current.providerMetadata,
      },
    },
@@ -553,6 +582,7 @@ const toolCallContent = (event: ToolCall): ContentPart =>
  ToolCallPart.make({
    id: event.id,
    name: event.name,
+    namespace: event.namespace,
    input: event.input,
    ...(event.providerExecuted === undefined ? {} : { providerExecuted: event.providerExecuted }),
    ...(event.providerMetadata === undefined ? {} : { providerMetadata: event.providerMetadata }),
@@ -562,6 +592,7 @@ const toolResultContent = (event: ToolResult): ContentPart =>
  ToolResultPart.make({
    id: event.id,
    name: event.name,
+    namespace: event.namespace,
    result: event.result,
    ...(event.providerExecuted === undefined ? {} : { providerExecuted: event.providerExecuted }),
    ...(event.providerMetadata === undefined ? {} : { providerMetadata: event.providerMetadata }),
@@ -135,6 +135,7 @@ export const ToolCallPart = Object.assign(
    type: Schema.Literal("tool-call"),
    id: Schema.String,
    name: Schema.String,
+    namespace: Schema.optional(Schema.String),
    input: Schema.Unknown,
    providerExecuted: Schema.optional(Schema.Boolean),
    cache: Schema.optional(CacheHint),
@@ -152,6 +153,7 @@ export const ToolResultPart = Object.assign(
    type: Schema.Literal("tool-result"),
    id: Schema.String,
    name: Schema.String,
+    namespace: Schema.optional(Schema.String),
    result: ToolResultValue,
    providerExecuted: Schema.optional(Schema.Boolean),
    cache: Schema.optional(CacheHint),
@@ -168,6 +170,7 @@ export const ToolResultPart = Object.assign(
      type: "tool-result",
      id: input.id,
      name: input.name,
+      namespace: input.namespace,
      result: ToolResultValue.make(input.result, input.resultType),
      providerExecuted: input.providerExecuted,
      cache: input.cache,
@@ -266,7 +269,7 @@ export namespace Message {
    make({ role: "tool", content: ["type" in result ? result : ToolResultPart.make(result)] })
 }

-export class ToolDefinition extends Schema.Class<ToolDefinition>("LLM.ToolDefinition")({
+const toolDefinitionFields = {
  name: Schema.String,
  description: Schema.String,
  inputSchema: JsonSchema,
@@ -274,15 +277,71 @@ export class ToolDefinition extends Schema.Class<ToolDefinition>("LLM.ToolDefini
  cache: Schema.optional(CacheHint),
  metadata: Schema.optional(Schema.Record(Schema.String, Schema.Unknown)),
  native: Schema.optional(Schema.Record(Schema.String, Schema.Unknown)),
-}) {}
+}
+
+export type ToolDefinitionInput = Schema.Struct.Type<typeof toolDefinitionFields>
+
+export class ToolDefinition extends Schema.Class<ToolDefinition>("LLM.ToolDefinition")({
+  type: Schema.Literal("tool"),
+  ...toolDefinitionFields,
+}) {
+  constructor(input: ToolDefinitionInput) {
+    super({ ...input, type: "tool" })
+  }
+}

 export namespace ToolDefinition {
-  export type Input = ToolDefinition | ConstructorParameters<typeof ToolDefinition>[0]
+  export type Input = ToolDefinition | ToolDefinitionInput

  /** Normalize tool definition input into the canonical `ToolDefinition` class. */
  export const make = (input: Input) => (input instanceof ToolDefinition ? input : new ToolDefinition(input))
 }

+export type ToolNamespace = {
+  readonly type: "namespace"
+  readonly name: string
+  readonly description?: string
+  readonly tools: ReadonlyArray<ToolEntry>
+}
+
+export type ToolNamespaceInput = Omit<ToolNamespace, "type" | "tools"> & {
+  readonly tools: ReadonlyArray<ToolEntryInput>
+}
+export type ToolNamespaceEntryInput = ToolNamespaceInput & { readonly type: "namespace" }
+
+export const ToolNamespace: Schema.Codec<ToolNamespace> & {
+  readonly make: (input: ToolNamespace | ToolNamespaceInput) => ToolNamespace
+} = Object.assign(
+  Schema.Struct({
+    type: Schema.Literal("namespace"),
+    name: Schema.String,
+    description: Schema.optional(Schema.UndefinedOr(Schema.String)),
+    tools: Schema.Array(Schema.suspend((): Schema.Codec<ToolEntry> => ToolEntry)),
+  }).annotate({ identifier: "LLM.ToolNamespace" }),
+  {
+    make: (input: ToolNamespace | ToolNamespaceInput): ToolNamespace => ({
+      ...input,
+      type: "namespace",
+      tools: input.tools.map(ToolEntry.make),
+    }),
+  },
+)
+
+export type ToolEntry = ToolDefinition | ToolNamespace
+export type ToolEntryInput = ToolDefinition.Input | ToolNamespaceEntryInput
+export const ToolEntry: Schema.Codec<ToolEntry> & {
+  readonly make: (input: ToolEntryInput) => ToolEntry
+} = Object.assign(
+  Schema.Union([ToolDefinition, ToolNamespace]).pipe(
+    Schema.toTaggedUnion("type"),
+    Schema.annotate({ identifier: "LLM.ToolEntry" }),
+  ),
+  {
+    make: (input: ToolEntryInput): ToolEntry =>
+      "type" in input && input.type === "namespace" ? ToolNamespace.make(input) : ToolDefinition.make(input),
+  },
+)
+
 export class ToolChoice extends Schema.Class<ToolChoice>("LLM.ToolChoice")({
  type: Schema.Literals(["auto", "none", "required", "tool"]),
  name: Schema.optional(Schema.String),
@@ -312,7 +371,7 @@ const requestSchema = Schema.Struct({
  model: LanguageModelSchema,
  system: Schema.Array(SystemPart),
  messages: Schema.Array(Message),
-  tools: Schema.Array(ToolDefinition),
+  tools: Schema.Array(ToolEntry),
  toolChoice: Schema.optional(ToolChoice),
  generation: Schema.optional(GenerationOptions),
  providerOptions: Schema.optional(ProviderOptions),
@@ -1,6 +1,6 @@
 import { Schema } from "effect"
 import { ModelID, ProviderID } from "./ids.js"
-import type { AnyRoute, CompactOperation } from "../route/client.js"
+import type { AnyRoute, CompactionOperations } from "../route/client.js"
 import { isRecord } from "../utils/record.js"

 export const JsonSchema = Schema.Record(Schema.String, Schema.Unknown)
@@ -175,7 +175,7 @@ export namespace LanguageModelCompatibility {

 export class LanguageModel<
  Options extends ProviderOptions = ProviderOptions,
-  Compact extends CompactOperation | undefined = CompactOperation | undefined,
+  Compact extends CompactionOperations | undefined = CompactionOperations | undefined,
 > {
  declare protected readonly _ProviderOptions: Options
  readonly id: ModelID
@@ -194,7 +194,7 @@ export class LanguageModel<

  static make<
    Options extends ProviderOptions = ProviderOptions,
-    Compact extends CompactOperation | undefined = CompactOperation | undefined,
+    Compact extends CompactionOperations | undefined = CompactionOperations | undefined,
  >(input: LanguageModel.Input<Compact>) {
    return new LanguageModel<Options, Compact>({
      id: ModelID.make(input.id),
@@ -206,7 +206,7 @@ export class LanguageModel<
    })
  }

-  static input<Options extends ProviderOptions, Compact extends CompactOperation | undefined>(
+  static input<Options extends ProviderOptions, Compact extends CompactionOperations | undefined>(
    model: LanguageModel<Options, Compact>,
  ): LanguageModel.ConstructorInput<Compact> {
    return {
@@ -218,11 +218,11 @@ export class LanguageModel<
    }
  }

-  static update<Options extends ProviderOptions, Compact extends CompactOperation | undefined>(
+  static update<Options extends ProviderOptions, Compact extends CompactionOperations | undefined>(
    model: LanguageModel<Options>,
    patch: Partial<LanguageModel.Input<Compact>> & { readonly route: AnyRoute<Compact> },
  ): LanguageModel<Options, Compact>
-  static update<Options extends ProviderOptions, Compact extends CompactOperation | undefined>(
+  static update<Options extends ProviderOptions, Compact extends CompactionOperations | undefined>(
    model: LanguageModel<Options, Compact>,
    patch: Partial<Omit<LanguageModel.Input, "route">> & { readonly route?: undefined },
  ): LanguageModel<Options, Compact>
@@ -241,7 +241,7 @@ export class LanguageModel<
 }

 export namespace LanguageModel {
-  export type ConstructorInput<Compact extends CompactOperation | undefined = CompactOperation | undefined> = {
+  export type ConstructorInput<Compact extends CompactionOperations | undefined = CompactionOperations | undefined> = {
    readonly id: ModelID
    readonly provider: ProviderID
    readonly route: AnyRoute<Compact>
@@ -249,7 +249,7 @@ export namespace LanguageModel {
    readonly compatibility?: LanguageModelCompatibility
  }

-  export type Input<Compact extends CompactOperation | undefined = CompactOperation | undefined> = Omit<
+  export type Input<Compact extends CompactionOperations | undefined = CompactionOperations | undefined> = Omit<
    ConstructorInput<Compact>,
    "id" | "provider" | "defaults" | "compatibility"
  > & {
@@ -1,10 +1,17 @@
 export * as TestLLM from "./testing.js"

-import { LLMClient } from "./route/client.js"
+import {
+  LLMClient,
+  type CompactionRequest,
+  type CheckpointRequest,
+  type EndpointCompactOptions,
+  type TriggerCompactOptions,
+} from "./route/client.js"
 import {
  LLMEvent,
  LLMResponse,
  CompactionResponse,
+  CompactionCheckpointResponse,
  type FinishReasonDetails,
  type AIError,
  type LLMRequest,
@@ -13,7 +20,11 @@ import {
 } from "./schema/index.js"
 import { Context, Deferred, Effect, Latch, Layer, Queue, Scope, Stream } from "effect"

-export type Response = readonly LLMEvent[] | Stream.Stream<LLMEvent, AIError> | CompactionResponse
+export type Response =
+  | readonly LLMEvent[]
+  | Stream.Stream<LLMEvent, AIError>
+  | CompactionResponse
+  | CompactionCheckpointResponse

 export type Gate = Readonly<{ started: Effect.Effect<void>; release: Effect.Effect<void> }>

@@ -132,21 +143,38 @@ const make = (options: LayerOptions) =>
      Stream.unwrap(
        take(request).pipe(
          Effect.map((response) => {
-            if (response instanceof CompactionResponse)
+            if (response instanceof CompactionResponse || response instanceof CompactionCheckpointResponse)
              return Stream.die("TestLLM generation requires an event response")
            return Stream.isStream(response) ? response : Stream.fromIterable(response)
          }),
        ),
      )
-    const test = Test.of({
-      compact: (request) =>
-        take(request).pipe(
-          Effect.flatMap((response) =>
-            response instanceof CompactionResponse
+    function compact(
+      request: CompactionRequest,
+      options?: EndpointCompactOptions,
+    ): Effect.Effect<CompactionResponse, AIError>
+    function compact(
+      request: CheckpointRequest,
+      options: TriggerCompactOptions,
+    ): Effect.Effect<CompactionCheckpointResponse, AIError>
+    function compact(
+      request: LLMRequest,
+      options?: EndpointCompactOptions | TriggerCompactOptions,
+    ): Effect.Effect<CompactionResponse | CompactionCheckpointResponse, AIError> {
+      return take(request).pipe(
+        Effect.flatMap((response): Effect.Effect<CompactionResponse | CompactionCheckpointResponse> => {
+          if (options?.mechanism === "trigger")
+            return response instanceof CompactionCheckpointResponse
              ? Effect.succeed(response)
-              : Effect.die("TestLLM compaction requires a CompactionResponse"),
-          ),
-        ),
+              : Effect.die("TestLLM trigger compaction requires a CompactionCheckpointResponse")
+          return response instanceof CompactionResponse
+            ? Effect.succeed(response)
+            : Effect.die("TestLLM compaction requires a CompactionResponse")
+        }),
+      )
+    }
+    const test = Test.of({
+      compact,
      stream,
      generate: (request) =>
        stream(request).pipe(
@@ -37,7 +37,13 @@ function missingToolResults(calls: Iterable<ToolCallPart>) {
  return new Message({
    role: "tool",
    content: [...calls].map((call) =>
-      ToolResultPart.make({ id: call.id, name: call.name, result: MISSING_TOOL_RESULT, resultType: "error" }),
+      ToolResultPart.make({
+        id: call.id,
+        name: call.name,
+        namespace: call.namespace,
+        result: MISSING_TOOL_RESULT,
+        resultType: "error",
+      }),
    ),
  })
 }
@@ -47,7 +53,7 @@ function normalizeToolMessage(message: Message, pending: Map<string, ToolCallPar
    if (part.type !== "tool-result" || part.providerExecuted === true) return part
    const call = pending.get(part.id)
    if (call) pending.delete(part.id)
-    return normalizeToolResult(part, call?.name ?? part.name)
+    return normalizeToolResult(part, call)
  })
  if (content.length === 0) return undefined
  if (content.every((part, index) => part === message.content[index])) return message
@@ -61,8 +67,11 @@ function normalizeToolMessage(message: Message, pending: Map<string, ToolCallPar
  })
 }

-function normalizeToolResult(part: ToolResultPart, name: string): ToolResultPart {
-  const named = part.name === name ? part : { ...part, name }
+function normalizeToolResult(part: ToolResultPart, call: ToolCallPart | undefined): ToolResultPart {
+  const named =
+    call === undefined || (part.name === call.name && part.namespace === call.namespace)
+      ? part
+      : { ...part, name: call.name, namespace: call.namespace }
  if (named.result.type === "text" && named.result.value === "")
    return { ...named, result: { type: "text", value: EMPTY_TOOL_OUTPUT } }
  if (named.result.type === "error" && named.result.value === "")
@@ -21,10 +21,11 @@ export interface DispatchResult extends ToolSettlement {

 /** Execute one canonical tool call without owning provider IO or continuation. */
 export const dispatch = (tools: Tools, call: ToolCallPart): Effect.Effect<DispatchResult> => {
-  const tool = tools[call.name]
-  if (!tool) return Effect.succeed(result(call, { type: "error", value: `Unknown tool: ${call.name}` }))
+  const name = call.namespace === undefined ? call.name : `${call.namespace}.${call.name}`
+  const tool = tools[name]
+  if (!tool) return Effect.succeed(result(call, { type: "error", value: `Unknown tool: ${name}` }))
  if (!tool.execute)
-    return Effect.succeed(result(call, { type: "error", value: `Tool has no execute handler: ${call.name}` }))
+    return Effect.succeed(result(call, { type: "error", value: `Tool has no execute handler: ${name}` }))

  return decodeAndExecute(tool, call).pipe(
    Effect.map((value) => result(call, value)),
@@ -38,7 +39,11 @@ const decodeAndExecute = (tool: AnyTool, call: ToolCallPart): Effect.Effect<Tool
  tool._decode(call.input).pipe(
    Effect.mapError((error) => new ToolFailure({ message: `Invalid tool input: ${error.message}` })),
    Effect.flatMap((decoded) =>
-      tool.execute!(decoded, { id: call.id, name: call.name }).pipe(
+      tool.execute!(decoded, {
+        id: call.id,
+        name: call.name,
+        namespace: call.namespace,
+      }).pipe(
        Effect.flatMap((value) =>
          tool._encode(value).pipe(
            Effect.mapError(
@@ -71,6 +76,7 @@ const result = (call: ToolCallPart, value: ToolResultValueType | ToolSettlement,
            LLMEvent.toolError({
              id: call.id,
              name: call.name,
+              namespace: call.namespace,
              message: String(settlement.result.value),
              error,
              providerMetadata: call.providerMetadata,
@@ -78,6 +84,7 @@ const result = (call: ToolCallPart, value: ToolResultValueType | ToolSettlement,
            LLMEvent.toolResult({
              id: call.id,
              name: call.name,
+              namespace: call.namespace,
              result: settlement.result,
              providerMetadata: call.providerMetadata,
            }),
@@ -86,6 +93,7 @@ const result = (call: ToolCallPart, value: ToolResultValueType | ToolSettlement,
            LLMEvent.toolResult({
              id: call.id,
              name: call.name,
+              namespace: call.namespace,
              result: settlement.result,
              output: settlement.output,
              providerMetadata: call.providerMetadata,
@@ -16,6 +16,7 @@ export type ToolSchema<T> = Schema.Codec<T, any, never, never>
 export interface ToolExecuteContext {
  readonly id: ToolCallPart["id"]
  readonly name: ToolCallPart["name"]
+  readonly namespace?: ToolCallPart["namespace"]
 }

 export type ToolExecute<Parameters extends ToolSchema<any>, Success extends ToolSchema<any>> = (
@@ -6,7 +6,8 @@ import * as AmazonBedrock from "../src/providers/amazon-bedrock.js"
 import * as Anthropic from "../src/providers/anthropic.js"
 import * as AnthropicCompatible from "../src/providers/anthropic-compatible.js"
 import * as Azure from "../src/providers/azure.js"
-import * as Cloudflare from "../src/providers/cloudflare.js"
+import { CloudflareWorkersAI } from "../src/providers/cloudflare-workers-ai.js"
+import { DeepSeek } from "../src/providers/deepseek.js"
 import * as Google from "../src/providers/google.js"
 import * as GoogleVertex from "../src/providers/google-vertex.js"
 import * as GoogleVertexChat from "../src/providers/google-vertex-chat.js"
@@ -263,10 +264,10 @@ XAI.configure({ apiKey: "xai-key" }).responses("grok-4", {})
 // @ts-expect-error xAI Chat selectors only accept model ids.
 XAI.configure({ apiKey: "xai-key" }).chat("grok-4", {})

-OpenAICompatible.deepseek.configure({ apiKey: "deepseek-key" }).model("deepseek-chat")
+DeepSeek.configure({ apiKey: "deepseek-key" }).model("deepseek-chat")
 // @ts-expect-error OpenAI-compatible family selectors only accept model ids.
-OpenAICompatible.deepseek.configure({ apiKey: "deepseek-key" }).model("deepseek-chat", {})
+DeepSeek.configure({ apiKey: "deepseek-key" }).model("deepseek-chat", {})

-Cloudflare.CloudflareWorkersAI.configure({ accountId: "account", apiKey: "cf-key" }).model("@cf/meta/llama")
+CloudflareWorkersAI.configure({ accountId: "account", apiKey: "cf-key" }).model("@cf/meta/llama")
 // @ts-expect-error Cloudflare Workers AI model selectors only accept model ids.
-Cloudflare.CloudflareWorkersAI.configure({ accountId: "account", apiKey: "cf-key" }).model("@cf/meta/llama", {})
+CloudflareWorkersAI.configure({ accountId: "account", apiKey: "cf-key" }).model("@cf/meta/llama", {})
@@ -215,6 +215,35 @@ describe("applyCachePolicy", () => {
    }),
  )

+  it.effect("deduplicates tools before counting cache hints", () =>
+    Effect.gen(function* () {
+      const manual = new CacheHint({ type: "ephemeral" })
+      const duplicate = (description: string) => ({
+        name: "lookup",
+        description,
+        inputSchema: { type: "object" },
+        cache: manual,
+      })
+      const prepared = yield* compileRequest(
+        LLM.request({
+          model: anthropicModel,
+          tools: [
+            duplicate("first"),
+            duplicate("second"),
+            duplicate("third"),
+            duplicate("fourth"),
+            { name: "lookup", description: "final", inputSchema: { type: "object" } },
+          ],
+          cache: { tools: true },
+        }),
+      )
+
+      expect(prepared.body.tools).toEqual([
+        expect.objectContaining({ name: "lookup", description: "final", cache_control: { type: "ephemeral" } }),
+      ])
+    }),
+  )
+
  it.effect("auto policy preserves manual CacheHints on other parts", () =>
    Effect.gen(function* () {
      const prepared = yield* compileRequest(
@@ -281,6 +310,30 @@ describe("applyCachePolicy", () => {
    }),
  )

+  test("marks the final leaf inside a tool namespace", () => {
+    const request = LLM.request({
+      model: anthropicModel,
+      tools: [
+        {
+          type: "namespace",
+          name: "crm",
+          tools: [
+            { name: "lookup", description: "lookup", inputSchema: {} },
+            { name: "orders", description: "orders", inputSchema: {} },
+          ],
+        },
+      ],
+      cache: { tools: true },
+    })
+    const applied = applyCachePolicy(request)
+    const namespace = applied.tools[0]
+
+    expect(namespace?.type).toBe("namespace")
+    if (namespace?.type !== "namespace") throw new Error("Expected namespace")
+    expect(namespace.tools[0]).not.toHaveProperty("cache")
+    expect(namespace.tools[1]).toHaveProperty("cache", { type: "ephemeral" })
+  })
+
  it.effect("ttlSeconds in the policy flows through to wire markers", () =>
    Effect.gen(function* () {
      const prepared = yield* compileRequest(
@@ -1,8 +1,16 @@
 import { describe, expect, test } from "bun:test"
 import { Effect, Ref, Schema } from "effect"
 import { HttpClientRequest, HttpClientResponse } from "effect/unstable/http"
-import { LLM, LLMRequest, Message, ToolCallPart, ToolDefinition, mergeProviderOptions } from "../src/index.js"
-import { AnthropicMessages, OpenAIChat } from "../src/protocols.js"
+import {
+  LLM,
+  LLMRequest,
+  Message,
+  ToolCallPart,
+  ToolDefinition,
+  ToolNamespace,
+  mergeProviderOptions,
+} from "../src/index.js"
+import { AnthropicMessages, OpenAIChat, OpenAIResponses } from "../src/protocols.js"
 import { Auth, LLMClient } from "../src/route.js"
 import { compileRequest } from "../src/route/client.js"
 import { it } from "./lib/effect.js"
@@ -106,6 +114,58 @@ describe("request option precedence", () => {
    }),
  )

+  it.effect("deduplicates tools within each namespace", () =>
+    Effect.gen(function* () {
+      const prepared = yield* compileRequest(
+        LLM.request({
+          model: OpenAIResponses.route.model({ id: "gpt-5.4" }),
+          tools: [
+            ToolDefinition.make({ name: "crm", description: "Top-level CRM tool", inputSchema: {} }),
+            ToolNamespace.make({
+              name: "crm",
+              description: "CRM tools",
+              tools: [
+                ToolDefinition.make({ name: "lookup", description: "old", inputSchema: {} }),
+                ToolDefinition.make({ name: "search", description: "search", inputSchema: {} }),
+                ToolDefinition.make({ name: "lookup", description: "new", inputSchema: {} }),
+              ],
+            }),
+            ToolNamespace.make({
+              name: "support",
+              description: "Support tools",
+              tools: [ToolDefinition.make({ name: "lookup", description: "support", inputSchema: {} })],
+            }),
+          ],
+        }),
+      )
+
+      expect(prepared.body.tools).toEqual([
+        {
+          type: "function",
+          name: "crm",
+          description: "Top-level CRM tool",
+          parameters: {},
+          strict: false,
+        },
+        {
+          type: "namespace",
+          name: "crm",
+          description: "CRM tools",
+          tools: [
+            { type: "function", name: "lookup", description: "new", parameters: {}, strict: false },
+            { type: "function", name: "search", description: "search", parameters: {}, strict: false },
+          ],
+        },
+        {
+          type: "namespace",
+          name: "support",
+          description: "Support tools",
+          tools: [{ type: "function", name: "lookup", description: "support", parameters: {}, strict: false }],
+        },
+      ])
+    }),
+  )
+
  it.effect("normalizes tool history before protocol lowering", () =>
    Effect.gen(function* () {
      const prepared = yield* compileRequest(
@@ -3,8 +3,11 @@ import { AIError, ImageInput, LanguageModel, LLM, LLMClient, Provider } from "@o
 import { Route, Protocol, WebSocketTransport } from "@opencode-ai/ai/route"
 import { Provider as ProviderSubpath } from "@opencode-ai/ai/provider"
 import {
+  Baseten,
  CloudflareAIGateway,
  CloudflareWorkersAI,
+  DeepSeek,
+  Fireworks,
  OpenAI,
  OpenAICompatible,
  OpenRouter,
@@ -48,7 +51,13 @@ describe("public exports", () => {
    expect(OpenAI.model).toBeFunction()
    expect(OpenAI.provider.responses).toBe(OpenAI.responses)
    expect(OpenAI.configure({ apiKey: "fixture" }).responses).toBeFunction()
-    expect(OpenAICompatible.deepseek.model).toBeFunction()
+    for (const provider of [Baseten, DeepSeek, Fireworks]) {
+      expect(provider.configure).toBeFunction()
+      expect(provider.model).toBeFunction()
+    }
+    for (const name of ["baseten", "cerebras", "deepinfra", "deepseek", "fireworks", "groq", "togetherai"]) {
+      expect(OpenAICompatible).not.toHaveProperty(name)
+    }
    expect(
      OpenAICompatibleResponses.configure({ baseURL: "https://responses.test/v1" }).model("fixture").route.id,
    ).toBe("openai-compatible-responses")
@@ -0,0 +1,30 @@
+import { Effect } from "effect"
+
+/**
+ * Run an effect with `process.env` overrides, restoring the previous values
+ * afterwards. Use for code that reads `process.env` directly, such as AWS SDK
+ * credential providers, where Effect `ConfigProvider` layers do not apply.
+ * `undefined` removes a variable for the duration of the effect.
+ */
+export const withProcessEnv =
+  (env: Record<string, string | undefined>) =>
+  <A, E, R>(effect: Effect.Effect<A, E, R>) =>
+    Effect.acquireUseRelease(
+      Effect.sync(() => {
+        const previous = Object.fromEntries(Object.keys(env).map((name) => [name, process.env[name]]))
+        apply(env)
+        return previous
+      }),
+      () => effect,
+      (previous) => Effect.sync(() => apply(previous)),
+    )
+
+const apply = (env: Record<string, string | undefined>) => {
+  for (const [name, value] of Object.entries(env)) {
+    if (value === undefined) {
+      delete process.env[name]
+      continue
+    }
+    process.env[name] = value
+  }
+}
@@ -1,5 +1,6 @@
 import { describe, expect, test } from "bun:test"
-import { CacheHint, LLM, LLMResponse } from "../src/index.js"
+import { Schema } from "effect"
+import { CacheHint, LLM, LLMResponse, ToolEntry, ToolNamespace } from "../src/index.js"
 import * as OpenAIChat from "../src/protocols/openai-chat.js"
 import * as OpenAIResponses from "../src/protocols/openai-responses.js"
 import {
@@ -17,6 +18,52 @@ const chatRoute = OpenAIChat.route
 const responsesRoute = OpenAIResponses.route

 describe("llm constructors", () => {
+  test("normalizes recursive tool namespaces", () => {
+    const request = LLM.request({
+      model: LanguageModel.make({ id: "fake-model", provider: "fake", route: responsesRoute }),
+      tools: [
+        {
+          type: "namespace",
+          name: "crm",
+          description: "Customer management",
+          tools: [
+            { name: "lookup", description: "Look up a customer", inputSchema: { type: "object" } },
+            {
+              type: "namespace",
+              name: "orders",
+              tools: [{ name: "list", description: "List orders", inputSchema: { type: "object" } }],
+            },
+          ],
+        },
+      ],
+    })
+
+    expect(request.tools[0]).toEqual({
+      type: "namespace",
+      name: "crm",
+      description: "Customer management",
+      tools: [
+        expect.objectContaining({ type: "tool", name: "lookup" }),
+        {
+          type: "namespace",
+          name: "orders",
+          description: undefined,
+          tools: [expect.objectContaining({ type: "tool", name: "list" })],
+        },
+      ],
+    })
+    expect(request.tools[0]).toEqual(
+      ToolNamespace.make({
+        name: "crm",
+        description: "Customer management",
+        tools: request.tools[0]!.type === "namespace" ? request.tools[0].tools : [],
+      }),
+    )
+    expect(Schema.decodeUnknownSync(ToolEntry)(Schema.encodeUnknownSync(ToolEntry)(request.tools[0]))).toEqual(
+      request.tools[0],
+    )
+  })
+
  test("builds canonical schema classes from ergonomic input", () => {
    const request = LLM.request({
      id: "req_1",
@@ -0,0 +1,129 @@
+import { Effect } from "effect"
+import {
+  CompactionCheckpointResponse,
+  CompactionResponse,
+  LanguageModel,
+  LLM,
+  LLMClient,
+  LLMRequest,
+} from "../../src/index.js"
+import {
+  Anthropic,
+  Azure,
+  AmazonBedrock,
+  AmazonBedrockMantle,
+  OpenAI,
+  OpenAICompatibleResponses,
+  XAI,
+} from "../../src/providers.js"
+import type { WebSocketChannelExecutor } from "../../src/route.js"
+import type { RoutePatch } from "../../src/route/client.js"
+import type { OpenAIResponsesBody } from "../../src/protocols/openai-responses.js"
+import type { Prepared } from "../../src/protocols/open-responses-channel.js"
+
+declare const webSocket: WebSocketChannelExecutor
+const model = OpenAI.configure().responses("fixture")
+const request = LLM.request({ model, prompt: "hello" })
+LLMClient.compact(request, { mechanism: "endpoint" }).pipe(Effect.map((result) => result satisfies CompactionResponse))
+LLMClient.compact(request, { mechanism: "trigger", webSocket }).pipe(
+  Effect.map((result) => {
+    result satisfies CompactionCheckpointResponse
+    result.checkpoint.encrypted satisfies string
+    result.responseID satisfies string
+    // @ts-expect-error A trigger does not return replacement history.
+    result.replacement
+  }),
+)
+// @ts-expect-error Endpoint compaction does not accept a WebSocket executor.
+LLMClient.compact(request, { mechanism: "endpoint", webSocket })
+// @ts-expect-error Omitting mechanism selects the HTTP endpoint.
+LLMClient.compact(request, { webSocket })
+// @ts-expect-error Unknown mechanisms do not have a permissive fallback overload.
+LLMClient.compact(request, { mechanism: "other" })
+
+for (const selected of [
+  model,
+  OpenAI.model("fixture", {}),
+  model.route.with({ headers: { fixture: "test" } }).model({ id: "fixture" }),
+  LanguageModel.make(LanguageModel.input(model)),
+  LanguageModel.update(model, { defaults: { generation: { maxTokens: 100 } } }),
+]) {
+  LLMClient.compact(LLM.request({ model: selected }), { mechanism: "trigger" })
+}
+LLMClient.compact(new LLMRequest(LLMRequest.input(request)), { mechanism: "trigger" })
+LLMClient.compact(LLMRequest.update(request, { messages: [] }), { mechanism: "trigger" })
+
+const azure = Azure.configure({ resourceName: "fixture" }).responses("fixture")
+const xai = XAI.configure().responses("fixture")
+// @ts-expect-error Azure must not inherit OpenAI's trigger operation.
+LLMClient.compact(LLM.request({ model: azure }), { mechanism: "trigger" })
+LLMClient.compact(LLM.request({ model: Azure.responsesModel("fixture", { resourceName: "fixture" }) }), {
+  // @ts-expect-error Azure's package entrypoint must preserve its narrower capability.
+  mechanism: "trigger",
+})
+// @ts-expect-error xAI endpoint support does not imply trigger support.
+LLMClient.compact(LLM.request({ model: xai }), { mechanism: "trigger" })
+// @ts-expect-error xAI's package entrypoint must preserve its narrower capability.
+LLMClient.compact(LLM.request({ model: XAI.model("fixture", {}) }), { mechanism: "trigger" })
+
+const unsupported = {
+  bedrock: LLM.request({ model: AmazonBedrock.configure().model("fixture") }),
+  mantle: LLM.request({ model: AmazonBedrockMantle.configure().responses("fixture") }),
+  anthropic: LLM.request({ model: Anthropic.configure().model("fixture") }),
+  openai: LLM.request({ model: OpenAI.configure().chat("fixture") }),
+  azure: LLM.request({ model: Azure.configure({ resourceName: "fixture" }).chat("fixture") }),
+  xai: LLM.request({ model: XAI.configure().chat("fixture") }),
+  compatible: LLM.request({
+    model: OpenAICompatibleResponses.configure({ baseURL: "https://example.com" }).model("fixture"),
+  }),
+}
+// @ts-expect-error Bedrock does not expose trigger compaction.
+LLMClient.compact(unsupported.bedrock, { mechanism: "trigger" })
+// @ts-expect-error Mantle must not inherit trigger support from OpenAI's protocol.
+LLMClient.compact(unsupported.mantle, { mechanism: "trigger" })
+// @ts-expect-error Anthropic does not expose trigger compaction.
+LLMClient.compact(unsupported.anthropic, { mechanism: "trigger" })
+// @ts-expect-error OpenAI Chat does not expose trigger compaction.
+LLMClient.compact(unsupported.openai, { mechanism: "trigger" })
+// @ts-expect-error Azure Chat does not expose trigger compaction.
+LLMClient.compact(unsupported.azure, { mechanism: "trigger" })
+// @ts-expect-error xAI Chat does not expose trigger compaction.
+LLMClient.compact(unsupported.xai, { mechanism: "trigger" })
+// @ts-expect-error Generic protocol compatibility does not grant trigger support.
+LLMClient.compact(unsupported.compatible, { mechanism: "trigger" })
+// @ts-expect-error Changing the model replaces its capability.
+LLMClient.compact(LLMRequest.update(request, { model: azure }), { mechanism: "trigger" })
+// @ts-expect-error Changing the route replaces its capability.
+LLMClient.compact(LLM.request({ model: LanguageModel.update(model, { route: azure.route }) }), { mechanism: "trigger" })
+LLMClient.compact(
+  LLM.request({
+    model: model.route.with({ compact: { endpoint: model.route.compact.endpoint } }).model({ id: "fixture" }),
+  }),
+  // @ts-expect-error Replacing route operations does not retain the old trigger capability.
+  { mechanism: "trigger" },
+)
+
+declare const dynamic: LLMRequest
+declare const patch: Partial<LLMRequest.Input>
+declare const routePatch: RoutePatch<OpenAIResponsesBody, Prepared>
+// @ts-expect-error A dynamic operation override cannot preserve trigger support.
+LLMClient.compact(LLM.request({ model: model.route.with(routePatch).model({ id: "fixture" }) }), {
+  mechanism: "trigger",
+})
+// @ts-expect-error Explicitly removing operations removes trigger support.
+LLMClient.compact(LLM.request({ model: model.route.with({ compact: undefined }).model({ id: "fixture" }) }), {
+  mechanism: "trigger",
+})
+// @ts-expect-error Dynamic models must be narrowed.
+LLMClient.compact(dynamic, { mechanism: "trigger" })
+if (LLMClient.canCompact(dynamic, { mechanism: "trigger" })) {
+  LLMClient.compact(dynamic, { mechanism: "trigger" })
+  LLMClient.Service.use((client) => client.compact(dynamic, { mechanism: "trigger" }))
+}
+if (LLMClient.canCompact(dynamic)) {
+  LLMClient.compact(dynamic)
+  // @ts-expect-error Endpoint narrowing does not grant trigger support.
+  LLMClient.compact(dynamic, { mechanism: "trigger" })
+}
+// @ts-expect-error A dynamic model override cannot preserve trigger support.
+LLMClient.compact(LLMRequest.update(request, patch), { mechanism: "trigger" })
@@ -1,7 +1,7 @@
 import { LLM } from "../../src/index.js"
 import { OpenAICompatible } from "../../src/providers.js"

-const model = OpenAICompatible.deepseek.model("deepseek-chat")
+const model = OpenAICompatible.configure({ baseURL: "https://compatible.example/v1" }).model("test-model")

 LLM.request({ model, prompt: "Hello", providerOptions: { store: false } })

@@ -32,6 +32,11 @@ describe("provider package entrypoints", () => {
      import("@opencode-ai/ai/providers/cerebras"),
      import("@opencode-ai/ai/providers/deepinfra"),
      import("@opencode-ai/ai/providers/groq"),
+      import("@opencode-ai/ai/providers/baseten"),
+      import("@opencode-ai/ai/providers/deepseek"),
+      import("@opencode-ai/ai/providers/fireworks"),
+      import("@opencode-ai/ai/providers/cloudflare-ai-gateway"),
+      import("@opencode-ai/ai/providers/cloudflare-workers-ai"),
    ])

    for (const module of modules) expect(module.model).toBeFunction()
@@ -60,6 +65,27 @@ describe("provider package entrypoints", () => {
    expect(deepinfra.route.defaults.http?.body).toEqual(settings.body)
  })

+  test("maps Cloudflare package settings onto provider-owned models", async () => {
+    const modules = await Promise.all([
+      import("@opencode-ai/ai/providers/cloudflare-ai-gateway"),
+      import("@opencode-ai/ai/providers/cloudflare-workers-ai"),
+    ])
+    for (const provider of modules) {
+      const selected = provider.model("provider-model", {
+        accountId: "account",
+        apiKey: "fixture",
+        headers: { "x-application": "opencode" },
+        body: { custom: true },
+        providerOptions: { reasoningEffort: "high" },
+      })
+      expect(selected.provider).toBe(provider.id)
+      expect(selected.route.endpoint.baseURL).toBe(provider.baseURL({ accountId: "account" }))
+      expect(selected.route.defaults.headers).toEqual({ "x-application": "opencode" })
+      expect(selected.route.defaults.http?.body).toEqual({ custom: true })
+      expect(selected.route.defaults.providerOptions).toEqual({ reasoningEffort: "high" })
+    }
+  })
+
  test("maps OpenRouter and xAI package settings onto executable models", async () => {
    const OpenRouter = await import("@opencode-ai/ai/providers/openrouter")
    const XAI = await import("@opencode-ai/ai/providers/xai")
@@ -0,0 +1,119 @@
+import { describe, expect } from "bun:test"
+import { Effect } from "effect"
+import { CacheHint, LLM, Message, ToolCallPart } from "../../src/index.js"
+import { AmazonBedrock } from "../../src/providers.js"
+import { compileRequest } from "../../src/route/client.js"
+import { it } from "../lib/effect.js"
+
+const bedrock = AmazonBedrock.configure({ apiKey: "fixture" })
+
+describe("Bedrock Converse cache policy", () => {
+  for (const id of [
+    "deepseek.r1-v1:0",
+    "meta.llama3-3-70b-instruct-v1:0",
+    "mistral.mistral-large-2402-v1:0",
+    "qwen.qwen3-coder-480b-a35b-v1:0",
+    "openai.gpt-oss-120b-1:0",
+    "cohere.command-r-v1:0",
+    "anthropic.claude-instant-v1",
+    "anthropic.claude-v1",
+    "anthropic.claude-v2",
+    "anthropic.claude-v2:1",
+    "anthropic.claude-3-haiku-20240307-v1:0",
+    "anthropic.claude-3-sonnet-20240229-v1:0",
+    "anthropic.claude-3-opus-20240229-v1:0",
+    "anthropic.claude-3-5-sonnet-20240620-v1:0",
+    "amazon.nova-lite-v1:0",
+    "global.amazon.nova-2-lite-v1:0",
+    "custom-model",
+    "arn:aws:bedrock:us-east-1:123456789012:application-inference-profile/abc123",
+  ]) {
+    for (const policy of [undefined, "auto", "none", { tools: true, system: true, messages: { tail: 3 } }] as const) {
+      it.effect(`omits checkpoints for ${id} (${JSON.stringify(policy)})`, () =>
+        Effect.gen(function* () {
+          // Exercise both automatic placement and manual hints at every lowering site.
+          const cache = policy === "none" ? new CacheHint({ type: "ephemeral", ttlSeconds: 3600 }) : undefined
+          const prepared = yield* compileRequest(
+            LLM.request({
+              model: bedrock.model(id),
+              cache: policy,
+              system: [{ type: "text", text: "System prefix", cache }],
+              tools: [{ name: "lookup", description: "Lookup", inputSchema: { type: "object" }, cache }],
+              messages: [
+                Message.user([{ type: "text", text: "Question", cache }]),
+                Message.system([{ type: "text", text: "Update", cache }]),
+                Message.assistant([
+                  { type: "text", text: "Answer", cache },
+                  { type: "reasoning", text: "Unsigned reasoning", cache },
+                  ToolCallPart.make({ id: "call_1", name: "lookup", input: {} }),
+                ]),
+                Message.tool({ id: "call_1", name: "lookup", result: "Result", cache }),
+              ],
+            }),
+          )
+
+          expect(JSON.stringify(prepared.body)).not.toContain("cachePoint")
+          expect(prepared.body).toMatchObject({
+            modelId: id,
+            system: [{ text: "System prefix" }],
+            toolConfig: { tools: [{ toolSpec: { name: "lookup" } }] },
+            messages: [
+              { role: "user", content: [{ text: "Question" }, { text: "<system-update>\nUpdate\n</system-update>" }] },
+              {
+                role: "assistant",
+                content: [{ text: "Answer" }, { text: "Unsigned reasoning" }, { toolUse: { name: "lookup" } }],
+              },
+              { role: "user", content: [{ toolResult: { content: [{ json: "Result" }] } }] },
+            ],
+          })
+        }),
+      )
+    }
+  }
+
+  for (const [id, ttl] of [
+    ["anthropic.claude-3-5-sonnet-20241022-v2:0", undefined],
+    ["us.anthropic.claude-3-5-haiku-20241022-v1:0", undefined],
+    ["eu.anthropic.claude-3-7-sonnet-20250219-v1:0", undefined],
+    ["apac.anthropic.claude-sonnet-4-20250514-v1:0", undefined],
+    ["anthropic.claude-opus-4-20250514-v1:0", undefined],
+    ["anthropic.claude-opus-4-1-20250805-v1:0", undefined],
+    ["anthropic.claude-sonnet-4-5-20250929-v1:0", "1h"],
+    ["global.anthropic.claude-sonnet-99", "1h"],
+    ["anthropic.claude-new-family-99", "1h"],
+    ["arn:aws:bedrock:us-east-1::foundation-model/anthropic.claude-sonnet-4-6", "1h"],
+    ["arn:aws:bedrock:us-east-1:123456789012:inference-profile/us.anthropic.claude-sonnet-4-6", "1h"],
+  ] as const) {
+    for (const ttlSeconds of [undefined, 3600]) {
+      it.effect(`preserves Claude checkpoints for ${id} (TTL: ${ttlSeconds ?? "default"})`, () =>
+        Effect.gen(function* () {
+          const prepared = yield* compileRequest(
+            LLM.request({
+              model: bedrock.model(id),
+              system: [
+                { type: "text", text: "Agent" },
+                { type: "text", text: "Project" },
+              ],
+              tools: [{ name: "lookup", description: "Lookup", inputSchema: { type: "object" } }],
+              prompt: "Question",
+              cache:
+                ttlSeconds === undefined ? undefined : { tools: true, system: true, messages: { tail: 1 }, ttlSeconds },
+            }),
+          )
+          const marker = {
+            cachePoint: ttlSeconds === undefined || ttl === undefined ? { type: "default" } : { type: "default", ttl },
+          }
+
+          expect(prepared.body).toMatchObject({
+            modelId: id,
+            toolConfig: { tools: [{ toolSpec: { name: "lookup" } }, marker] },
+            system: [{ text: "Agent" }, marker, { text: "Project" }, marker],
+            messages: [{ role: "user", content: [{ text: "Question" }, marker] }],
+          })
+          if (ttlSeconds === undefined || ttl === undefined)
+            expect(JSON.stringify(prepared.body)).not.toContain('"ttl"')
+        }),
+      )
+    }
+  }
+})
@@ -2,9 +2,11 @@ import { EventStreamCodec } from "@smithy/eventstream-codec"
 import { fromUtf8, toUtf8 } from "@smithy/util-utf8"
 import { describe, expect } from "bun:test"
 import { Effect, Encoding, Ref, Stream } from "effect"
+import { HttpClientRequest } from "effect/unstable/http"
 import {
  CacheHint,
  GenerationOptions,
+  type LanguageModel,
  LLM,
  LLMEvent,
  LLMRequest,
@@ -18,7 +20,8 @@ import { compileRequest } from "../../src/route/client.js"
 import { AmazonBedrock } from "../../src/providers.js"
 import * as BedrockConverse from "../../src/protocols/bedrock-converse.js"
 import { it } from "../lib/effect.js"
-import { fixedResponse } from "../lib/http.js"
+import { withProcessEnv } from "../lib/env.js"
+import { dynamicResponse, fixedResponse } from "../lib/http.js"
 import {
  eventSummary,
  expectWeatherToolLoop,
@@ -95,10 +98,45 @@ const fixedByteChunks = (...chunks: ReadonlyArray<Uint8Array>) =>
    { headers: { "content-type": "application/vnd.amazon.eventstream" } },
  )

+// Clears every input the AWS default chain reads so tests do not pick up the
+// developer's profiles, SSO cache, or instance metadata.
+const noAmbientAWS = {
+  AWS_ACCESS_KEY_ID: undefined,
+  AWS_SECRET_ACCESS_KEY: undefined,
+  AWS_SESSION_TOKEN: undefined,
+  AWS_BEARER_TOKEN_BEDROCK: undefined,
+  AWS_PROFILE: undefined,
+  AWS_REGION: undefined,
+  AWS_DEFAULT_REGION: undefined,
+  AWS_WEB_IDENTITY_TOKEN_FILE: undefined,
+  AWS_CONTAINER_CREDENTIALS_RELATIVE_URI: undefined,
+  AWS_CONTAINER_CREDENTIALS_FULL_URI: undefined,
+  AWS_EC2_METADATA_DISABLED: "true",
+}
+
+const captureHeaders = (target: LanguageModel) =>
+  Effect.gen(function* () {
+    const seen: Array<Headers> = []
+    yield* LLMClient.generate(LLMRequest.update(baseRequest, { model: target })).pipe(
+      Effect.provide(
+        dynamicResponse((input) =>
+          Effect.gen(function* () {
+            const request = yield* HttpClientRequest.toWeb(input.request)
+            seen.push(request.headers)
+            return input.respond(eventStreamBody(["messageStop", { stopReason: "end_turn" }]), {
+              headers: { "content-type": "application/vnd.amazon.eventstream" },
+            })
+          }),
+        ),
+      ),
+    )
+    return seen[0]!
+  })
+
 const model = AmazonBedrock.configure({
  baseURL: "https://bedrock-runtime.test",
  apiKey: "test-bearer",
-}).model("anthropic.claude-3-5-sonnet-20240620-v1:0")
+}).model("anthropic.claude-sonnet-4-5-20250929-v1:0")

 const baseRequest = LLM.request({
  id: "req_1",
@@ -117,7 +155,7 @@ describe("Bedrock Converse route", () => {
      const prepared = yield* compileRequest(baseRequest)

      expect(prepared.body).toEqual({
-        modelId: "anthropic.claude-3-5-sonnet-20240620-v1:0",
+        modelId: "anthropic.claude-sonnet-4-5-20250929-v1:0",
        system: [{ text: "You are concise." }],
        messages: [{ role: "user", content: [{ text: "Say hello." }] }],
        inferenceConfig: { maxTokens: 64, temperature: 0 },
@@ -713,9 +751,10 @@ describe("Bedrock Converse route", () => {
    }),
  )

-  it.effect("ignores late tool deltas after contentBlockStop", () =>
+  it.effect("ignores tool deltas without an open tool block", () =>
    Effect.gen(function* () {
      const body = eventStreamBody(
+        ["contentBlockDelta", { contentBlockIndex: 5, delta: { toolUse: { input: "{}" } } }],
        [
          "contentBlockStart",
          {
@@ -745,27 +784,6 @@ describe("Bedrock Converse route", () => {
    }),
  )

-  it.effect("rejects tool deltas without contentBlockStart", () =>
-    Effect.gen(function* () {
-      const error = yield* LLMClient.generate(baseRequest).pipe(
-        Effect.provide(
-          fixedBytes(
-            eventStreamBody(
-              ["contentBlockDelta", { contentBlockIndex: 0, delta: { toolUse: { input: "{}" } } }],
-              ["messageStop", { stopReason: "tool_use" }],
-            ),
-          ),
-        ),
-        Effect.flip,
-      )
-
-      expect(error).toMatchObject({
-        reason: { _tag: "InvalidProviderOutput" },
-        message: "Bedrock Converse tool delta is missing its tool call",
-      })
-    }),
-  )
-
  it.effect("recovers incomplete tool input at finalization", () =>
    Effect.gen(function* () {
      const body = eventStreamBody(
@@ -1331,35 +1349,137 @@ describe("Bedrock Converse route", () => {
    }),
  )

-  it.effect("rejects requests with no auth path", () =>
+  it.effect("fails with an authentication error when the default chain cannot resolve credentials", () =>
    Effect.gen(function* () {
      const unsignedModel = AmazonBedrock.configure({
        baseURL: "https://bedrock-runtime.test",
+        profile: "opencode-test-missing-profile",
      }).model("anthropic.claude-3-5-sonnet-20240620-v1:0")
      const error = yield* LLMClient.generate(LLMRequest.update(baseRequest, { model: unsignedModel })).pipe(
        Effect.provide(fixedBytes(eventStreamBody(["messageStop", { stopReason: "end_turn" }]))),
        Effect.flip,
      )

-      expect(error.message).toContain("Bedrock Converse requires either route bearer auth or AWS credentials")
-    }),
+      expect(error.reason._tag).toBe("Authentication")
+      expect(error.message).toContain("AWS default credential chain failed")
+    }).pipe(withProcessEnv(noAmbientAWS)),
  )

-  it.effect("signs requests with SigV4 when AWS credentials are provided (deterministic plumbing check)", () =>
+  it.effect("signs with static credentials using the configured region for the SigV4 scope", () =>
    Effect.gen(function* () {
      const signed = AmazonBedrock.configure({
        baseURL: "https://bedrock-runtime.test",
+        region: "eu-west-1",
        credentials: {
          region: "us-east-1",
          accessKeyId: "AKIAIOSFODNN7EXAMPLE",
          secretAccessKey: "wJalrXUtnFEMI/K7MDENG/bPxRfiCYEXAMPLEKEY",
        },
      }).model("anthropic.claude-3-5-sonnet-20240620-v1:0")
-      const prepared = yield* compileRequest(LLMRequest.update(baseRequest, { model: signed }))
+      const headers = yield* captureHeaders(signed)

-      expect(prepared.route).toBe("bedrock-converse")
-      expect(prepared.model).toBe(signed)
-    }),
+      expect(headers.get("authorization")).toContain("Credential=AKIAIOSFODNN7EXAMPLE/")
+      expect(headers.get("authorization")).toContain("/eu-west-1/bedrock/aws4_request")
+      expect(headers.get("x-amz-security-token")).toBeNull()
+    }).pipe(withProcessEnv(noAmbientAWS)),
+  )
+
+  it.effect("resolves SigV4 credentials through the AWS default chain on every request", () =>
+    Effect.gen(function* () {
+      const chained = AmazonBedrock.configure({ baseURL: "https://bedrock-runtime.test", region: "us-west-2" }).model(
+        "anthropic.claude-3-5-sonnet-20240620-v1:0",
+      )
+      const headers = yield* captureHeaders(chained)
+
+      expect(headers.get("authorization")).toContain("Credential=AKIACHAINEXAMPLE/")
+      expect(headers.get("authorization")).toContain("/us-west-2/bedrock/aws4_request")
+      expect(headers.get("x-amz-security-token")).toBe("chain-session-token")
+    }).pipe(
+      withProcessEnv({
+        ...noAmbientAWS,
+        AWS_ACCESS_KEY_ID: "AKIACHAINEXAMPLE",
+        AWS_SECRET_ACCESS_KEY: "chain-secret",
+        AWS_SESSION_TOKEN: "chain-session-token",
+      }),
+    ),
+  )
+
+  it.effect("re-resolves rotated chain credentials on the next request without rebuilding the model", () =>
+    Effect.gen(function* () {
+      const chained = AmazonBedrock.configure({ baseURL: "https://bedrock-runtime.test", region: "us-west-2" }).model(
+        "anthropic.claude-3-5-sonnet-20240620-v1:0",
+      )
+      const first = yield* captureHeaders(chained)
+      const second = yield* captureHeaders(chained).pipe(
+        withProcessEnv({ AWS_ACCESS_KEY_ID: "AKIAROTATEDEXAMPLE", AWS_SECRET_ACCESS_KEY: "rotated-secret" }),
+      )
+
+      expect(first.get("authorization")).toContain("Credential=AKIACHAINEXAMPLE/")
+      expect(second.get("authorization")).toContain("Credential=AKIAROTATEDEXAMPLE/")
+    }).pipe(
+      withProcessEnv({
+        ...noAmbientAWS,
+        AWS_ACCESS_KEY_ID: "AKIACHAINEXAMPLE",
+        AWS_SECRET_ACCESS_KEY: "chain-secret",
+      }),
+    ),
+  )
+
+  it.effect("prefers AWS_BEARER_TOKEN_BEDROCK over the credential chain", () =>
+    Effect.gen(function* () {
+      const bearer = AmazonBedrock.configure({ baseURL: "https://bedrock-runtime.test" }).model(
+        "anthropic.claude-3-5-sonnet-20240620-v1:0",
+      )
+      const headers = yield* captureHeaders(bearer)
+
+      expect(headers.get("authorization")).toBe("Bearer env-bearer-token")
+    }).pipe(
+      withProcessEnv({
+        ...noAmbientAWS,
+        AWS_BEARER_TOKEN_BEDROCK: "env-bearer-token",
+        AWS_ACCESS_KEY_ID: "AKIACHAINEXAMPLE",
+        AWS_SECRET_ACCESS_KEY: "chain-secret",
+      }),
+    ),
+  )
+
+  it.effect("auth: sigv4 ignores an ambient bearer token from configure and settings", () =>
+    Effect.gen(function* () {
+      const configured = AmazonBedrock.configure({ auth: "sigv4", baseURL: "https://bedrock-runtime.test" }).model(
+        "anthropic.claude-3-5-sonnet-20240620-v1:0",
+      )
+      const fromSettings = AmazonBedrock.model("anthropic.claude-3-5-sonnet-20240620-v1:0", {
+        auth: "sigv4",
+        baseURL: "https://bedrock-runtime.test",
+      })
+
+      for (const target of [configured, fromSettings]) {
+        const headers = yield* captureHeaders(target)
+        expect(headers.get("authorization")).toContain("Credential=AKIACHAINEXAMPLE/")
+        expect(headers.get("authorization")).toContain("/ap-southeast-2/bedrock/aws4_request")
+      }
+      expect(() => AmazonBedrock.configure({ auth: "sigv4", apiKey: "k" })).toThrow("does not accept apiKey")
+    }).pipe(
+      withProcessEnv({
+        ...noAmbientAWS,
+        AWS_BEARER_TOKEN_BEDROCK: "env-bearer-token",
+        AWS_REGION: "ap-southeast-2",
+        AWS_ACCESS_KEY_ID: "AKIACHAINEXAMPLE",
+        AWS_SECRET_ACCESS_KEY: "chain-secret",
+      }),
+    ),
+  )
+
+  it.effect("derives the endpoint region from AWS_REGION then AWS_DEFAULT_REGION", () =>
+    Effect.gen(function* () {
+      const fromRegion = AmazonBedrock.configure({ apiKey: "k" }).model("anthropic.claude-3-5-sonnet-20240620-v1:0")
+      expect(fromRegion.route.endpoint.baseURL).toBe("https://bedrock-runtime.eu-central-1.amazonaws.com")
+
+      const fromDefault = yield* Effect.sync(() =>
+        AmazonBedrock.configure({ apiKey: "k" }).model("anthropic.claude-3-5-sonnet-20240620-v1:0"),
+      ).pipe(withProcessEnv({ AWS_REGION: undefined }))
+      expect(fromDefault.route.endpoint.baseURL).toBe("https://bedrock-runtime.us-gov-west-1.amazonaws.com")
+    }).pipe(withProcessEnv({ ...noAmbientAWS, AWS_REGION: "eu-central-1", AWS_DEFAULT_REGION: "us-gov-west-1" })),
  )

  it.effect("emits cachePoint markers after system, user-text, and assistant-text with cache hints", () =>
@@ -7,6 +7,7 @@ import { model } from "../../src/providers/amazon-bedrock/mantle.js"
 import { OpenAIResponses } from "../../src/protocols/openai-responses.js"
 import { compileRequest, LLMClient } from "../../src/route/client.js"
 import { it } from "../lib/effect.js"
+import { withProcessEnv } from "../lib/env.js"
 import { dynamicResponse, fixedResponse } from "../lib/http.js"
 import { sseEvents } from "../lib/sse.js"
 import { recordedTests } from "../recorded-test.js"
@@ -82,6 +83,36 @@ describe("Amazon Bedrock Mantle provider", () => {
    }),
  )

+  it.effect("signs with the Mantle service using default-chain credentials", () =>
+    Effect.gen(function* () {
+      const seen: Array<string | undefined> = []
+      const model = AmazonBedrockMantle.configure({ region: "us-west-1" }).responses("openai.gpt-oss-120b")
+      yield* LLMClient.generate(LLM.request({ model, prompt: "Hi" })).pipe(
+        Effect.provide(
+          dynamicResponse((input) =>
+            Effect.gen(function* () {
+              const request = yield* HttpClientRequest.toWeb(input.request)
+              seen.push(request.headers.get("authorization") ?? undefined)
+              return input.respond("", { headers: { "content-type": "text/event-stream" } })
+            }),
+          ),
+        ),
+        Effect.flip,
+      )
+
+      expect(seen[0]).toContain("Credential=AKIACHAINEXAMPLE/")
+      expect(seen[0]).toContain("/us-west-1/bedrock-mantle/aws4_request")
+    }).pipe(
+      withProcessEnv({
+        AWS_BEARER_TOKEN_BEDROCK: undefined,
+        AWS_PROFILE: undefined,
+        AWS_ACCESS_KEY_ID: "AKIACHAINEXAMPLE",
+        AWS_SECRET_ACCESS_KEY: "chain-secret",
+        AWS_SESSION_TOKEN: undefined,
+      }),
+    ),
+  )
+
  it.effect("supports bearer authentication and custom base URLs", () =>
    Effect.gen(function* () {
      const seen: Array<{ readonly url: string; readonly authorization: string | undefined }> = []
@@ -0,0 +1,243 @@
+import { expect } from "bun:test"
+import { Effect, Stream } from "effect"
+import { LLM, LLMClient, LLMRequest } from "../../src/index.js"
+import { Azure, OpenAI } from "../../src/providers.js"
+import { testEffect } from "../lib/effect.js"
+import { dynamicResponse, fixedResponse } from "../lib/http.js"
+import { sseEvents } from "../lib/sse.js"
+
+const checkpoint = { type: "compaction", id: "cmp_1", encrypted_content: "opaque" }
+const request = LLM.request({ model: OpenAI.configure({ apiKey: "fixture" }).responses("fixture"), prompt: "hello" })
+const trigger = { mechanism: "trigger" } as const
+
+testEffect(
+  dynamicResponse(({ request, text, respond }) =>
+    Effect.sync(() => {
+      expect(new URL(request.url).pathname).toBe("/v1/responses")
+      expect(new URL(request.url).searchParams.get("trace")).toBe("request")
+      expect(request.headers.authorization).toBe("Bearer fixture")
+      expect(request.headers["chatgpt-account-id"]).toBe("fixture-account")
+      expect(request.headers["x-codex-beta-features"]).toBe("remote_compaction_v2")
+      const body = JSON.parse(text)
+      expect(body).toMatchObject({
+        model: "fixture",
+        stream: true,
+        store: false,
+        instructions: "Keep the context",
+        parallel_tool_calls: true,
+        prompt_cache_key: "session-key",
+        service_tier: "priority",
+        reasoning: { effort: "high", summary: "auto" },
+        prompt_cache_retention: "24h",
+        prompt_cache_options: { mode: "session", ttl: "1h" },
+        input: [{ role: "user", content: [{ type: "input_text", text: "hello" }] }, { type: "compaction_trigger" }],
+      })
+      expect(body.tools).toHaveLength(1)
+      expect(body.tools[0].name).toBe("lookup")
+      expect(body.tool_choice).toBeUndefined()
+      expect(body.context_management).toBeUndefined()
+      expect(body.text).toBeUndefined()
+      expect(body.max_output_tokens).toBeUndefined()
+      expect(body.previous_response_id).toBeUndefined()
+      return respond(
+        sseEvents({
+          type: "response.completed",
+          response: {
+            id: "resp_1",
+            output: [checkpoint],
+            usage: {
+              input_tokens: 100,
+              input_tokens_details: { cached_tokens: 40 },
+              output_tokens: 5,
+              total_tokens: 105,
+            },
+          },
+        }),
+        { headers: { "content-type": "text/event-stream" } },
+      )
+    }),
+  ),
+).effect("trigger uses normal request preparation, configured deployment, and supplied subscription headers", () =>
+  Effect.gen(function* () {
+    const calls: string[] = []
+    const input = LLM.request({
+      model: request.model,
+      system: "Keep the context",
+      prompt: "hello",
+      promptCacheKey: "session-key",
+      tools: [{ name: "lookup", description: "Lookup", inputSchema: { type: "object", properties: {} } }],
+      toolChoice: { type: "tool", name: "lookup" },
+      generation: { maxTokens: 1 },
+      providerOptions: {
+        store: true,
+        reasoningEffort: "high",
+        reasoningSummary: "auto",
+        contextManagement: [{ type: "compaction" }],
+      },
+      http: {
+        headers: { "chatgpt-account-id": "fixture-account", "x-codex-beta-features": "remote_compaction_v2" },
+        query: { trace: "request" },
+        body: {
+          service_tier: "priority",
+          prompt_cache_retention: "24h",
+          prompt_cache_options: { mode: "session", ttl: "1h" },
+          store: true,
+          stream: false,
+          text: { format: { type: "json_object" } },
+          tool_choice: "required",
+        },
+      },
+    })
+    const original = LLMRequest.input(input)
+    const result = yield* LLMClient.compact(input, {
+      ...trigger,
+      http: (request, next) => {
+        calls.push("http")
+        return next(request)
+      },
+    })
+    expect(result.checkpoint).toMatchObject({
+      type: "compaction",
+      provider: "openai",
+      id: "cmp_1",
+      encrypted: "opaque",
+    })
+    expect(result.responseID).toBe("resp_1")
+    expect(result.usage).toMatchObject({
+      inputTokens: 100,
+      outputTokens: 5,
+      totalTokens: 105,
+      cacheReadInputTokens: 40,
+    })
+    expect(LLMRequest.input(input)).toEqual(original)
+    expect(calls).toEqual(["http"])
+  }),
+)
+
+const idless = { type: "compaction", encrypted_content: "opaque" }
+testEffect(
+  fixedResponse(
+    sseEvents(
+      {
+        type: "response.output_item.done",
+        output_index: 0,
+        item: { type: "function_call", id: "fc_1", name: "unexpected", arguments: "not JSON" },
+      },
+      { type: "response.output_text.delta", delta: "do not expose this" },
+      { type: "response.output_item.added", output_index: 1, item: { type: "compaction" } },
+      { type: "response.output_item.done", output_index: 1, item: idless },
+      { type: "response.output_item.done", output_index: 1, item: idless },
+      {
+        type: "response.completed",
+        response: {
+          id: "resp_1",
+          output: [{ type: "function_call", id: "fc_1", name: "unexpected", arguments: "not JSON" }, idless],
+        },
+      },
+    ),
+  ),
+).effect("correlates an ID-less checkpoint across repeated events and ignores other output", () =>
+  Effect.gen(function* () {
+    const result = yield* LLMClient.compact(request, trigger)
+    expect(result.checkpoint.encrypted).toBe("opaque")
+    expect(result.checkpoint.id).toBeString()
+    expect("message" in result).toBe(false)
+  }),
+)
+
+for (const [name, events] of Object.entries({
+  missing: [{ type: "response.completed", response: { id: "resp_1", output: [] } }],
+  multiple: [
+    { type: "response.completed", response: { id: "resp_1", output: [checkpoint, { ...checkpoint, id: "cmp_2" }] } },
+  ],
+  duplicateSlots: [{ type: "response.completed", response: { id: "resp_1", output: [checkpoint, checkpoint] } }],
+  malformed: [
+    { type: "response.completed", response: { id: "resp_1", output: [{ type: "compaction", id: "cmp_1" }] } },
+  ],
+  noResponseID: [{ type: "response.completed", response: { output: [checkpoint] } }],
+  incomplete: [
+    { type: "response.output_item.done", item: checkpoint },
+    { type: "response.incomplete", response: { id: "resp_1", incomplete_details: { reason: "max_output_tokens" } } },
+  ],
+  failed: [
+    { type: "response.output_item.done", item: checkpoint },
+    { type: "response.failed", response: { id: "resp_1", error: { code: "server_error", message: "failed" } } },
+  ],
+})) {
+  const wire = events.map((event) => ({ ...event, fixture_extra: "preserved" }))
+  testEffect(
+    fixedResponse(sseEvents(...wire), { headers: { "content-type": "text/event-stream", "x-fixture": "preserved" } }),
+  ).effect(`rejects ${name} checkpoint response and preserves original error context`, () =>
+    Effect.gen(function* () {
+      const error = yield* LLMClient.compact(request, trigger).pipe(Effect.flip)
+      expect(error.reason.body).toBe(JSON.stringify(wire.at(-1)))
+      expect(error.reason.http).toMatchObject({ status: 200, headers: { "x-fixture": "preserved" } })
+    }),
+  )
+}
+
+testEffect(fixedResponse(sseEvents({ type: "response.output_item.done", item: checkpoint }))).effect(
+  "rejects clean EOF without response.completed",
+  () =>
+    Effect.gen(function* () {
+      const error = yield* LLMClient.compact(request, trigger).pipe(Effect.flip)
+      expect(error.reason._tag).toBe("InvalidProviderOutput")
+    }),
+)
+for (const body of [{ input: [] }, { previous_response_id: "stale" }]) {
+  testEffect(dynamicResponse(() => Effect.die("Must reject before sending"))).effect(
+    `rejects caller-supplied ${Object.keys(body)[0]} before sending trigger`,
+    () =>
+      Effect.gen(function* () {
+        const error = yield* LLMClient.compact(LLMRequest.update(request, { http: { body } }), trigger).pipe(
+          Effect.flip,
+        )
+        expect(error.reason._tag).toBe("InvalidRequest")
+      }),
+  )
+}
+
+testEffect(dynamicResponse(() => Effect.die("Must reject before sending"))).effect(
+  "endpoint support does not grant trigger support, including for untyped callers",
+  () =>
+    Effect.gen(function* () {
+      const client = yield* LLMClient.Service
+      const unsupported = LLM.request({ model: Azure.configure({ resourceName: "fixture" }).responses("fixture") })
+      expect(LLMClient.canCompact(unsupported)).toBe(true)
+      expect(LLMClient.canCompact(unsupported, trigger)).toBe(false)
+      // @ts-expect-error Exercise untyped consumers; the service must reject before sending.
+      const serviceError = yield* client.compact(unsupported, trigger).pipe(Effect.flip)
+      expect(serviceError.reason._tag).toBe("UnsupportedOperation")
+      // @ts-expect-error Exercise an unknown mechanism supplied by JavaScript.
+      const error = yield* LLMClient.compact(request, { mechanism: "other" }).pipe(Effect.flip)
+      expect(error.reason._tag).toBe("InvalidRequest")
+    }),
+)
+
+testEffect(fixedResponse("must not use HTTP")).effect(
+  "invalid checkpoint output is rejected before acknowledging channel completion",
+  () =>
+    Effect.gen(function* () {
+      let completed = 0
+      const operation = LLMClient.compact(request, {
+        mechanism: "trigger",
+        webSocket: {
+          execute: () =>
+            Effect.succeed({
+              frames: Stream.make(
+                JSON.stringify({
+                  type: "response.completed",
+                  response: { id: "resp_1", output: [] },
+                }),
+              ),
+              complete: Effect.sync(() => {
+                completed++
+              }),
+            }),
+        },
+      })
+      const error = yield* operation.pipe(Effect.flip)
+      expect(error.reason._tag).toBe("InvalidProviderOutput")
+      expect(completed).toBe(0)
+    }),
+)
@@ -2,7 +2,8 @@ import { describe, expect, test } from "bun:test"
 import { ConfigProvider, Effect, Schema } from "effect"
 import { HttpClientRequest } from "effect/unstable/http"
 import { LLM, LLMEvent } from "../../src/index.js"
-import { CloudflareAIGateway, CloudflareWorkersAI } from "../../src/providers/cloudflare.js"
+import { CloudflareAIGateway } from "../../src/providers/cloudflare-ai-gateway.js"
+import { CloudflareWorkersAI } from "../../src/providers/cloudflare-workers-ai.js"
 import { compileRequest } from "../../src/route/client.js"
 import { it } from "../lib/effect.js"
 import { dynamicResponse } from "../lib/http.js"
@@ -1,6 +1,6 @@
 import { expect } from "bun:test"
 import { Effect, Schema } from "effect"
-import { LLM, LLMRequest, Message } from "../../src/index.js"
+import { LLM, LLMRequest, Message, ToolDefinition } from "../../src/index.js"
 import { LLMClient, Route } from "../../src/route/client.js"
 import { Auth } from "../../src/route/auth.js"
 import { Endpoint } from "../../src/route/endpoint.js"
@@ -134,7 +134,12 @@ for (const model of [
        [
          LLMRequest.update(request, {
            tools: [
-              { name: "unsupported", description: "Generation only", inputSchema: {}, native: { unsupported: {} } },
+              ToolDefinition.make({
+                name: "unsupported",
+                description: "Generation only",
+                inputSchema: {},
+                native: { unsupported: {} },
+              }),
            ],
          }),
          "InvalidRequest",
@@ -1643,19 +1643,14 @@ describe("Gemini route", () => {
    }),
  )

-  it.effect("maps current blocking and invalid-output finish reasons", () =>
+  it.effect("preserves blocking finishes and rejects invalid-output finish reasons", () =>
    Effect.gen(function* () {
      const reasons = [
        ["MODEL_ARMOR", "content-filter"],
        ["IMAGE_PROHIBITED_CONTENT", "content-filter"],
        ["IMAGE_RECITATION", "content-filter"],
        ["LANGUAGE", "content-filter"],
-        ["UNEXPECTED_TOOL_CALL", "error"],
-        ["NO_IMAGE", "error"],
        ["IMAGE_OTHER", "unknown"],
-        ["TOO_MANY_TOOL_CALLS", "error"],
-        ["MISSING_THOUGHT_SIGNATURE", "error"],
-        ["MALFORMED_RESPONSE", "error"],
      ] as const

      for (const [raw, normalized] of reasons) {
@@ -1666,6 +1661,26 @@ describe("Gemini route", () => {
        )
        expect(response.finishReason).toEqual({ normalized, raw })
      }
+
+      for (const raw of [
+        "MALFORMED_FUNCTION_CALL",
+        "UNEXPECTED_TOOL_CALL",
+        "NO_IMAGE",
+        "TOO_MANY_TOOL_CALLS",
+        "MISSING_THOUGHT_SIGNATURE",
+        "MALFORMED_RESPONSE",
+      ]) {
+        const event = { candidates: [{ finishReason: raw, finishMessage: "Provider detail" }], responseId: "failure" }
+        const error = yield* LLMClient.generate(request).pipe(
+          Effect.provide(fixedResponse(sseEvents(event))),
+          Effect.flip,
+        )
+        expect(error).toMatchObject({
+          _tag: "AI.Error",
+          reason: { _tag: "InvalidProviderOutput", body: JSON.stringify(event), http: { status: 200 } },
+          message: `Gemini stopped with ${raw}`,
+        })
+      }
    }),
  )

@@ -1,7 +1,13 @@
 import * as Anthropic from "../../src/providers/anthropic.js"
 import * as AnthropicCompatible from "../../src/providers/anthropic-compatible.js"
-import { Cerebras, DeepInfra, TogetherAI } from "../../src/providers/index.js"
-import { CloudflareAIGateway, CloudflareWorkersAI } from "../../src/providers/cloudflare.js"
+import {
+  Cerebras,
+  CloudflareAIGateway,
+  CloudflareWorkersAI,
+  DeepInfra,
+  DeepSeek,
+  TogetherAI,
+} from "../../src/providers/index.js"
 import * as Google from "../../src/providers/google.js"
 import * as OpenAI from "../../src/providers/openai.js"
 import * as OpenAICompatible from "../../src/providers/openai-compatible.js"
@@ -45,16 +51,18 @@ const cloudflareAIGatewayWorkers = cloudflareAIGateway.model("workers-ai/@cf/met
 const cloudflareAIGatewayWorkersTools = cloudflareAIGateway.model("workers-ai/@cf/openai/gpt-oss-20b")
 const cloudflareWorkersAI = cloudflareWorkers.model("@cf/meta/llama-3.1-8b-instruct")
 const cloudflareWorkersAITools = cloudflareWorkers.model("@cf/openai/gpt-oss-20b")
-const deepseek = OpenAICompatible.deepseek
-  .configure({ apiKey: process.env.DEEPSEEK_API_KEY ?? "fixture" })
-  .model("deepseek-chat")
+const deepseek = DeepSeek.configure({ apiKey: process.env.DEEPSEEK_API_KEY ?? "fixture" }).model("deepseek-chat")
 const together = TogetherAI.configure({
  apiKey: process.env.TOGETHER_API_KEY ?? process.env.TOGETHER_AI_API_KEY ?? "fixture",
 }).model("meta-llama/Llama-3.3-70B-Instruct-Turbo")
 const cerebras = Cerebras.configure({ apiKey: process.env.CEREBRAS_API_KEY ?? "fixture" }).model("gpt-oss-120b")
-const groq = OpenAICompatible.groq
-  .configure({ apiKey: process.env.GROQ_API_KEY ?? "fixture" })
-  .model("llama-3.3-70b-versatile")
+// These older cassettes exercise generic Chat compatibility. Native Groq request
+// shaping and reasoning are covered by groq.recorded.test.ts.
+const groq = OpenAICompatible.configure({
+  provider: "groq",
+  baseURL: "https://api.groq.com/openai/v1",
+  apiKey: process.env.GROQ_API_KEY ?? "fixture",
+}).model("llama-3.3-70b-versatile")
 const deepInfra = DeepInfra.configure({ apiKey: process.env.DEEPINFRA_API_KEY ?? "fixture" }).model(
  "meta-llama/Llama-3.3-70B-Instruct-Turbo",
 )
@@ -600,7 +600,6 @@ describe("Mistral Chat", () => {
        ["stop", "stop"],
        ["model_length", "length"],
        ["tool_calls", "tool-calls"],
-        ["error", "error"],
        ["future_reason", "unknown"],
      ] as const) {
        const response = yield* LLMClient.generate(request).pipe(
@@ -609,6 +608,22 @@ describe("Mistral Chat", () => {
        expect(response.finishReason).toEqual({ normalized, raw })
      }

+      for (const [raw, tag] of [
+        ["error", "UnknownProvider"],
+        ["network_error", "ProviderInternal"],
+      ] as const) {
+        const event = { ...chunk({}, raw), diagnostics: { trace: "failure" } }
+        const error = yield* LLMClient.generate(request).pipe(
+          Effect.provide(fixedResponse(sseEvents(event))),
+          Effect.flip,
+        )
+        expect(error).toMatchObject({
+          _tag: "AI.Error",
+          reason: { _tag: tag, body: JSON.stringify(event), http: { status: 200 } },
+          message: `Mistral Chat stopped with ${raw}`,
+        })
+      }
+
      const truncated = yield* LLMClient.generate(request).pipe(
        Effect.provide(
          fixedResponse(
@@ -8,10 +8,13 @@ import {
  Anthropic,
  AnthropicCompatible,
  Azure,
+  Baseten,
  Cerebras,
  CloudflareAIGateway,
  CloudflareWorkersAI,
  DeepInfra,
+  DeepSeek,
+  Fireworks,
  Google,
  GoogleVertex,
  GoogleVertexChat,
@@ -57,6 +60,9 @@ describe("native OpenAI-compatible providers", () => {
        "custom",
      ],
      [Cerebras.configure({ apiKey: "test" }).model("model"), "cerebras"],
+      [Baseten.configure({ apiKey: "test" }).model("model"), "baseten"],
+      [DeepSeek.configure({ apiKey: "test" }).model("model"), "deepseek"],
+      [Fireworks.configure({ apiKey: "test" }).model("model"), "fireworks"],
      [DeepInfra.configure({ apiKey: "test" }).model("model"), "deepinfra"],
      [TogetherAI.configure({ apiKey: "test" }).model("model"), "togetherai"],
      [CloudflareAIGateway.configure({ accountId: "account" }).model("model"), "cloudflare-ai-gateway"],
@@ -87,6 +93,36 @@ describe("native OpenAI-compatible providers", () => {
    expect(cerebras.route.endpoint.baseURL).toBe("https://api.cerebras.ai/v1")
  })

+  it.effect("preserves extracted providers' Chat requests and identity through custom endpoints", () =>
+    Effect.gen(function* () {
+      for (const provider of [Baseten, DeepSeek, Fireworks]) {
+        const settings = {
+          apiKey: "fixture",
+          baseURL: "https://gateway.example/v1",
+          providerOptions: { reasoningEffort: "high" },
+        }
+        const selected = provider.configure(settings).model("test-model")
+        expect(selected.provider).toBe(provider.id)
+        expect(selected.route.id).toBe(`${provider.id}-chat`)
+        expect(selected.route.protocol).toBe("openai-chat")
+        expect(selected.route.endpoint.baseURL).toBe(settings.baseURL)
+        const input = {
+          prompt: "Use a tool.",
+          generation: { maxTokens: 48 },
+          tools: [ToolDefinition.make({ name: "lookup", description: "Lookup data", inputSchema: { type: "object" } })],
+        }
+        const native = yield* compileRequest(LLM.request({ ...input, model: selected }))
+        const generic = yield* compileRequest(
+          LLM.request({
+            ...input,
+            model: OpenAICompatible.configure({ ...settings, provider: provider.id }).model("test-model"),
+          }),
+        )
+        expect(native.body).toEqual(generic.body)
+      }
+    }),
+  )
+
  test("preserves native DeepInfra provider and route identity", () => {
    const deepinfra = DeepInfra.configure({ apiKey: "fixture" }).model("google/gemma-3-27b-it")
    expect(deepinfra).toMatchObject({
@@ -164,7 +200,7 @@ describe("native OpenAI-compatible providers", () => {
  })

  test("maps package settings onto native executable models", () => {
-    for (const native of [TogetherAI, Cerebras]) {
+    for (const native of [Baseten, Cerebras, DeepSeek, Fireworks, TogetherAI]) {
      const selected = native.model("provider-model", {
        apiKey: "fixture",
        baseURL: "https://gateway.example/v1",
@@ -183,6 +219,24 @@ describe("native OpenAI-compatible providers", () => {
  it.effect("resolves provider environment credentials and preserves deprecated Together credentials", () =>
    Effect.gen(function* () {
      const scenarios = [
+        {
+          model: Baseten.configure().model("model"),
+          env: { BASETEN_API_KEY: "baseten-secret" },
+          token: "baseten-secret",
+          url: "https://inference.baseten.co/v1/chat/completions",
+        },
+        {
+          model: DeepSeek.configure().model("deepseek-chat"),
+          env: { DEEPSEEK_API_KEY: "deepseek-secret" },
+          token: "deepseek-secret",
+          url: "https://api.deepseek.com/v1/chat/completions",
+        },
+        {
+          model: Fireworks.configure().model("model"),
+          env: { FIREWORKS_API_KEY: "fireworks-secret" },
+          token: "fireworks-secret",
+          url: "https://api.fireworks.ai/inference/v1/chat/completions",
+        },
        {
          model: TogetherAI.configure().model("llama"),
          env: { TOGETHER_API_KEY: "together-primary", TOGETHER_AI_API_KEY: "together-legacy" },
@@ -71,7 +71,7 @@ function expectLifecycle(events: ReadonlyArray<LLMEvent>, completed: boolean) {
 }

 describe("Open Responses basic-item lifecycles", () => {
-  it.effect("closes implicit summary boundaries and ignores late events for completed reasoning", () =>
+  it.effect("closes implicit summary boundaries", () =>
    Effect.gen(function* () {
      const item = { type: "reasoning", id: "rs_1", encrypted_content: "encrypted-state" }
      const events = yield* collect(
@@ -90,12 +90,6 @@ describe("Open Responses basic-item lifecycles", () => {
          delta: "Third",
        },
        { type: "response.output_item.done", item },
-        { type: "response.output_item.done", item },
-        { type: "response.output_item.added", item },
-        { type: "response.reasoning_summary_part.added", item_id: "rs_1", summary_index: 3 },
-        { type: "response.reasoning_summary_text.delta", item_id: "rs_1", summary_index: 3, delta: "late" },
-        { type: "response.reasoning_summary_text.done", item_id: "rs_1", summary_index: 2, text: "late final" },
-        { type: "response.reasoning_summary_part.done", item_id: "rs_1", summary_index: 3 },
        completed,
      )

@@ -129,7 +123,7 @@ describe("Open Responses basic-item lifecycles", () => {
    }),
  )

-  it.effect("preserves done-only reasoning text and encryption without replaying late events", () =>
+  it.effect("preserves done-only reasoning text and encryption", () =>
    Effect.gen(function* () {
      const item = {
        type: "reasoning",
@@ -139,11 +133,6 @@ describe("Open Responses basic-item lifecycles", () => {
      }
      const events = yield* collect(
        { type: "response.output_item.done", item },
-        { type: "response.output_item.done", item },
-        { type: "response.output_item.added", item },
-        { type: "response.reasoning_summary_text.delta", item_id: "rs_1", delta: "late" },
-        { type: "response.reasoning_summary_part.added", item_id: "rs_1", summary_index: 1 },
-        { type: "response.reasoning_summary_text.done", item_id: "rs_1", summary_index: 1, text: "late final" },
        completed,
        // Route termination must also prevent events after response completion.
        { type: "response.output_item.added", item: { type: "reasoning", id: "rs_after" } },
@@ -337,7 +326,6 @@ describe("Open Responses basic-item lifecycles", () => {
      ])
    }),
  )
-
  it.effect("mints an id for a done-only tool that never had one", () =>
    Effect.gen(function* () {
      const events = yield* collect(
@@ -368,9 +356,16 @@ describe("Open Responses basic-item lifecycles", () => {
      const events = yield* collect({ type: "response.output_item.done", item }, completed)
      const providerMetadata = { "openai-compatible": { itemId: "fc_1" } }
      expect(events.filter((event) => event.type.startsWith("tool-"))).toEqual([
-        { type: "tool-input-start", id: "call_1", name: "lookup", providerMetadata },
-        { type: "tool-input-end", id: "call_1", name: "lookup", providerMetadata },
-        { type: "tool-call", id: "call_1", name: "lookup", input: { query: "weather" }, providerMetadata },
+        { type: "tool-input-start", id: "call_1", name: "lookup", namespace: undefined, providerMetadata },
+        { type: "tool-input-end", id: "call_1", name: "lookup", namespace: undefined, providerMetadata },
+        {
+          type: "tool-call",
+          id: "call_1",
+          name: "lookup",
+          namespace: undefined,
+          input: { query: "weather" },
+          providerMetadata,
+        },
      ])
      expect(events.filter(LLMEvent.is.finish)).toEqual([
        {
@@ -4,7 +4,6 @@ import { HttpClientRequest } from "effect/unstable/http"
 import { LLM, LLMRequest, Message, ToolCallPart, ToolChoice, ToolDefinition } from "../../src/index.js"
 import { Auth, LLMClient } from "../../src/route.js"
 import { compileRequest } from "../../src/route/client.js"
-import * as OpenAICompatible from "../../src/providers/openai-compatible.js"
 import * as OpenAICompatibleChat from "../../src/protocols/openai-compatible-chat.js"
 import { it } from "../lib/effect.js"
 import { dynamicResponse, fixedResponse } from "../lib/http.js"
@@ -41,15 +40,6 @@ const usageChunk = (usage: object) => ({
  usage,
 })

-const providerFamilies = [
-  ["baseten", OpenAICompatible.baseten, "https://inference.baseten.co/v1"],
-  ["cerebras", OpenAICompatible.cerebras, "https://api.cerebras.ai/v1"],
-  ["deepinfra", OpenAICompatible.deepinfra, "https://api.deepinfra.com/v1/openai"],
-  ["deepseek", OpenAICompatible.deepseek, "https://api.deepseek.com/v1"],
-  ["fireworks", OpenAICompatible.fireworks, "https://api.fireworks.ai/inference/v1"],
-  ["togetherai", OpenAICompatible.togetherai, "https://api.together.xyz/v1"],
-] as const
-
 describe("OpenAI-compatible Chat route", () => {
  it.effect("prepares generic Chat target", () =>
    Effect.gen(function* () {
@@ -91,39 +81,6 @@ describe("OpenAI-compatible Chat route", () => {
    }),
  )

-  test("provides model helpers for compatible provider families", () => {
-    expect(
-      providerFamilies.map(([provider, family]) => {
-        const model = family.configure({ apiKey: "test-key" }).model(`${provider}-model`)
-        return {
-          id: String(model.id),
-          provider: String(model.provider),
-          route: model.route.id,
-          baseURL: model.route.endpoint.baseURL,
-        }
-      }),
-    ).toEqual(
-      providerFamilies.map(([provider, _, baseURL]) => ({
-        id: `${provider}-model`,
-        provider,
-        route: "openai-compatible-chat",
-        baseURL,
-      })),
-    )
-
-    const custom = OpenAICompatible.deepseek
-      .configure({
-        apiKey: "test-key",
-        baseURL: "https://custom.deepseek.test/v1",
-      })
-      .model("deepseek-chat")
-    expect(custom).toMatchObject({
-      provider: "deepseek",
-      route: { id: "openai-compatible-chat" },
-    })
-    expect(custom.route.endpoint.baseURL).toBe("https://custom.deepseek.test/v1")
-  })
-
  it.effect("matches AI SDK compatible basic request body fixture", () =>
    Effect.gen(function* () {
      const prepared = yield* compileRequest(request)
@@ -171,6 +171,66 @@ describe("Open Responses-compatible route", () => {
    }),
  )

+  it.effect("flattens tool namespaces", () =>
+    Effect.gen(function* () {
+      const model = configure({ apiKey: "test-key", baseURL: "https://responses.example.test/v1" }).model(
+        "example-model",
+      )
+      const prepared = yield* compileRequest(
+        LLM.request({
+          model,
+          tools: [
+            {
+              type: "namespace",
+              name: "acme",
+              tools: [
+                {
+                  type: "namespace",
+                  name: "billing",
+                  tools: [ToolDefinition.make({ name: "lookup", description: "Lookup billing", inputSchema: {} })],
+                },
+                ToolDefinition.make({ name: "users", description: "Lookup users", inputSchema: {} }),
+              ],
+            },
+          ],
+        }),
+      )
+
+      expect(prepared.body.tools).toEqual([
+        {
+          type: "function",
+          name: "acme_billing_lookup",
+          description: "Lookup billing",
+          parameters: {},
+          strict: false,
+        },
+        { type: "function", name: "acme_users", description: "Lookup users", parameters: {}, strict: false },
+      ])
+    }),
+  )
+
+  it.effect("flattens tool namespaces in history", () =>
+    Effect.gen(function* () {
+      const model = configure({ apiKey: "test-key", baseURL: "https://responses.example.test/v1" }).model(
+        "example-model",
+      )
+      const prepared = yield* compileRequest(
+        LLM.request({
+          model,
+          messages: [
+            Message.assistant({ type: "tool-call", id: "call_1", name: "lookup", namespace: "crm", input: {} }),
+            Message.tool({ id: "call_1", name: "lookup", namespace: "crm", result: "done", resultType: "text" }),
+          ],
+        }),
+      )
+
+      expect(prepared.body.input).toEqual([
+        { type: "function_call", call_id: "call_1", name: "crm_lookup", namespace: undefined, arguments: "{}" },
+        { type: "function_call_output", call_id: "call_1", output: "done" },
+      ])
+    }),
+  )
+
  it.effect("lowers canonical parallel tool control", () =>
    Effect.gen(function* () {
      const model = configure({
@@ -12,6 +12,7 @@ import {
  LanguageModel,
  ToolCallPart,
  ToolDefinition,
+  ToolNamespace,
  ToolResultPart,
  TransportError,
  Usage,
@@ -143,6 +144,94 @@ describe("OpenAI Responses route", () => {
    }),
  )

+  it.effect("lowers tool namespaces without flattening leaf names", () =>
+    Effect.gen(function* () {
+      const prepared = yield* compileRequest(
+        LLM.request({
+          model,
+          prompt: "Find a customer and their orders.",
+          tools: [
+            ToolNamespace.make({
+              name: "crm",
+              description: "Customer management",
+              tools: [
+                ToolDefinition.make({ name: "lookup", description: "Look up a customer", inputSchema: {} }),
+                ToolDefinition.make({ name: "orders", description: "List customer orders", inputSchema: {} }),
+              ],
+            }),
+          ],
+        }),
+      )
+
+      expect(prepared.body.tools).toEqual([
+        {
+          type: "namespace",
+          name: "crm",
+          description: "Customer management",
+          tools: [
+            { type: "function", name: "lookup", description: "Look up a customer", parameters: {}, strict: false },
+            { type: "function", name: "orders", description: "List customer orders", parameters: {}, strict: false },
+          ],
+        },
+      ])
+    }),
+  )
+
+  it.effect("flattens nested levels within a native tool namespace", () =>
+    Effect.gen(function* () {
+      const prepared = yield* compileRequest(
+        LLM.request({
+          model,
+          tools: [
+            {
+              type: "namespace",
+              name: "crm",
+              description: "Customer management",
+              tools: [
+                {
+                  type: "namespace",
+                  name: "orders",
+                  description: "Order management",
+                  tools: [ToolDefinition.make({ name: "list", description: "List orders", inputSchema: {} })],
+                },
+              ],
+            },
+          ],
+        }),
+      )
+
+      expect(prepared.body.tools).toEqual([
+        {
+          type: "namespace",
+          name: "crm",
+          description: "Customer management",
+          tools: [{ type: "function", name: "orders_list", description: "List orders", parameters: {}, strict: false }],
+        },
+      ])
+    }),
+  )
+
+  it.effect("defaults tool namespace descriptions", () =>
+    Effect.gen(function* () {
+      const prepared = yield* compileRequest(
+        LLM.request({
+          model,
+          tools: [
+            {
+              type: "namespace",
+              name: "crm",
+              tools: [ToolDefinition.make({ name: "lookup", description: "Look up a customer", inputSchema: {} })],
+            },
+          ],
+        }),
+      )
+
+      expect(prepared.body.tools).toEqual([
+        expect.objectContaining({ type: "namespace", name: "crm", description: "Tools in the crm namespace." }),
+      ])
+    }),
+  )
+
  it.effect("rejects invalid hosted image generation options locally", () =>
    Effect.gen(function* () {
      const error = yield* compileRequest(
@@ -437,7 +526,7 @@ describe("OpenAI Responses route", () => {
    }),
  )

-  it.effect("tolerates keepalive frames before response.created", () =>
+  it.effect("tolerates keepalive and provider notifications before response.created", () =>
    Effect.gen(function* () {
      const webSocket = WebSocketTransport.makeDirect({
        open: () =>
@@ -445,6 +534,7 @@ describe("OpenAI Responses route", () => {
            sendText: () => Effect.void,
            messages: Stream.fromArray([
              ProviderShared.encodeJson({ type: "keepalive", sequence_number: 0 }),
+              ProviderShared.encodeJson({ type: "codex.rate_limits" }),
              ProviderShared.encodeJson({ type: "response.created", response: { id: "resp_alive" } }),
              ProviderShared.encodeJson({
                type: "response.completed",
@@ -568,7 +658,7 @@ describe("OpenAI Responses route", () => {
    }),
  )

-  it.effect("continues a promoted steer after the completed assistant output", () =>
+  it.effect("continues a promoted steer after assistant output with response-only text metadata", () =>
    Effect.gen(function* () {
      const firstInput = [{ role: "user", content: [{ type: "input_text", text: "First" }] }]
      const first = continuationDriver({ type: "response.create", model: "gpt-5.2", store: false, input: firstInput })
@@ -582,7 +672,7 @@ describe("OpenAI Responses route", () => {
            id: "msg_1",
            status: "completed",
            role: "assistant",
-            content: [{ type: "output_text", text: "Hello" }],
+            content: [{ type: "output_text", text: "Hello", annotations: [], logprobs: [] }],
          },
        }),
      )
@@ -610,42 +700,37 @@ describe("OpenAI Responses route", () => {
    }),
  )

-  it.effect("continues store-false reasoning while retaining the output item ID", () =>
+  it.effect("continues streamed reasoning when completion re-encrypts the same item", () =>
    Effect.gen(function* () {
      const firstInput = [{ role: "user", content: [{ type: "input_text", text: "Think" }] }]
      const request = { type: "response.create", model: "gpt-5.2", store: false, input: firstInput }
+      const reasoning = {
+        type: "reasoning",
+        id: "rs_1",
+        summary: [{ type: "summary_text", text: "Thought" }],
+        encrypted_content: "encrypted",
+      }
      const first = continuationDriver(request)
      const create = yield* first.create(undefined)
      yield* first.observe(
        create,
        ProviderShared.encodeJson({
          type: "response.output_item.done",
-          item: {
-            type: "reasoning",
-            id: "rs_1",
-            summary: [{ type: "summary_text", text: "Thought" }],
-            encrypted_content: "encrypted",
-          },
+          item: reasoning,
        }),
      )
      const saved = checkpoint(
        yield* first.observe(
          create,
-          ProviderShared.encodeJson({ type: "response.completed", response: { id: "resp_1" } }),
+          ProviderShared.encodeJson({
+            type: "response.completed",
+            response: { id: "resp_1", output: [{ ...reasoning, encrypted_content: "terminal-encrypted" }] },
+          }),
        ),
      )
      const next = continuationDriver({
        ...request,
-        input: [
-          ...firstInput,
-          {
-            type: "reasoning",
-            id: "rs_1",
-            summary: [{ type: "summary_text", text: "Thought" }],
-            encrypted_content: "encrypted",
-          },
-          { role: "user", content: [{ type: "input_text", text: "Continue" }] },
-        ],
+        input: [...firstInput, reasoning, { role: "user", content: [{ type: "input_text", text: "Continue" }] }],
      })

      const continued = yield* next.create(saved)
@@ -655,6 +740,11 @@ describe("OpenAI Responses route", () => {
        previous_response_id: "resp_1",
        input: [{ role: "user", content: [{ type: "input_text", text: "Continue" }] }],
      })
+      const edited = yield* continuationDriver({
+        ...request,
+        input: [...firstInput, { ...reasoning, encrypted_content: "edited" }, { role: "user", content: "Continue" }],
+      }).create(saved)
+      expect(edited.mode).toBe("full")
    }),
  )

@@ -2130,6 +2220,71 @@ describe("OpenAI Responses route", () => {
    }),
  )

+  it.effect("preserves tool namespaces through streaming and history replay", () =>
+    Effect.gen(function* () {
+      const item = {
+        type: "function_call",
+        id: "fc_1",
+        call_id: "call_1",
+        namespace: "crm",
+        name: "lookup",
+        arguments: "",
+      }
+      const response = yield* LLMClient.generate(request).pipe(
+        Effect.provide(
+          fixedResponse(
+            sseEvents(
+              { type: "response.output_item.added", output_index: 0, item },
+              {
+                type: "response.function_call_arguments.delta",
+                output_index: 0,
+                item_id: "fc_1",
+                delta: '{"id":"123"}',
+              },
+              {
+                type: "response.output_item.done",
+                output_index: 0,
+                item: { ...item, arguments: '{"id":"123"}' },
+              },
+              { type: "response.completed", response: { id: "resp_1" } },
+            ),
+          ),
+        ),
+      )
+
+      const toolEvents = response.events.filter((event) => event.type.startsWith("tool-"))
+      expect(toolEvents).toEqual([
+        expect.objectContaining({ type: "tool-input-start", name: "lookup", namespace: "crm" }),
+        expect.objectContaining({ type: "tool-input-delta", name: "lookup", namespace: "crm" }),
+        expect.objectContaining({ type: "tool-input-end", name: "lookup", namespace: "crm" }),
+        expect.objectContaining({ type: "tool-call", name: "lookup", namespace: "crm", input: { id: "123" } }),
+      ])
+      expect(response.message.content).toEqual([
+        expect.objectContaining({ type: "tool-call", name: "lookup", namespace: "crm", input: { id: "123" } }),
+      ])
+
+      const prepared = yield* compileRequest(
+        LLM.request({
+          model,
+          messages: [
+            response.message,
+            Message.tool({ id: "call_1", name: "lookup", namespace: "crm", result: { customer: "Ada" } }),
+          ],
+        }),
+      )
+      expect(prepared.body.input).toEqual([
+        {
+          type: "function_call",
+          id: "fc_1",
+          call_id: "call_1",
+          namespace: "crm",
+          name: "lookup",
+          arguments: '{"id":"123"}',
+        },
+        { type: "function_call_output", call_id: "call_1", output: '{"customer":"Ada"}' },
+      ])
+    }),
+  )
  it.effect("routes reasoning summary events by output index", () =>
    Effect.gen(function* () {
      const response = yield* LLMClient.generate(request).pipe(
@@ -2861,7 +3016,6 @@ describe("OpenAI Responses route", () => {
              { type: "response.output_item.added", item: { type: "reasoning", id: "rs_1" } },
              { type: "response.reasoning_summary_text.delta", item_id: "rs_1", summary_index: 0, delta: "Think" },
              { type: "response.output_item.done", item: { type: "reasoning", id: "rs_1" } },
-              { type: "response.output_item.done", item: { type: "reasoning", id: "rs_1" } },
              {
                type: "response.output_item.added",
                item: { type: "function_call", id: "fc_1", call_id: "call_1", name: "lookup", arguments: "" },
@@ -3,6 +3,7 @@ import {
  AIError,
  CompactionPart,
  CompactionResponse,
+  CompactionCheckpointResponse,
  LanguageModel,
  LLM,
  LLMClient,
@@ -79,6 +80,31 @@ describe("TestLLM legacy client", () => {
 })

 describe("TestLLM first-class client", () => {
+  it.effect("returns scripted checkpoints only for trigger compaction", () =>
+    Effect.gen(function* () {
+      const client = yield* TestLLM.Test
+      const request = LLM.request({ model: OpenAI.configure().responses("fixture"), prompt: "hello" })
+      const checkpoint = new CompactionCheckpointResponse({
+        checkpoint: { type: "compaction", provider: ProviderID.make("openai"), encrypted: "opaque" },
+        responseID: "resp_fixture",
+      })
+      const endpoint = new CompactionResponse({ replacement: [] })
+      yield* client.push(checkpoint, endpoint, checkpoint, checkpoint)
+      const operation = LLMClient.compact(request, { mechanism: "trigger" })
+      expect(yield* client.requests()).toEqual([])
+      expect(yield* operation).toBe(checkpoint)
+      expect(yield* client.compact(request, { mechanism: "trigger" }).pipe(Effect.catchDefect(Effect.succeed))).toBe(
+        "TestLLM trigger compaction requires a CompactionCheckpointResponse",
+      )
+      expect(yield* client.compact(request).pipe(Effect.catchDefect(Effect.succeed))).toBe(
+        "TestLLM compaction requires a CompactionResponse",
+      )
+      expect(yield* client.generate(request).pipe(Effect.catchDefect(Effect.succeed))).toBe(
+        "TestLLM generation requires an event response",
+      )
+    }),
+  )
+
  it.effect("rejects response fixtures for the wrong operation", () =>
    Effect.gen(function* () {
      const client = yield* TestLLM.Test
@@ -15,13 +15,7 @@ describe("tool history normalization", () => {
      Message.assistant(toolCall("trailing")),
    ])

-    expect(normalized.map((message) => message.role)).toEqual([
-      "assistant",
-      "tool",
-      "tool",
-      "user",
-      "assistant",
-    ])
+    expect(normalized.map((message) => message.role)).toEqual(["assistant", "tool", "tool", "user", "assistant"])
    expect(normalized[1]?.content[0]).toMatchObject({ type: "tool-result", id: "first", name: "first" })
    expect(normalized[2]?.content).toEqual([
      { type: "tool-result", id: "second", name: "second", result: { type: "error", value: "Tool result missing" } },
@@ -74,4 +68,13 @@ describe("tool history normalization", () => {

    expect(normalizeToolHistory([orphan, hosted])).toEqual([orphan, hosted])
  })
+
+  test("uses a matching call as the complete tool identity", () => {
+    const normalized = normalizeToolHistory([
+      Message.assistant(ToolCallPart.make({ id: "call_1", name: "lookup", input: {} })),
+      Message.tool(ToolResultPart.make({ id: "call_1", name: "wrong", namespace: "stale", result: "done" })),
+    ])
+
+    expect(normalized[1]?.content[0]).toMatchObject({ name: "lookup", namespace: undefined })
+  })
 })
@@ -7,6 +7,7 @@ import {
  LLMEvent,
  LLMRequest,
  LLMResponse,
+  ToolCallPart,
  ToolChoice,
  ToolOutput,
  toDefinitions,
@@ -36,6 +37,27 @@ const baseRequest = LLM.request({
 })
 const weatherFailureCause = new Error("weather lookup denied")

+test("dispatches namespaced calls by qualified identity", async () => {
+  let context: ToolExecuteContext | undefined
+  const lookup = Tool.make({
+    description: "Look up a customer.",
+    parameters: Schema.Struct({}),
+    success: Schema.String,
+    execute: (_, value) => {
+      context = value
+      return Effect.succeed("customer")
+    },
+  })
+  const call = ToolCallPart.make({ id: "call_1", namespace: "crm", name: "lookup", input: {} })
+  const result = await Effect.runPromise(
+    ToolRuntime.dispatch({ "crm.lookup": lookup, lookup: schema_only_weather }, call),
+  )
+
+  expect(result.result).toEqual({ type: "text", value: "customer" })
+  expect(context).toEqual({ id: "call_1", namespace: "crm", name: "lookup" })
+  expect(result.events).toEqual([expect.objectContaining({ type: "tool-result", namespace: "crm", name: "lookup" })])
+})
+
 const get_weather = Tool.make({
  description: "Get current weather for a city.",
  parameters: Schema.Struct({ city: Schema.String }),
@@ -5,5 +5,5 @@
    "noEmit": true,
    "rootDir": "."
  },
-  "include": ["test/**/*.types.ts", "test/testing.test.ts"]
+  "include": ["test/**/*.types.ts", "test/testing.test.ts", "test/provider/checkpoint.test.ts"]
 }
@@ -1,16 +1,139 @@
 import { expect, story } from "../../storybook/playwright/story"

-story("raises the docked composer only in dark mode", async ({ mount, page }) => {
-  const component = await mount("opencode-composer-flow--empty-draft")
-  const composer = component.locator('[data-component="composer"]')
+for (const direction of ["ltr", "rtl"]) {
+  for (const alternate of ["none", "queue", "steer"]) {
+    story(`scrolls overflowing controls beside fixed ${alternate} actions in ${direction}`, async ({ mount, page }) => {
+      const component = await mount("opencode-composer-flow--toolbar-overflow", {
+        args: { alternate },
+        globals: { direction },
+      })
+      const controls = component.locator('[data-slot="composer-controls"]')
+      const actions = component.locator('[data-slot="composer-actions"]')
+      const submit = component.locator('[data-action="composer-submit"]')
+      const add = component.locator('[data-action="composer-attach"]')
+      const agent = controls.getByRole("button", { name: "Choose agent" })
+      const variant = controls.getByRole("button", { name: "Choose model variant" })
+      await expect(controls).toHaveCSS("overflow-x", "auto")
+      await expect(controls).toHaveCSS("overscroll-behavior-x", "contain")
+      await expect(controls).toHaveCSS("direction", direction)
+      await expect(controls).toHaveCSS("padding-inline-start", "0px")
+      await expect(controls).toHaveCSS("padding-inline-end", "0px")
+      await expect(component.locator('[data-action="composer-alternate-delivery"]')).toHaveCount(
+        alternate === "none" ? 0 : 1,
+      )

-  await page.locator("html").evaluate((root) => root.setAttribute("data-color-scheme", "light"))
-  await expect(composer).toHaveCSS("background-color", "rgb(255, 255, 255)")
+      for (const width of [1024, 360]) {
+        await page.setViewportSize({ width, height: 720 })
+        await expect
+          .poll(() => controls.evaluate((element) => element.scrollWidth - element.clientWidth))
+          .toBeGreaterThan(0)
+        const fixed = await submit.boundingBox()
+        const fixedAdd = await add.boundingBox()
+        const viewport = await controls.boundingBox()
+        const action = await actions.boundingBox()
+        expect(fixed).not.toBeNull()
+        expect(viewport).not.toBeNull()
+        expect(action).not.toBeNull()
+        expect(fixedAdd).not.toBeNull()
+        if (!fixed || !viewport || !action || !fixedAdd) return
+        expect(direction === "ltr" ? fixedAdd.x + fixedAdd.width : viewport.x + viewport.width).toBeCloseTo(
+          direction === "ltr" ? viewport.x - 4 : fixedAdd.x - 4,
+          1,
+        )
+        expect(direction === "ltr" ? viewport.x + viewport.width : action.x + action.width).toBeCloseTo(
+          direction === "ltr" ? action.x - 12 : viewport.x - 12,
+          1,
+        )

-  await page.locator("html").evaluate((root) => root.setAttribute("data-color-scheme", "dark"))
-  await expect(composer).toHaveCSS("background-color", "rgb(36, 36, 36)")
+        await controls.evaluate((element) => {
+          element.scrollLeft = 0
+        })
+        await expect(controls).toHaveAttribute("data-overflow-start", "false")
+        await expect(controls).toHaveAttribute("data-overflow-end", "true")
+        await expect(controls).toHaveCSS(
+          "mask-image",
+          `linear-gradient(to ${direction === "ltr" ? "right" : "left"}, rgba(0, 0, 0, 0), rgb(0, 0, 0) 0px, rgb(0, 0, 0) calc(100% - 16px), rgba(0, 0, 0, 0))`,
+        )
+        const first = await agent.boundingBox()
+        if (!first) throw new Error("Missing agent control")
+        expect(
+          direction === "ltr" ? first.x - viewport.x : viewport.x + viewport.width - first.x - first.width,
+        ).toBeCloseTo(0, 0)
+        await controls.evaluate((element) => {
+          element.scrollLeft =
+            ((getComputedStyle(element).direction === "rtl" ? -1 : 1) * (element.scrollWidth - element.clientWidth)) / 2
+        })
+        await expect(controls).toHaveAttribute("data-overflow-start", "true")
+        await expect(controls).toHaveAttribute("data-overflow-end", "true")
+        const scrolled = (await controls.boundingBox())!
+        expect(scrolled.x).toBeCloseTo(viewport.x, 1)
+        expect(scrolled.width).toBeCloseTo(viewport.width, 1)
+        await controls.hover()
+        await page.mouse.wheel(direction === "ltr" ? 1000 : -1000, 0)
+        await expect.poll(() => controls.evaluate((element) => Math.abs(element.scrollLeft))).toBeGreaterThan(0)
+        expect(await submit.boundingBox()).toEqual(fixed)
+        expect(await add.boundingBox()).toEqual(fixedAdd)
+
+        // The fade disappears at the endpoint instead of reserving padding.
+        await variant.focus()
+        await controls.evaluate((element) => {
+          element.scrollLeft =
+            getComputedStyle(element).direction === "rtl" ? -element.scrollWidth : element.scrollWidth
+        })
+        await expect(controls).toHaveAttribute("data-overflow-start", "true")
+        await expect(controls).toHaveAttribute("data-overflow-end", "false")
+        const last = await variant.boundingBox()
+        if (!last) throw new Error("Missing variant control")
+        expect(
+          Math.abs(direction === "ltr" ? viewport.x + viewport.width - last.x - last.width : last.x - viewport.x),
+        ).toBeLessThan(1)
+        expect(
+          Math.abs((direction === "ltr" ? action.x - last.x - last.width : last.x - action.x - action.width) - 12),
+        ).toBeLessThan(1)
+        await page.keyboard.press("Enter")
+        await expect(page.getByRole("menuitemradio", { name: "high", exact: true })).toBeVisible()
+        await page.keyboard.press("Escape")
+        await expect(variant).toBeFocused()
+        await expect(submit).toBeInViewport()
+        await expect(add).toBeInViewport()
+        await add.click()
+        await expect(page.getByRole("menuitem", { name: "Images and files" })).toBeVisible()
+        await page.keyboard.press("Escape")
+        expect(await add.boundingBox()).toEqual(fixedAdd)
+      }
+
+      if (alternate !== "none") {
+        const width = (await controls.boundingBox())!.width
+        await component.getByRole("textbox", { name: "Prompt", exact: true }).fill("")
+        await expect(component.locator('[data-action="composer-alternate-delivery"]')).toHaveCount(0)
+        await expect.poll(async () => (await controls.boundingBox())!.width).toBeGreaterThan(width)
+      }
+    })
+  }
+}
+
+story("does not mask or pad controls when they fit", async ({ mount }) => {
+  const component = await mount("opencode-composer-flow--model-and-variant")
+  const controls = component.locator('[data-slot="composer-controls"]')
+  await expect(controls).toHaveAttribute("data-overflow-start", "false")
+  await expect(controls).toHaveAttribute("data-overflow-end", "false")
+  await expect(controls).toHaveCSS("mask-image", "none")
+  await expect(controls).toHaveCSS("padding-inline-start", "0px")
+  await expect(controls).toHaveCSS("padding-inline-end", "0px")
 })

+// ThemeProvider writes resolved token values into a <style> block, so toggling data-color-scheme by hand
+// leaves every --v2-* variable at its previous value. Switch themes through the Storybook global instead.
+for (const [theme, background] of [
+  ["light", "rgb(255, 255, 255)"],
+  ["dark", "rgb(36, 36, 36)"],
+] as const) {
+  story(`raises the docked composer only in dark mode (${theme})`, async ({ mount }) => {
+    const component = await mount("opencode-composer-flow--empty-draft", { globals: { theme } })
+    await expect(component.locator('[data-component="composer"]')).toHaveCSS("background-color", background)
+  })
+}
+
 story("centers add menu shortcuts in a consistent column", async ({ mount, page }) => {
  const component = await mount("opencode-composer-flow--empty-draft")
  await component.locator('[data-action="composer-attach"]').click()
@@ -88,27 +211,88 @@ story("renders a draft once and supports editing, caret restoration, and failure
  await expect(input).toHaveText("Preserve this draft on failure")
 })

-// Moved from packages/app/e2e/regression/prompt-thinking-level.spec.ts
-story("shows the thinking level control while relevant", async ({ mount, page }) => {
+story("shows thinking on composer hover or when a non-default variant is selected", async ({ mount, page }) => {
  const component = await mount("opencode-composer-flow--model-and-variant")
  const composer = component.locator('[data-component="composer"]')
-  const input = composer.locator('[data-component="composer-editor"]')
+  const input = composer.getByRole("textbox", { name: "Prompt", exact: true })
  const control = composer.getByRole("button", { name: "Choose model variant" })

+  await component.getByRole("status").click()
  await page.mouse.move(0, 0)
-  await page.evaluate(() => (document.activeElement as HTMLElement | null)?.blur())
-  await expect(control).toBeVisible()
+  await expect(control).toHaveText("balanced")
+  await expect(control).toHaveCSS("opacity", "1")

+  await control.click()
+  await page.getByRole("menuitemradio", { name: "default", exact: true }).click()
+  await component.getByRole("status").click()
+  await expect(control).toHaveText("default")
+  await expect(control).toHaveCSS("opacity", "0")
+  await expect(control).toHaveCSS("pointer-events", "none")
+
+  await input.hover()
+  await expect(control).toHaveCSS("opacity", "1")
+  await expect(control).toHaveCSS("pointer-events", "auto")
+  await input.click()
+  await page.mouse.move(0, 0)
+  await expect(input).toBeFocused()
+  await expect(control).toHaveCSS("opacity", "0")
+
+  await input.hover()
  await control.click()
  const high = page.getByRole("menuitemradio", { name: "high" })
  await expect(high).toBeVisible()
  await page.mouse.move(0, 0)
-  await expect(control).toBeVisible()
+  await expect(control).toHaveAttribute("aria-expanded", "true")
+  await expect(control).toHaveCSS("opacity", "1")
  await expect(high).toBeVisible()
  await high.click()
+  await component.getByRole("status").click()
+  await expect(control).toHaveText("high")
+  await expect(control).toHaveCSS("opacity", "1")

-  await input.focus()
-  await expect(control).toBeVisible()
-  await input.blur()
-  await expect(control).toBeVisible()
+  await control.click()
+  await page.getByRole("menuitemradio", { name: "default", exact: true }).click()
+  await component.getByRole("status").click()
+  await expect(control).toHaveCSS("opacity", "0")
+})
+
+story("keeps default thinking accessible by keyboard without composer hover", async ({ mount, page }) => {
+  const component = await mount("opencode-composer-flow--model-and-variant")
+  const input = component.getByRole("textbox", { name: "Prompt", exact: true })
+  const control = component.getByRole("button", { name: "Choose model variant" })
+
+  await control.click()
+  await page.getByRole("menuitemradio", { name: "default", exact: true }).click()
+  await input.click()
+  await page.mouse.move(0, 0)
+  await expect(control).toHaveCSS("opacity", "0")
+
+  // Tab through Add, Agent, and Model to the visually hidden thinking trigger.
+  for (let count = 0; count < 4; count++) await page.keyboard.press("Tab")
+  await expect(control).toBeFocused()
+  await expect(control).toHaveCSS("opacity", "1")
+  await page.keyboard.press("Enter")
+  await expect(control).toHaveAttribute("aria-expanded", "true")
+  await expect(page.getByRole("menuitemradio", { name: "default", exact: true })).toBeFocused()
+  await expect(control).toHaveCSS("opacity", "1")
+  await page.keyboard.press("Escape")
+  await expect(control).toHaveAttribute("aria-expanded", "false")
+  await expect(control).toBeFocused()
+  await expect(control).toHaveCSS("opacity", "1")
+  await page.keyboard.press("Enter")
+  await expect(page.getByRole("menuitemradio", { name: "default", exact: true })).toBeFocused()
+  await page.keyboard.press("End")
+  await expect(page.getByRole("menuitemradio", { name: "high", exact: true })).toBeFocused()
+  await page.keyboard.press("Enter")
+  await expect(control).toHaveText("high")
+  await expect(control).toBeFocused()
+  await page.keyboard.press("Enter")
+  await expect(page.getByRole("menuitemradio", { name: "default", exact: true })).toBeFocused()
+  await page.keyboard.press("Enter")
+  await expect(control).toHaveText("default")
+  await expect(control).toBeFocused()
+  await expect(control).toHaveCSS("opacity", "1")
+  await page.keyboard.press("Tab")
+  await expect(control).not.toBeFocused()
+  await expect(control).toHaveCSS("opacity", "0")
 })
@@ -0,0 +1,86 @@
+import { expect, story } from "../../storybook/playwright/story"
+
+for (const theme of ["light", "dark"]) {
+  for (const direction of ["ltr", "rtl"]) {
+    story(`shimmers once over the visible wordmark (${theme}, ${direction})`, async ({ mount, page }, testInfo) => {
+      await page.emulateMedia({ reducedMotion: "no-preference" })
+      const component = await mount("app-new-session-wordmark--reveal", { globals: { theme, direction } })
+      const logo = component.locator('[data-component="new-session-wordmark"]')
+      const reveal = logo.locator('[data-slot="wordmark-reveal"]')
+      const shimmer = logo.locator(".wordmark-shimmer")
+      const base = logo.locator("svg").first()
+      await expect(base).toHaveCSS("opacity", theme === "dark" ? "0.5" : "0.6")
+      await expect(shimmer).toHaveCSS("animation-iteration-count", "1")
+      await expect(logo.locator("g[mask]")).toHaveCount(0)
+      await expect(shimmer).toHaveCSS("color", "rgb(255, 255, 255)")
+      await expect(shimmer).toHaveCSS("mix-blend-mode", "screen")
+      await expect(base.locator("g[fill]")).toHaveAttribute("fill", "currentColor")
+
+      await logo.evaluate((element) => {
+        element.getAnimations({ subtree: true }).forEach((animation) => {
+          animation.pause()
+          animation.currentTime = 0
+        })
+      })
+      await expect(reveal).toHaveCSS("opacity", "1")
+      await expect(reveal).toHaveCSS("animation-name", "none")
+      await expect(shimmer).toHaveCSS("mask-position", "100% 0px")
+
+      await logo.evaluate((element) => {
+        element.getAnimations({ subtree: true }).forEach((animation) => (animation.currentTime = 400))
+      })
+      await expect(shimmer).toHaveCSS("mask-position", "50% 0px")
+      await expect(shimmer).toHaveCSS("opacity", theme === "dark" ? "0.0672" : "0.45")
+      const swept = await logo.screenshot({ path: testInfo.outputPath("wordmark-shimmer.png") })
+
+      await logo.evaluate((element) => {
+        element.getAnimations({ subtree: true }).forEach((animation) => animation.finish())
+      })
+      await expect(reveal).toHaveCSS("opacity", "1")
+      await expect(shimmer).toHaveCSS("opacity", "0")
+      await expect(shimmer).toHaveCSS("mask-position", "0% 0px")
+      const settled = await logo.screenshot({ path: testInfo.outputPath("wordmark-settled.png") })
+      const brightness = await page.evaluate(
+        async (screenshots) => {
+          const pixels = await Promise.all(
+            screenshots.map(async (screenshot) => {
+              const image = new Image()
+              image.src = `data:image/png;base64,${screenshot}`
+              await image.decode()
+              const canvas = new OffscreenCanvas(image.width, image.height)
+              const context = canvas.getContext("2d")
+              if (!context) throw new Error("Cannot read screenshot pixels")
+              context.drawImage(image, 0, 0)
+              return context.getImageData(0, 0, image.width, image.height).data
+            }),
+          )
+          return {
+            min: pixels[0].reduce(
+              (min, value, index) => (index % 4 === 3 ? min : Math.min(min, value - pixels[1][index])),
+              0,
+            ),
+            max: pixels[0].reduce(
+              (max, value, index) => (index % 4 === 3 ? max : Math.max(max, value - pixels[1][index])),
+              0,
+            ),
+          }
+        },
+        [swept.toString("base64"), settled.toString("base64")],
+      )
+      expect(brightness.min).toBeGreaterThanOrEqual(0)
+      expect(brightness.max).toBeGreaterThanOrEqual(5)
+    })
+  }
+
+  story(`shows a static wordmark with reduced motion (${theme})`, async ({ mount, page }) => {
+    await page.emulateMedia({ reducedMotion: "reduce" })
+    await page.setViewportSize({ width: 360, height: 640 })
+    const component = await mount("app-new-session-wordmark--reveal", { globals: { theme } })
+    const logo = component.locator('[data-component="new-session-wordmark"]')
+    await expect(logo.locator("svg").first()).toHaveCSS("opacity", theme === "dark" ? "0.5" : "0.6")
+    await expect(logo.locator('[data-slot="wordmark-reveal"]')).toHaveCSS("opacity", "1")
+    await expect(logo.locator(".wordmark-shimmer")).toHaveCSS("opacity", "0")
+    expect(await logo.evaluate((element) => element.getAnimations({ subtree: true }).length)).toBe(0)
+    expect(await logo.evaluate((element) => element.getBoundingClientRect().right <= innerWidth)).toBe(true)
+  })
+}
@@ -0,0 +1,18 @@
+import { expect, story } from "../../storybook/playwright/story"
+
+for (const theme of ["light", "dark"]) {
+  story(`keeps the Open in border visible without hovering (${theme})`, async ({ mount, page }, testInfo) => {
+    const component = await mount("ui-split-button--open-in", { globals: { theme } })
+    const control = component.locator('[data-component="split-button-v2"]')
+    await page.mouse.move(0, 0)
+    await expect(control).toBeVisible()
+    await expect(control).not.toHaveCSS("box-shadow", "none")
+    const border = await control.evaluate((element) => getComputedStyle(element).boxShadow)
+
+    await component.getByRole("button", { name: "Open options" }).hover()
+    await expect(control).toHaveCSS("box-shadow", border)
+    await page.mouse.move(0, 0)
+    await expect(control).toHaveCSS("box-shadow", border)
+    await control.screenshot({ path: testInfo.outputPath(`open-in-${theme}.png`) })
+  })
+}
@@ -0,0 +1,264 @@
+import { expect, story } from "../../storybook/playwright/story"
+
+story("maps grouped and collapsed switches to the timeline settings", async ({ mount }) => {
+  const component = await mount("settings-timeline-detail--interactive")
+  await component.getByRole("button", { name: "Advanced", exact: true }).click()
+  const shell = component.getByRole("group", { name: "Shell", exact: true })
+  const grouped = shell.getByRole("switch", { name: "Shell grouped", exact: true })
+  const collapsed = shell.getByRole("switch", { name: "Shell collapsed", exact: true })
+  const value = component.locator('[data-slot="timeline-detail-fixture-value"]')
+
+  await expect(component.getByRole("switch")).toHaveCount(9)
+  await expect(component.getByText("Activity", { exact: true })).toHaveCount(0)
+  await expect(grouped).toBeChecked()
+  await expect(collapsed).toBeChecked()
+  await shell.locator('[data-field="placement"] [data-slot="switch-control"]').click()
+  await shell.locator('[data-field="details"] [data-slot="switch-control"]').click()
+  await expect(value).toContainText('"shell":{"placement":"separate","details":"expanded"}')
+  await grouped.focus()
+  await grouped.press("Space")
+  await collapsed.focus()
+  await collapsed.press("Space")
+  await expect(value).toContainText('"shell":{"placement":"grouped","details":"collapsed"}')
+  await expect(component.getByRole("group", { name: "Subagents", exact: true }).getByRole("switch")).toHaveCount(1)
+})
+
+story("replaces hidden switches with solid lines and restores options", async ({ mount, page }) => {
+  const component = await mount("settings-timeline-detail--interactive")
+  await component.getByRole("button", { name: "Advanced", exact: true }).click()
+  const shell = component.getByRole("group", { name: "Shell", exact: true })
+  const visibility = shell.getByRole("button", { name: "Shell visibility" })
+  const label = shell.locator('[data-slot="timeline-detail-activity"] > label')
+  const color = await label.evaluate((element) => getComputedStyle(element).color)
+  const iconColor = await visibility.evaluate((element) => getComputedStyle(element).color)
+  await shell.locator('[data-field="placement"] [data-slot="switch-control"]').click()
+  await shell.locator('[data-field="details"] [data-slot="switch-control"]').click()
+  await visibility.hover()
+  await expect(page.getByRole("tooltip")).toHaveText("Hide")
+  expect(await page.getByRole("tooltip").evaluate((element) => element.getBoundingClientRect().bottom)).toBeLessThan(
+    await visibility.evaluate((element) => element.getBoundingClientRect().top),
+  )
+  await visibility.click()
+
+  await expect(visibility).toHaveAttribute("aria-pressed", "false")
+  await expect(shell.getByRole("switch")).toHaveCount(0)
+  await expect(shell.locator('[data-slot="timeline-detail-unavailable"]')).toHaveCount(2)
+  await expect(shell.locator('[data-slot="timeline-detail-unavailable"]').first()).toHaveCSS(
+    "border-top-style",
+    "solid",
+  )
+  await expect(component.locator('[data-slot="timeline-detail-fixture-value"]')).toContainText(
+    '"shell":{"placement":"hidden","details":"expanded"}',
+  )
+  await expect(label).not.toHaveCSS("color", color)
+  await expect(visibility.locator("use")).toHaveAttribute("href", "#opencode-v2-icon-outline-eye-slash")
+  await component.getByRole("button", { name: "Advanced", exact: true }).hover()
+  await expect(visibility).not.toHaveCSS("color", iconColor)
+  await expect(shell.locator('[data-slot="timeline-detail-unavailable"]').first()).toHaveCSS(
+    "border-top-color",
+    await visibility.evaluate((element) => getComputedStyle(element).color),
+  )
+  await visibility.hover()
+  await expect(page.getByRole("tooltip")).toHaveText("Show")
+  expect(await page.getByRole("tooltip").evaluate((element) => element.getBoundingClientRect().bottom)).toBeLessThan(
+    await visibility.evaluate((element) => element.getBoundingClientRect().top),
+  )
+  await visibility.click()
+
+  await expect(visibility).toHaveAttribute("aria-pressed", "true")
+  await expect(label).toHaveCSS("color", color)
+  await expect(visibility.locator("use")).toHaveAttribute("href", "#opencode-v2-icon-outline-eye")
+  await expect(shell.locator('[data-slot="timeline-detail-unavailable"]')).toHaveCount(0)
+  await expect(shell.getByRole("switch", { name: "Shell grouped", exact: true })).toBeEnabled()
+  await expect(shell.getByRole("switch", { name: "Shell collapsed", exact: true })).toBeEnabled()
+  await expect(shell.getByRole("switch", { name: "Shell grouped", exact: true })).not.toBeChecked()
+  await expect(shell.getByRole("switch", { name: "Shell collapsed", exact: true })).not.toBeChecked()
+})
+
+story("toggles visibility by clicking the activity label", async ({ mount }) => {
+  const component = await mount("settings-timeline-detail--interactive")
+  await component.getByRole("button", { name: "Advanced", exact: true }).click()
+  const shell = component.getByRole("group", { name: "Shell", exact: true })
+  const label = shell.locator('[data-slot="timeline-detail-activity"] > label')
+  const visibility = shell.getByRole("button", { name: "Shell visibility" })
+
+  await expect(label).toHaveCSS("cursor", "default")
+  await label.click()
+  await expect(visibility).toHaveAttribute("aria-pressed", "false")
+  await expect(shell.getByRole("switch")).toHaveCount(0)
+  await expect(shell.locator('[data-slot="timeline-detail-unavailable"]')).toHaveCount(2)
+  await label.click()
+  await expect(visibility).toHaveAttribute("aria-pressed", "true")
+  await expect(shell.getByRole("switch", { name: "Shell grouped", exact: true })).toBeChecked()
+  await expect(shell.getByRole("switch", { name: "Shell collapsed", exact: true })).toBeChecked()
+})
+
+story("only highlights the eye when hovering the icon, not its activity label", async ({ mount }) => {
+  const component = await mount("settings-timeline-detail--interactive")
+  await component.getByRole("button", { name: "Advanced", exact: true }).click()
+  const shell = component.getByRole("group", { name: "Shell", exact: true })
+  const visibility = shell.getByRole("button", { name: "Shell visibility" })
+  const label = shell.locator('[data-slot="timeline-detail-activity"] > label')
+
+  for (const hidden of [false, true]) {
+    if (hidden) await label.click()
+    await visibility.hover()
+    await expect(visibility).not.toHaveCSS("background-color", "rgba(0, 0, 0, 0)")
+    await label.hover()
+    await expect(visibility).toHaveCSS("background-color", "rgba(0, 0, 0, 0)")
+  }
+})
+
+story("opens advanced on returning to custom settings but not presets", async ({ mount }) => {
+  const component = await mount("settings-timeline-detail--interactive")
+  const advanced = component.getByRole("button", { name: "Advanced", exact: true })
+  await expect(advanced).toHaveAttribute("aria-expanded", "false")
+  await advanced.click()
+  await component.locator('[data-category="shell"][data-field="placement"] [data-slot="switch-control"]').click()
+  await expect(component.getByRole("slider")).toHaveAttribute("aria-valuetext", "Custom")
+  await advanced.click()
+  await expect(advanced).toHaveAttribute("aria-expanded", "false")
+  await component.getByRole("button", { name: "Leave settings" }).click()
+  await expect(advanced).toHaveCount(0)
+  await component.getByRole("button", { name: "Return to settings" }).click()
+  await expect(advanced).toHaveAttribute("aria-expanded", "true")
+  await expect(component.getByRole("switch", { name: "Shell grouped", exact: true })).not.toBeChecked()
+
+  await component.getByRole("slider").press("End")
+  await expect(component.getByRole("slider")).toHaveAttribute("aria-valuetext", "Everything")
+  await component.getByRole("button", { name: "Leave settings" }).click()
+  await component.getByRole("button", { name: "Return to settings" }).click()
+  await expect(advanced).toHaveAttribute("aria-expanded", "false")
+})
+
+story("keeps visibility and switches in sync with the preset slider", async ({ mount }) => {
+  const component = await mount("settings-timeline-detail--interactive")
+  await component.getByRole("button", { name: "Advanced", exact: true }).click()
+  const slider = component.getByRole("slider", { name: "Timeline detail" })
+  await slider.focus()
+  await slider.press("Home")
+  await expect(slider).toHaveAttribute("aria-valuetext", "Messages only")
+  await expect(component.getByRole("switch")).toHaveCount(0)
+  await expect(component.locator('[data-slot="timeline-detail-unavailable"]')).toHaveCount(9)
+  await expect(component.locator('[data-action="timeline-detail-visibility"][aria-pressed="false"]')).toHaveCount(6)
+  await component.getByRole("button", { name: "Shell visibility" }).click()
+  await expect(component.getByRole("switch", { name: "Shell grouped", exact: true })).toBeChecked()
+  await expect(slider).toHaveAttribute("aria-valuetext", "Custom")
+  await slider.focus()
+  await slider.press("End")
+  await expect(slider).toHaveAttribute("aria-valuetext", "Everything")
+  await expect(component.getByRole("switch")).toHaveCount(9)
+  await expect(component.locator('[data-slot="timeline-detail-unavailable"]')).toHaveCount(0)
+  await expect(component.locator('[data-action="timeline-detail-visibility"][aria-pressed="true"]')).toHaveCount(6)
+  await expect(component.getByRole("switch", { checked: true })).toHaveCount(0)
+})
+
+for (const direction of ["ltr", "rtl"]) {
+  for (const theme of ["light", "dark"]) {
+    story(`fits narrow and wide layouts in ${direction}, ${theme}`, async ({ mount, page }, testInfo) => {
+      await page.setViewportSize({ width: 900, height: 900 })
+      const component = await mount("settings-timeline-detail--interactive", { globals: { direction, theme } })
+      await expect(component.locator('[data-slot="timeline-detail-summary"]')).toHaveCSS(
+        "color",
+        await component
+          .locator('[data-slot="settings-row-title"]')
+          .evaluate((element) => getComputedStyle(element).color),
+      )
+      await component.getByRole("button", { name: "Advanced", exact: true }).click()
+      const track = component.locator('[data-slot="timeline-detail-track"]')
+      await expect(track).toHaveCSS(
+        "--timeline-detail-track-background",
+        await track.evaluate(
+          (element, theme) =>
+            getComputedStyle(element)
+              .getPropertyValue(theme === "light" ? "--v2-background-bg-layer-04" : "--v2-background-bg-layer-03")
+              .trim(),
+          theme,
+        ),
+      )
+      if (theme === "light") {
+        await expect(track.locator("span").first()).toHaveCSS("background-image", "none")
+        await expect(track).toHaveCSS(
+          "--timeline-detail-marker-background",
+          await track.evaluate((element) => getComputedStyle(element).getPropertyValue("--v2-grey-500").trim()),
+        )
+      }
+      if (theme === "dark") {
+        await expect(track.locator("span").first()).not.toHaveCSS("background-image", "none")
+      }
+      await page.screenshot({ path: testInfo.outputPath(`timeline-${theme}-${direction}.png`) })
+      const list = component.locator('[data-slot="timeline-detail-list"]')
+      await expect(component.locator('[data-slot="timeline-detail-categories"]')).toHaveCSS("margin-top", "0px")
+      for (const [column, field] of [
+        [2, "placement"],
+        [3, "details"],
+      ] as const) {
+        const heading = await component
+          .locator(`[data-slot="timeline-detail-columns"] > :nth-child(${column})`)
+          .evaluate((element) => {
+            const rect = element.getBoundingClientRect()
+            return rect.x + rect.width / 2
+          })
+        const toggle = await component
+          .locator(`[data-category="shell"][data-field="${field}"] [data-slot="switch-control"]`)
+          .evaluate((element) => {
+            const rect = element.getBoundingClientRect()
+            return rect.x + rect.width / 2
+          })
+        expect(Math.abs(heading - toggle)).toBeLessThan(1)
+      }
+      await expect(component.locator('[data-slot="timeline-detail-activity"]').first()).toHaveCSS("gap", "12px")
+      for (const width of [900, 320]) {
+        await page.setViewportSize({ width, height: 900 })
+        await expect(component.getByRole("switch", { name: "Shell grouped", exact: true })).toBeVisible()
+        expect(await list.evaluate((element) => element.scrollWidth <= element.clientWidth)).toBe(true)
+        const visibility = component.getByRole("button", { name: "Shell visibility" })
+        await visibility.focus()
+        await visibility.press("Space")
+        await expect(visibility).toHaveAttribute("aria-pressed", "false")
+        await expect(component.getByRole("switch", { name: "Shell grouped", exact: true })).toHaveCount(0)
+        await expect(
+          component
+            .getByRole("group", { name: "Shell", exact: true })
+            .locator('[data-slot="timeline-detail-unavailable"]'),
+        ).toHaveCount(2)
+        await visibility.press("Space")
+        await expect(visibility).toHaveAttribute("aria-pressed", "true")
+
+        const slider = component.getByRole("slider", { name: "Timeline detail" })
+        const track = component.locator('[data-slot="timeline-detail-track"]')
+        expect(await track.evaluate((element) => element.getBoundingClientRect().width)).toBe(
+          await component
+            .locator('[data-slot="timeline-detail-scale"]')
+            .evaluate((element) => element.getBoundingClientRect().width),
+        )
+        await slider.focus()
+        await slider.press("Home")
+        for (const position of [0, 1, 2, 3, 4]) {
+          if (position > 0) await slider.press("ArrowUp")
+          await expect(track).toHaveCSS("--timeline-detail-progress", `${position * 25}%`)
+          const fill = await track.evaluate((element) => {
+            const style = getComputedStyle(element, "::before")
+            return {
+              fraction: parseFloat(style.width) / element.getBoundingClientRect().width,
+              start: style.getPropertyValue("inset-inline-start"),
+              color: style.backgroundColor,
+              remainder: getComputedStyle(element).backgroundColor,
+            }
+          })
+          expect(fill.fraction).toBeCloseTo(position / 4, 2)
+          expect(fill.start).toBe("0px")
+          expect(fill.color).not.toBe(fill.remainder)
+          const marker = await track.locator("span").nth(position).boundingBox()
+          const bounds = await track.boundingBox()
+          expect(marker).not.toBeNull()
+          expect(bounds).not.toBeNull()
+          expect((marker!.x + marker!.width / 2 - bounds!.x) / bounds!.width).toBeCloseTo(
+            direction === "rtl" ? 1 - position / 4 : position / 4,
+            2,
+          )
+        }
+      }
+    })
+  }
+}
@@ -0,0 +1,79 @@
+import { expect, story } from "../../storybook/playwright/story"
+
+story("enables Any only after confirmation and supports resetting the preview", async ({ mount }) => {
+  const component = await mount("app-current-session-surface--web-search-request")
+  const card = component.getByRole("region", { name: "Third-party web search" })
+  await expect(card.getByRole("button", { name: "Enable", exact: true })).toBeEnabled()
+  await expect(card.getByRole("button", { name: "Search provider Any", exact: true })).toBeVisible()
+  await card.getByRole("button", { name: "Enable", exact: true }).click()
+  await expect(component.getByRole("status")).toHaveText("Web search selection (local only): random")
+  await expect(card).toHaveCount(0)
+  await component.getByRole("button", { name: "Reset", exact: true }).click()
+  await expect(card.getByRole("button", { name: "Enable", exact: true })).toBeEnabled()
+})
+
+story("declining search is an explicit disabled selection", async ({ mount }) => {
+  const component = await mount("app-current-session-surface--web-search-request")
+  const card = component.getByRole("region", { name: "Third-party web search" })
+  await card.getByRole("button", { name: "Don’t use search", exact: true }).click()
+  await expect(component.getByRole("status")).toHaveText("Web search selection (local only): false")
+  await expect(card).toHaveCount(0)
+})
+
+story("sizes provider options to their content", async ({ mount, page }) => {
+  const component = await mount("app-current-session-surface--web-search-request")
+  await component.getByRole("button", { name: "Search provider Any", exact: true }).click()
+  const menu = page.getByRole("listbox", { name: "Search provider", exact: true })
+  await expect(menu).toBeVisible()
+  const metrics = await menu.evaluate((listbox) => {
+    const items = Array.from(listbox.querySelectorAll('[data-component="menu-v2-item"]'))
+    const longest = items
+      .flatMap((item) => {
+        const label = item.querySelector('[data-slot="menu-v2-item-content"]')?.getBoundingClientRect()
+        const check = item.querySelector('[data-slot="menu-v2-item-indicator"]')?.getBoundingClientRect()
+        return label && check ? [{ label, check }] : []
+      })
+      .toSorted((a, b) => b.label.width - a.label.width)[0]
+    return {
+      width: listbox.getBoundingClientRect().width,
+      gap: longest ? longest.check.left - longest.label.right : 0,
+    }
+  })
+  expect(metrics.width).toBeLessThan(160)
+  expect(metrics.gap).toBe(24)
+})
+
+for (const width of [360, 1200]) {
+  for (const direction of ["ltr", "rtl"]) {
+    for (const theme of ["light", "dark"]) {
+      story(`selects and confirms a provider at ${width}px in ${direction} ${theme}`, async ({ mount, page }) => {
+        await page.setViewportSize({ width, height: 900 })
+        const component = await mount("app-current-session-surface--web-search-request", {
+          globals: { theme, direction },
+        })
+        const card = component.getByRole("region", { name: "Third-party web search" })
+        const select = card.getByRole("button", { name: /^Search provider/ })
+        await expect(select).toBeEnabled()
+        await expect(card).toHaveCSS("direction", direction)
+        expect(await card.evaluate((node) => node.scrollWidth <= node.clientWidth)).toBe(true)
+        await select.focus()
+        await select.press("Enter")
+        const list = page.getByRole("listbox", { name: "Search provider", exact: true })
+        await expect(list).toHaveCSS("direction", direction)
+        await list.getByRole("option", { name: "Parallel", exact: true }).click()
+        await expect(select).toHaveText("Parallel")
+        await expect(card).toBeVisible()
+        await expect(component.getByRole("status")).toHaveText("Ready")
+        await expect(select).toBeFocused()
+        await select.press("Tab")
+        await expect(card.getByRole("button", { name: "Don’t use search", exact: true })).toBeFocused()
+        await page.keyboard.press("Tab")
+        const enable = card.getByRole("button", { name: "Enable", exact: true })
+        await expect(enable).toBeFocused()
+        await enable.press("Enter")
+        await expect(component.getByRole("status")).toHaveText("Web search selection (local only): parallel")
+        await expect(card).toHaveCount(0)
+      })
+    }
+  }
+}
@@ -43,6 +43,7 @@ The suite contains:
 - home-session click timing split between content and titlebar-tab paint
 - single-session tab close timing through stable home restoration
 - cached session repaint and mutation tracing
+- large-session search scan, first-result reveal, and highlight stabilization
 - streaming timeline throughput, RAF-gap, long-task, geometry, and remount diagnostics
 - retained renderer heap with a large model catalog across repeated session navigation

@@ -0,0 +1,47 @@
+import { benchmark, expect } from "./benchmark"
+import { openCommandPalette } from "../utils/command-palette"
+
+benchmark.use({
+  viewport: { width: 1440, height: 900 },
+  serviceWorkers: "block",
+  traceScope: "interaction",
+  trace: "off",
+  video: "off",
+})
+
+for (const home of [false, true]) {
+  benchmark(`command lookup from ${home ? "home" : "session"}`, async ({ page, report }) => {
+    const { dialog, input } = await openCommandPalette(page, home)
+    const title = home ? "Open settings" : "Copy Session ID"
+    const query = home ? "open settings" : "copy session"
+    // Measure input-to-selected-result in the renderer, without assertion polling overhead.
+    await input.evaluate((element, title) => {
+      element.addEventListener(
+        "input",
+        () => {
+          performance.mark("palette-input")
+          const observer = new MutationObserver(() => {
+            if (document.querySelectorAll('[role="dialog"] [role="option"]').length !== 1) return
+            const selected = document.querySelector('[role="dialog"] [role="option"][aria-selected="true"]')
+            if (!selected?.textContent?.includes(title)) return
+            performance.measure("palette-result", "palette-input")
+            observer.disconnect()
+          })
+          observer.observe(document, { subtree: true, childList: true, attributes: true, characterData: true })
+        },
+        { once: true, capture: true },
+      )
+    }, title)
+    await input.fill(query)
+    await expect(dialog.getByRole("option")).toHaveCount(1)
+    await expect(dialog.getByRole("option", { name: new RegExp(`^${title}(?:$| )`) })).toHaveAttribute(
+      "aria-selected",
+      "true",
+    )
+    const result = await page.evaluate(() =>
+      performance.getEntriesByName("palette-result").map((entry) => entry.duration),
+    )
+    expect(result).toHaveLength(1)
+    report({ inputToResultMs: result[0] }, { home, query, data: "fixture; immediate server responses" })
+  })
+}
@@ -0,0 +1,69 @@
+import { benchmark, expect } from "../benchmark"
+import { buildInitialStreamEvent, setupTimelineBenchmark, textPartID } from "./session-timeline-benchmark.fixture"
+import {
+  collectTimelineSearchMetrics,
+  installTimelineSearchProbe,
+  waitForStableTimelineSearch,
+} from "./session-timeline-search-probe"
+
+benchmark("searches a large virtualized session and reveals the first result", async ({ page, report }) => {
+  benchmark.setTimeout(180_000)
+  const historyTurns = Number(process.env.TIMELINE_SEARCH_HISTORY_TURNS ?? 320)
+  const completionTimeout = Number(process.env.TIMELINE_SEARCH_COMPLETION_TIMEOUT_MS ?? 60_000)
+  const query = "Historical prompt"
+  const targetPartID = "msg_0000_0000_a_user:text:0"
+  const expectedCounter = `1/${historyTurns}`
+  const fixture = await setupTimelineBenchmark(page, {
+    historyTurns,
+    eventBatch: 1,
+  })
+
+  fixture.transport.enqueue(buildInitialStreamEvent(1))
+  await expect(fixture.text).toContainText("Implementation plan")
+  await fixture.scrollToBottom()
+  await fixture.waitForStableGeometry()
+  // Chromium reserves the physical shortcut for its native find overlay, so request the same controller path directly.
+  await page.evaluate(() => document.dispatchEvent(new Event("opencode:timeline-search-open")))
+
+  const search = page.locator('[data-component="timeline-search-bar"]')
+  const field = search.getByRole("searchbox", { name: "Find..." })
+  const count = search.locator('[data-slot="timeline-search-count"]')
+  const target = page.locator(`[data-timeline-part-id="${targetPartID}"]`)
+  await expect(field).toBeVisible()
+  await expect(field).toBeFocused()
+  await installTimelineSearchProbe(page, { targetPartID })
+
+  await field.fill(query)
+  await expect(count).toHaveText(expectedCounter)
+  await expect(target).toBeVisible({ timeout: completionTimeout })
+  await waitForStableTimelineSearch(page, { counter: expectedCounter, targetPartID, timeout: completionTimeout })
+  const metrics = await collectTimelineSearchMetrics(page, { counter: expectedCounter, targetPartID })
+
+  expect(metrics.summary.handlerDurationMs).toBeDefined()
+  expect(metrics.summary.firstCountObservedMs).toBeDefined()
+  expect(metrics.summary.firstTargetVisibleMs).toBeDefined()
+  expect(metrics.summary.firstActiveHighlightObservedMs).toBeDefined()
+  expect(metrics.summary.stableResultObservedMs).toBeDefined()
+  expect(metrics.summary.activeHighlightRanges).toBe(1)
+  report(metrics, { historyTurns, query, expectedMatches: historyTurns })
+
+  // Check navigation and the V2 assistant content IDs outside the measured interval.
+  await field.press("Enter")
+  await expect(count).toHaveText(`2/${historyTurns}`)
+  await field.press("Shift+Enter")
+  await expect(count).toHaveText(expectedCounter)
+  await field.fill("Implementation plan")
+  await expect(count).toHaveText("1/1")
+  await expect(fixture.text).toBeInViewport()
+  await expect
+    .poll(() =>
+      page.evaluate(() => {
+        const range = [...(CSS.highlights.get("timeline-search-hit-active") ?? [])][0]
+        return range?.startContainer.parentElement?.closest<HTMLElement>("[data-timeline-part-id]")?.dataset
+          .timelinePartId
+      }),
+    )
+    .toBe(textPartID)
+  await field.press("Escape")
+  await expect(search).toBeHidden()
+})
@@ -0,0 +1,176 @@
+import type { Page } from "@playwright/test"
+
+export type TimelineSearchSample = {
+  observedAtMs: number
+  counter: string
+  targetMounted: boolean
+  targetVisible: boolean
+  targetTopPx?: number
+  activeRanges: number
+  activePartID?: string
+  activeVisible: boolean
+  scrollTopPx: number
+}
+
+type TimelineSearchProbe = {
+  samples: TimelineSearchSample[]
+  handlerDurationMs?: number
+  initialScrollTopPx: number
+  stop: () => void
+}
+
+export async function installTimelineSearchProbe(page: Page, input: { targetPartID: string }) {
+  await page.evaluate(({ targetPartID }) => {
+    const search = document.querySelector<HTMLElement>('[data-component="timeline-search-bar"]')
+    const field = search?.querySelector<HTMLInputElement>('[data-slot="text-input-v2-input"]')
+    const root = [...document.querySelectorAll<HTMLElement>(".scroll-view__viewport")].find((element) =>
+      element.querySelector("[data-timeline-row]"),
+    )
+    if (!search || !field || !root) throw new Error("missing timeline search benchmark nodes")
+
+    const samples: TimelineSearchSample[] = []
+    const initialScrollTopPx = root.scrollTop
+    let startedAt: number | undefined
+    let handlerDurationMs: number | undefined
+    let frame: number | undefined
+    let running = true
+
+    const visibleInRoot = (rect: DOMRect) => {
+      const viewport = root.getBoundingClientRect()
+      return rect.width > 0 && rect.height > 0 && rect.bottom > viewport.top && rect.top < viewport.bottom
+    }
+    const sample = () => {
+      if (!running || startedAt === undefined) return
+      frame = requestAnimationFrame(() => {
+        frame = undefined
+        setTimeout(() => {
+          if (!running || startedAt === undefined) return
+          const target = root.querySelector<HTMLElement>(`[data-timeline-part-id="${targetPartID}"]`)
+          const targetRect = target?.getBoundingClientRect()
+          const highlight = CSS.highlights.get("timeline-search-hit-active")
+          const ranges = highlight ? [...highlight] : []
+          const active = ranges.find((range): range is Range => range instanceof Range)
+          const activeRect = active?.getBoundingClientRect()
+          const activeElement =
+            active?.startContainer instanceof Element ? active.startContainer : active?.startContainer.parentElement
+
+          samples.push({
+            observedAtMs: performance.now() - startedAt,
+            counter:
+              search.querySelector<HTMLElement>('[data-slot="timeline-search-count"]')?.textContent?.trim() ?? "",
+            targetMounted: !!target,
+            targetVisible: !!targetRect && visibleInRoot(targetRect),
+            targetTopPx: targetRect?.top,
+            activeRanges: ranges.length,
+            activePartID: activeElement?.closest<HTMLElement>("[data-timeline-part-id]")?.dataset.timelinePartId,
+            activeVisible: !!activeRect && visibleInRoot(activeRect),
+            scrollTopPx: root.scrollTop,
+          })
+          sample()
+        }, 0)
+      })
+    }
+    const onInputCapture = (event: Event) => {
+      if (event.target !== field || startedAt !== undefined) return
+      startedAt = performance.now()
+      sample()
+    }
+    const onInput = (event: Event) => {
+      if (event.target !== field || startedAt === undefined || handlerDurationMs !== undefined) return
+      handlerDurationMs = performance.now() - startedAt
+    }
+    document.addEventListener("input", onInputCapture, { capture: true })
+    document.addEventListener("input", onInput)
+    ;(window as Window & { __timelineSearchBenchmark?: TimelineSearchProbe }).__timelineSearchBenchmark = {
+      samples,
+      initialScrollTopPx,
+      get handlerDurationMs() {
+        return handlerDurationMs
+      },
+      stop: () => {
+        running = false
+        document.removeEventListener("input", onInputCapture, { capture: true })
+        document.removeEventListener("input", onInput)
+        if (frame !== undefined) cancelAnimationFrame(frame)
+      },
+    }
+  }, input)
+}
+
+export async function waitForStableTimelineSearch(
+  page: Page,
+  input: { counter: string; targetPartID: string; timeout: number },
+) {
+  await page.waitForFunction(
+    ({ counter, targetPartID }) => {
+      const samples = (window as Window & { __timelineSearchBenchmark?: TimelineSearchProbe }).__timelineSearchBenchmark
+        ?.samples
+      if (!samples) return false
+      return samples.some((_, index) => {
+        const stable = samples.slice(index, index + 3)
+        if (stable.length !== 3) return false
+        return stable.every(
+          (sample, sampleIndex) =>
+            sample.counter === counter &&
+            sample.targetVisible &&
+            sample.activeRanges === 1 &&
+            sample.activePartID === targetPartID &&
+            sample.activeVisible &&
+            (sampleIndex === 0 ||
+              (Math.abs(sample.scrollTopPx - stable[sampleIndex - 1]!.scrollTopPx) <= 1 &&
+                Math.abs((sample.targetTopPx ?? Infinity) - (stable[sampleIndex - 1]!.targetTopPx ?? -Infinity)) <= 1)),
+        )
+      })
+    },
+    { counter: input.counter, targetPartID: input.targetPartID },
+    { timeout: input.timeout },
+  )
+}
+
+export async function collectTimelineSearchMetrics(page: Page, input: { counter: string; targetPartID: string }) {
+  const result = await page.evaluate(() => {
+    const probe = (window as Window & { __timelineSearchBenchmark?: TimelineSearchProbe }).__timelineSearchBenchmark
+    if (!probe) throw new Error("missing timeline search benchmark probe")
+    probe.stop()
+    return {
+      samples: probe.samples,
+      handlerDurationMs: probe.handlerDurationMs,
+      initialScrollTopPx: probe.initialScrollTopPx,
+    }
+  })
+  const first = (predicate: (sample: TimelineSearchSample) => boolean) => result.samples.find(predicate)?.observedAtMs
+  const stable = result.samples.findIndex((_, index) => {
+    const samples = result.samples.slice(index, index + 3)
+    if (samples.length !== 3) return false
+    return samples.every(
+      (sample, sampleIndex) =>
+        sample.counter === input.counter &&
+        sample.targetVisible &&
+        sample.activeRanges === 1 &&
+        sample.activePartID === input.targetPartID &&
+        sample.activeVisible &&
+        (sampleIndex === 0 ||
+          (Math.abs(sample.scrollTopPx - samples[sampleIndex - 1]!.scrollTopPx) <= 1 &&
+            Math.abs((sample.targetTopPx ?? Infinity) - (samples[sampleIndex - 1]!.targetTopPx ?? -Infinity)) <= 1)),
+    )
+  })
+  const final = result.samples.at(-1)
+
+  return {
+    summary: {
+      handlerDurationMs: result.handlerDurationMs,
+      firstCountObservedMs: first((sample) => sample.counter === input.counter),
+      firstTargetMountedMs: first((sample) => sample.targetMounted),
+      firstTargetVisibleMs: first((sample) => sample.targetVisible),
+      firstActiveHighlightObservedMs: first(
+        (sample) => sample.activeRanges === 1 && sample.activePartID === input.targetPartID && sample.activeVisible,
+      ),
+      stableResultObservedMs: stable >= 0 ? result.samples[stable + 2]?.observedAtMs : undefined,
+      initialScrollTopPx: result.initialScrollTopPx,
+      finalScrollTopPx: final?.scrollTopPx,
+      scrollDistancePx: final === undefined ? undefined : Math.abs(result.initialScrollTopPx - final.scrollTopPx),
+      activeHighlightRanges: final?.activeRanges,
+    },
+    samples: result.samples,
+  }
+}
@@ -0,0 +1,50 @@
+import { expect, test } from "@playwright/test"
+import { openCommandPalette, paletteSession } from "../utils/command-palette"
+
+test.use({ serviceWorkers: "block" })
+
+test("failed event-driven reads report an error and recover without an unhandled rejection", async ({ page }) => {
+  const errors: string[] = []
+  page.on("pageerror", (error) => errors.push(error.message))
+  const palette = await openCommandPalette(page)
+  const path = `**/api/session/${paletteSession.id}`
+  await page.route(path, (route) => route.abort("failed"))
+  const requested = page.waitForRequest(path)
+  await page.evaluate((sessionID) => {
+    const host = window as Window & { __mockServerStream?: { push: (events: unknown[]) => void } }
+    if (!host.__mockServerStream) throw new Error("Missing fixture event stream")
+    host.__mockServerStream.push([
+      {
+        id: "evt_failed_refresh",
+        created: 2,
+        type: "session.viewed",
+        durable: { aggregateID: sessionID, seq: 1, version: 1 },
+        data: { sessionID, idle: 2 },
+      },
+    ])
+  }, paletteSession.id)
+  await requested
+  await expect(page.getByText("Request failed", { exact: true })).toBeVisible()
+  await palette.input.fill("copy session")
+  await expect(palette.dialog.getByRole("option", { name: "Copy Session ID", exact: true })).toHaveAttribute(
+    "aria-selected",
+    "true",
+  )
+  await palette.input.press("Escape")
+  await page.unroute(path)
+  await page.evaluate((sessionID) => {
+    const host = window as Window & { __mockServerStream?: { push: (events: unknown[]) => void } }
+    if (!host.__mockServerStream) throw new Error("Missing fixture event stream")
+    host.__mockServerStream.push([
+      {
+        id: "evt_recovered_refresh",
+        created: 3,
+        type: "session.renamed",
+        durable: { aggregateID: sessionID, seq: 2, version: 1 },
+        data: { sessionID, title: "Recovered session" },
+      },
+    ])
+  }, paletteSession.id)
+  await expect(page.getByRole("heading", { name: "Recovered session", exact: true })).toBeVisible()
+  expect(errors).toEqual([])
+})
@@ -0,0 +1,102 @@
+import { expect, test } from "@playwright/test"
+import { captureConsoleWarnings, openCommandPalette, paletteSession } from "../utils/command-palette"
+
+test.use({ serviceWorkers: "block", permissions: ["clipboard-read", "clipboard-write"] })
+
+test("copies the session ID while file and session searches are still pending", async ({ page }) => {
+  const warnings = captureConsoleWarnings(page)
+  const { dialog, input } = await openCommandPalette(page)
+  const release = Promise.withResolvers<void>()
+  await page.route(/\/api\/(session\?|fs\/find\?)/, async (route) => {
+    await release.promise
+    await route.fallback()
+  })
+  await input.pressSequentially("copy session")
+  const copy = dialog.getByRole("option", { name: "Copy Session ID", exact: true })
+  await expect(copy).toHaveAttribute("aria-selected", "true")
+  await input.press("Enter")
+  await expect(dialog).toHaveCount(0)
+  await expect.poll(() => page.evaluate(() => navigator.clipboard.readText())).toBe(paletteSession.id)
+  await expect(page.locator('[data-testid^="toast-v2-"] [data-slot="icon-svg"]')).toBeVisible()
+  expect(warnings).toEqual([])
+  release.resolve()
+})
+
+test("home commands do not wait for session search", async ({ page }) => {
+  const { dialog, input } = await openCommandPalette(page, true)
+  const release = Promise.withResolvers<void>()
+  await page.route("**/api/session?*", async (route) => {
+    await release.promise
+    await route.fallback()
+  })
+  await input.fill("open settings")
+  await expect(dialog.getByRole("option")).toHaveCount(1)
+  await expect(dialog.getByRole("option", { name: /^Open settings/ })).toHaveAttribute("aria-selected", "true")
+  await input.press("Enter")
+  await expect(page).toHaveURL("/settings")
+  await expect(page.getByTestId("settings-screen").getByRole("tab", { name: "Preferences", exact: true })).toBeVisible()
+  release.resolve()
+})
+
+test("appends search results without resetting the selected command", async ({ page }) => {
+  const { dialog, input } = await openCommandPalette(page)
+  const files = Promise.withResolvers<void>()
+  const sessions = Promise.withResolvers<void>()
+  await page.route("**/api/fs/find?*", async (route) => {
+    await files.promise
+    await route.fulfill({ json: { data: [{ path: "copy.txt", type: "file" }] } })
+  })
+  await page.route("**/api/session?*", async (route) => {
+    await sessions.promise
+    await route.fulfill({
+      json: {
+        data: [{ ...paletteSession, location: { directory: paletteSession.directory }, title: "Copy fixture" }],
+      },
+    })
+  })
+  await input.fill("copy")
+  const project = dialog.getByRole("option", { name: "Copy Project ID", exact: true })
+  await expect(project).toBeVisible()
+  // Select a non-first command with the keyboard before remote results arrive.
+  await input.press("ArrowDown")
+  await expect(project).toHaveAttribute("aria-selected", "true")
+  files.resolve()
+  await expect(dialog.getByRole("option", { name: "/ copy.txt", exact: true })).toBeVisible()
+  await expect(project).toHaveAttribute("aria-selected", "true")
+  // File results are usable even while sessions are still pending.
+  sessions.resolve()
+  await expect(dialog.getByRole("option", { name: /Copy fixture/ })).toBeVisible()
+  await expect(project).toHaveAttribute("aria-selected", "true")
+  await input.fill("copy session")
+  await expect(dialog.getByRole("option", { name: "Copy Session ID", exact: true })).toHaveAttribute(
+    "aria-selected",
+    "true",
+  )
+  await expect(dialog.getByRole("option", { name: "Copy Project ID", exact: true })).toHaveCount(0)
+})
+
+test("keeps the automatically selected file when session results arrive later", async ({ page }) => {
+  const { dialog, input } = await openCommandPalette(page)
+  const sessions = Promise.withResolvers<void>()
+  await page.route("**/api/fs/find?*", (route) =>
+    route.fulfill({ json: { data: [{ path: "README.md", type: "file" }] } }),
+  )
+  await page.route("**/api/session?*", async (route) => {
+    await sessions.promise
+    await route.fulfill({
+      json: {
+        data: [{ ...paletteSession, location: { directory: paletteSession.directory }, title: "README work" }],
+      },
+    })
+  })
+  await input.fill("README")
+  const file = dialog.getByRole("option", { name: "/ README.md", exact: true })
+  await expect(file).toHaveAttribute("aria-selected", "true")
+  sessions.resolve()
+  await expect(dialog.getByRole("option", { name: /README work/ })).toBeVisible()
+  await expect(file).toHaveAttribute("aria-selected", "true")
+  await input.press("Enter")
+  await expect(dialog).toHaveCount(0)
+  await expect(page.getByRole("tab", { name: "README.md", exact: true })).toBeVisible()
+  await expect(page.getByRole("heading", { name: paletteSession.title, exact: true })).toBeVisible()
+})
@@ -0,0 +1,77 @@
+import { expect, test } from "@playwright/test"
+import { captureConsoleWarnings, openCommandPalette } from "../utils/command-palette"
+
+test.use({ serviceWorkers: "block", video: "off" })
+
+test("opening and closing files does not duplicate tab commands", async ({ page }) => {
+  const warnings = captureConsoleWarnings(page)
+  const palette = await openCommandPalette(page)
+  await page.route("**/api/fs/find?*", (route) =>
+    route.fulfill({
+      headers: { "access-control-allow-origin": "*" },
+      json: { data: [{ path: "fixture.txt", type: "file" }] },
+    }),
+  )
+  await palette.input.fill("fixture.txt")
+  await palette.dialog.getByRole("option", { name: /fixture\.txt/ }).click()
+  const file = page.getByRole("tab", { name: /fixture\.txt/ })
+  await expect(file).toBeVisible()
+  await expect(palette.dialog).toHaveCount(0)
+  await page
+    .getByRole("complementary", { name: "Review and files" })
+    .getByRole("button", { name: "Close tab", exact: true })
+    .click()
+  await expect(file).toHaveCount(0)
+  await expect(page.getByRole("heading", { name: "Palette fixture session", exact: true })).toBeVisible()
+  expect(warnings).toEqual([])
+})
+
+test("navigation replaces commands without retaining disposed owners", async ({ page }) => {
+  const warnings = captureConsoleWarnings(page)
+  const palette = await openCommandPalette(page, true)
+  await palette.input.press("Escape")
+  await expect(palette.dialog).toHaveCount(0)
+  await page
+    .getByRole("region", { name: "Recent sessions" })
+    .getByRole("button", { name: /Palette fixture session/ })
+    .click()
+  await expect(page.locator('[data-component="composer-editor"]')).toBeEditable()
+  await page.keyboard.press("ControlOrMeta+t")
+  await expect(page).toHaveURL(/\/new-session\?/)
+  await expect(page.locator('[data-component="composer-editor"]')).toBeEditable()
+  await page.locator('[data-component="composer-editor"]').blur()
+  await page.keyboard.press("Control+l")
+  await expect(page.locator('[data-component="composer-editor"]')).toBeFocused()
+  await page.keyboard.press("ControlOrMeta+Shift+P")
+  const dialog = page.getByRole("dialog")
+  await expect(dialog.getByRole("textbox")).toBeFocused()
+  await expect(dialog.getByRole("textbox")).toHaveAttribute("placeholder", "Search files, commands, and sessions")
+  await dialog.getByRole("textbox").fill("copy session")
+  await expect(dialog.getByRole("option", { name: "Copy Session ID", exact: true })).toHaveCount(0)
+  await dialog.getByRole("textbox").press("Escape")
+  await expect(dialog).toHaveCount(0)
+  await page.locator("[data-titlebar-tab-link]").filter({ hasText: "Palette fixture session" }).click()
+  await expect(page.getByRole("heading", { name: "Palette fixture session", exact: true })).toBeVisible()
+  for (const count of [3, 4]) {
+    await page.getByRole("button", { name: "New session", exact: true }).click()
+    await expect(page.locator("[data-titlebar-tab-link]")).toHaveCount(count)
+    await expect(page.locator('[data-component="composer-editor"]')).toBeEditable()
+  }
+  await page.setViewportSize({ width: 600, height: 800 })
+  await page.locator('[data-slot="mobile-tabs-trigger"]').click()
+  await expect(page.locator('[data-slot="mobile-tabs-drawer"] [data-titlebar-tab-link]')).toHaveCount(4)
+  await page.setViewportSize({ width: 1280, height: 800 })
+  await expect(page.locator('[data-slot="titlebar-tabs"] [data-titlebar-tab-link]')).toHaveCount(4)
+  await page.keyboard.press("ControlOrMeta+w")
+  await expect(page.locator("[data-titlebar-tab-link]")).toHaveCount(3)
+  await page.locator("[data-titlebar-tab-link]").filter({ hasText: "Palette fixture session" }).click()
+  await expect(page.getByRole("heading", { name: "Palette fixture session", exact: true })).toBeVisible()
+  await page.keyboard.press("ControlOrMeta+Shift+P")
+  await expect(dialog.getByRole("textbox")).toBeFocused()
+  await dialog.getByRole("textbox").fill("copy session")
+  await expect(dialog.getByRole("option", { name: "Copy Session ID", exact: true })).toHaveAttribute(
+    "aria-selected",
+    "true",
+  )
+  expect(warnings).toEqual([])
+})
@@ -0,0 +1,195 @@
+import { expect, test, type Page } from "@playwright/test"
+import { base64Encode } from "@opencode-ai/util/encode"
+import { mockOpenCodeServer } from "../utils/mock-server"
+
+const directory = "/repo/slash-skills"
+const projectID = "proj_slash_skills"
+const sessionID = "ses_slash_skills"
+const server = `http://${process.env.PLAYWRIGHT_SERVER_HOST ?? "127.0.0.1"}:${process.env.PLAYWRIGHT_SERVER_PORT ?? "4096"}`
+
+async function setup(page: Page, queued = false) {
+  const prompts: Record<string, unknown>[] = []
+  await mockOpenCodeServer(page, {
+    directory,
+    project: {
+      id: projectID,
+      worktree: directory,
+      vcs: "git",
+      name: "slash-skills",
+      time: { created: 1700000000000, updated: 1700000000000 },
+      sandboxes: [],
+    },
+    provider: {
+      all: [
+        {
+          id: "opencode",
+          name: "OpenCode",
+          models: { test: { id: "test", name: "Test", limit: { context: 200_000 } } },
+        },
+      ],
+      connected: ["opencode"],
+      default: { providerID: "opencode", modelID: "test" },
+    },
+    sessions: [
+      {
+        id: sessionID,
+        projectID,
+        directory,
+        title: "Slash skills",
+        version: "dev",
+        time: { created: 1700000000000, updated: 1700000000000 },
+      },
+    ],
+    pageMessages: () => ({ items: [] }),
+    sessionStatus: queued ? { [sessionID]: { type: "running" } } : {},
+    inbox: queued
+      ? [
+          {
+            id: "inb_slash_skill",
+            sessionID,
+            timeCreated: 1700000000000,
+            type: "user",
+            payload: { text: "Explain caching" },
+            delivery: "queue",
+          },
+        ]
+      : [],
+    findFiles: ({ query }) =>
+      query.includes("cache")
+        ? [
+            {
+              name: "cache.ts",
+              path: "src/cache.ts",
+              absolute: `${directory}/src/cache.ts`,
+              type: "file",
+              ignored: false,
+            },
+          ]
+        : [],
+    onPrompt: ({ body }) => prompts.push(body),
+  })
+  await page.route("**/api/skill?*", (route) =>
+    route.fulfill({
+      json: {
+        location: { directory, project: { id: projectID, directory, canonical: directory } },
+        data: [
+          { id: "show-me", slash: true, autoinvoke: false },
+          { id: "hidden", slash: false },
+          { id: "implicit" },
+          { id: "review", slash: true },
+          { id: "model", slash: true },
+        ].map((skill) => ({
+          name: skill.id === "show-me" ? "Show Me" : skill.id,
+          description: "Explain the current topic visually",
+          location: `/skills/${skill.id}/SKILL.md`,
+          content: "Explain visually",
+          ...skill,
+        })),
+      },
+      headers: { "access-control-allow-origin": "*" },
+    }),
+  )
+  await page.route("**/api/command?*", (route) =>
+    route.fulfill({
+      json: {
+        location: { directory, project: { id: projectID, directory, canonical: directory } },
+        data: [{ name: "review", description: "Review changes" }],
+      },
+      headers: { "access-control-allow-origin": "*" },
+    }),
+  )
+  await page.goto(`/server/${base64Encode(server)}/session/${sessionID}`)
+  const composer = page.locator('[data-component="composer"]')
+  const editor = composer.locator('[data-component="composer-editor"]')
+  await expect(editor).toBeEditable()
+  return { prompts, composer, editor }
+}
+
+for (const selection of ["keyboard", "pointer"]) {
+  test(`selects and submits a manual-only slash skill with ${selection}`, async ({ page }) => {
+    const { prompts, editor } = await setup(page)
+    await editor.fill("/show")
+    const skill = page.locator('[data-suggestion-id="skill:show-me"]')
+    await expect(skill).toContainText("/show-me")
+    await page.screenshot({ path: test.info().outputPath("slash-menu.png") })
+    if (selection === "keyboard") await editor.press("Enter")
+    if (selection === "pointer") await skill.click()
+    await expect(editor).toHaveText("/show-me ")
+    await expect(editor).toBeFocused()
+    await editor.pressSequentially("explain caching")
+    await editor.press("Enter")
+    await expect.poll(() => prompts.length).toBe(1)
+    expect(prompts[0]).toMatchObject({
+      text: "/show-me explain caching",
+      skills: [{ id: "show-me", mention: { start: 0, end: 8, text: "/show-me" } }],
+    })
+    await expect(editor).toBeEmpty()
+  })
+}
+
+test("respects slash flags and command precedence without hiding context skills", async ({ page }) => {
+  const { editor } = await setup(page)
+  await editor.fill("/")
+  await expect(page.locator('[data-suggestion-id="skill:show-me"]')).toBeVisible()
+  await expect(page.locator('[data-suggestion-id="custom.review"]')).toBeVisible()
+  await expect(page.locator('[data-suggestion-id="model.choose"]')).toBeVisible()
+  for (const id of ["hidden", "implicit", "review", "model"]) {
+    await expect(page.locator(`[data-suggestion-id="skill:${id}"]`)).toHaveCount(0)
+  }
+  await editor.press("Escape")
+  await expect(page.locator('[data-suggestion-id="skill:show-me"]')).toHaveCount(0)
+  await expect(editor).toBeFocused()
+  await editor.fill("@hidden")
+  const hidden = page.locator('[data-suggestion-id="skill:hidden"]')
+  await expect(hidden).toContainText("@hidden")
+  await hidden.click()
+  await expect(editor).toHaveText("@hidden ")
+  await expect(editor).toBeFocused()
+})
+
+test("preserves structured attachments when adding a slash skill from the command menu", async ({ page }) => {
+  const { prompts, composer, editor } = await setup(page)
+  await editor.fill("explain @cache")
+  const file = page.locator('[data-suggestion-id="file:src/cache.ts"]')
+  await expect(file).toBeVisible()
+  await file.click()
+  await expect(editor).toHaveText("explain @src/cache.ts ")
+  await composer.getByRole("button", { name: "Add images and files" }).click()
+  await page.getByRole("menuitem", { name: "Commands" }).click()
+  const skill = page.locator('[data-suggestion-id="skill:show-me"]')
+  await expect(skill).toBeVisible()
+  await skill.click()
+  await expect(editor).toHaveText("/show-me explain @src/cache.ts ")
+  await expect(editor).toBeFocused()
+  await editor.press("Enter")
+  await expect.poll(() => prompts.length).toBe(1)
+  expect(prompts[0]).toMatchObject({
+    skills: [{ id: "show-me", mention: { start: 0, end: 8, text: "/show-me" } }],
+    files: [{ name: "cache.ts", mention: { start: 17, end: 30, text: "@src/cache.ts" } }],
+  })
+})
+
+for (const select of [true, false]) {
+  test(`keeps slash skills in queued edits (${select ? "selected" : "typed"})`, async ({ page }) => {
+    const { prompts, editor } = await setup(page, true)
+    const row = page.locator('[data-component="session-queue-row"]')
+    await row.getByText("Explain caching", { exact: true }).click()
+    await expect(editor).toHaveText("Explain caching")
+    if (select) {
+      await editor.fill("/show")
+      const skill = page.locator('[data-suggestion-id="skill:show-me"]')
+      await expect(skill).toBeVisible()
+      await skill.click()
+      await expect(editor).toHaveText("/show-me ")
+      await editor.pressSequentially("explain caching")
+    }
+    if (!select) await editor.fill("/show-me explain caching")
+    await editor.press("Enter")
+    await expect.poll(() => prompts.length).toBe(1)
+    expect(prompts[0]).toMatchObject({
+      delivery: "queue",
+      text: "/show-me explain caching",
+      skills: [{ id: "show-me", mention: { start: 0, end: 8, text: "/show-me" } }],
+    })
+  })
+}
@@ -86,7 +86,12 @@ async function mockServers(page: Page, requests: string[]) {
      }
      return json(route, url.pathname === "/api/project" ? [project] : { id: project.id, directory: current.directory })
    }
-    if (url.pathname === "/api/location") return json(route, { directory: current.directory })
+    if (url.pathname === "/api/location")
+      return json(route, {
+        directory: current.directory,
+        project: { id: current.projectID, directory: current.directory, canonical: current.directory },
+      })
+    if (url.pathname === "/api/worktree") return json(route, [{ directory: current.directory }])
    if (url.pathname === "/api/vcs")
      return json(route, {
        location: { directory: current.directory },
@@ -0,0 +1,60 @@
+import { expect, test } from "@playwright/test"
+import { mockOpenCodeServer } from "../utils/mock-server"
+import { expectAppVisible } from "../utils/waits"
+
+const draftID = "draft_new_session_dropzone"
+const directory = "C:/OpenCode/NewSessionDropzone"
+const server = `http://${process.env.PLAYWRIGHT_SERVER_HOST ?? "127.0.0.1"}:${process.env.PLAYWRIGHT_SERVER_PORT ?? "4096"}`
+
+test("shows the dropzone and attaches a dropped file", async ({ page }) => {
+  await mockOpenCodeServer(page, {
+    directory,
+    project: {
+      id: "proj_new_session_dropzone",
+      worktree: directory,
+      vcs: "git",
+      name: "new-session-dropzone",
+      time: { created: 1700000000000, updated: 1700000000000 },
+      sandboxes: [],
+    },
+    provider: { all: [], connected: [], default: {} },
+    sessions: [],
+    pageMessages: () => ({ items: [] }),
+  })
+  await page.addInitScript(
+    ({ directory, draftID, server }) => {
+      localStorage.setItem(
+        "opencode.global.dat:server",
+        JSON.stringify({
+          projects: { local: [{ worktree: directory, expanded: true }] },
+          lastProject: { local: directory },
+        }),
+      )
+      localStorage.setItem(
+        "opencode.window.browser.dat:tabs",
+        JSON.stringify([{ type: "draft", draftID, server, directory }]),
+      )
+    },
+    { directory, draftID, server },
+  )
+
+  await page.goto(`/new-session?draftId=${draftID}`)
+  await expectAppVisible(page.locator('[data-component="composer-editor"]'))
+
+  const surface = page.locator('[data-component="new-session"]')
+  const dropzone = page.locator('[data-component="session-dropzone"]')
+  const transfer = await page.evaluateHandle(() => {
+    const value = new DataTransfer()
+    value.items.add(new File(["Dropzone fixture"], "dropzone.txt", { type: "text/plain" }))
+    return value
+  })
+
+  await expect(dropzone).toHaveCount(0)
+  await surface.dispatchEvent("dragover", { dataTransfer: transfer })
+  await expect(dropzone).toHaveAttribute("data-visible", "true")
+  await expect(dropzone).toContainText("Drop files to add")
+
+  await surface.dispatchEvent("drop", { dataTransfer: transfer })
+  await expect(page.locator('[data-component="composer-attachments"]')).toContainText("dropzone.txt")
+  await expect(dropzone).toHaveCount(0)
+})
@@ -1,4 +1,5 @@
 import { expect, test, type Page } from "@playwright/test"
+import type { OpenCodeEvent } from "@opencode-ai/client/promise"
 import { base64Encode } from "@opencode-ai/util/encode"
 import { currentSession, mockOpenCodeServer } from "../utils/mock-server"
 import { expectAppVisible } from "../utils/waits"
@@ -25,6 +26,53 @@ for (const viewport of [
  { name: "desktop", width: 1280, height: 900 },
  { name: "mobile", width: 390, height: 844 },
 ]) {
+  test(`keeps Session in the tab until the generated title arrives on ${viewport.name}`, async ({ page }, testInfo) => {
+    await page.setViewportSize(viewport)
+    const mock = await openDraft(page, { untitled: true })
+    const label = page.locator(
+      viewport.name === "mobile"
+        ? '[data-slot="mobile-tab-title"]'
+        : '[data-titlebar-tab-slot][data-active="true"] [data-titlebar-tab-title]',
+    )
+    await expect(label).toHaveText("Session")
+    const pending = await submitPending(page, mock)
+    const spinner = page.locator(
+      viewport.name === "mobile"
+        ? '[data-slot="mobile-tabs-trigger"] [data-component="session-progress-indicator-v2"]'
+        : `[data-titlebar-tab-link][href="${sessionPath}${pending.sessionID}"] [data-component="session-progress-indicator-v2"]`,
+    )
+    await expect(spinner).toBeVisible()
+    await testInfo.attach("pending-tab-title", {
+      body: await page.screenshot(),
+      contentType: "image/png",
+    })
+    await expect(label).toHaveText("Session")
+
+    mock.worktree.resolve({ status: 200, json: { directory: workspace } })
+    await expect(pending.shimmer).toHaveCount(0)
+    await expect(page.locator('[data-action="composer-submit"]')).toBeEnabled()
+    await expect(label).toHaveText("Session")
+
+    if (viewport.name === "mobile") {
+      await label.click()
+      const drawer = page.locator('[data-slot="mobile-tabs-drawer"]')
+      const tab = drawer.locator(`[data-titlebar-tab-link][href="${sessionPath}${pending.sessionID}"]`)
+      await expect(tab.locator("[data-titlebar-tab-title]")).toHaveText("Session")
+      await tab.click()
+      await expect(drawer).toBeHidden()
+    }
+
+    mock.events.push({
+      id: "evt_generated_title",
+      type: "session.renamed",
+      created: Date.now(),
+      location: { directory: workspace },
+      durable: { aggregateID: pending.sessionID, seq: 1, version: 1 },
+      data: { sessionID: pending.sessionID, title: "Generated session title" },
+    })
+    await expect(label).toHaveText("Generated session title")
+  })
+
  test(`shows a pending workspace session immediately on ${viewport.name}`, async ({ page }, testInfo) => {
    await page.setViewportSize(viewport)
    const mock = await openDraft(page)
@@ -86,7 +134,7 @@ for (const viewport of [
      await expect(page).toHaveURL(pending.url)
      await expect(pending.message).toHaveAttribute("data-timeline-part-id", `${pending.messageID}:text:0`)
      await expect(pending.shimmer).toHaveAttribute("data-active", "true")
-      await expect(pending.title).toHaveText("New session")
+      await expect(pending.title).toHaveText("Session")
      await expect(editor).toHaveText(followUp)
      expect(mock.calls).toEqual(["worktree"])

@@ -190,7 +238,7 @@ for (const direction of ["ltr", "rtl"]) {
          !frame.message ||
          !frame.spinner ||
          frame.draft !== followUp ||
-          !["New session", "Created workspace session"].includes(frame.title ?? ""),
+          !["Session", "Created workspace session"].includes(frame.title ?? ""),
      ),
    ).toEqual([])
    const after = await title.boundingBox()
@@ -335,6 +383,61 @@ test("restores the draft after closing and revisiting a pending session that fai
  expect(mock.prompts).toEqual([])
 })

+test("executes a selected slash command after creating its worktree", async ({ page }, testInfo) => {
+  const events: OpenCodeEvent[] = []
+  const mock = await openDraft(page, { command: true, events: () => events.splice(0) })
+  const commands: { sessionID: string; body: Record<string, unknown> }[] = []
+  const expanded =
+    "Review the latest commit for correctness and regressions. Check the relevant tests and report actionable findings."
+  await page.route("**/api/session/*/command", async (route) => {
+    if (route.request().method() !== "POST") return route.fallback()
+    const sessionID = new URL(route.request().url()).pathname.split("/")[3]
+    commands.push({ sessionID, body: route.request().postDataJSON() })
+    // The server owns command expansion; the client receives the expanded inbox item.
+    events.push({
+      id: "evt_workspace_review",
+      type: "session.inbox.enqueued",
+      created: Date.now(),
+      durable: { aggregateID: sessionID, seq: 1, version: 1 },
+      data: {
+        sessionID,
+        inboxID: "msg_workspace_review",
+        item: { type: "user", payload: { text: expanded }, delivery: "steer" },
+      },
+    })
+    await route.fulfill({ status: 204, headers })
+  })
+  const editor = page.locator('[data-component="composer-editor"]')
+  await editor.fill("/review")
+  const suggestion = page.getByRole("button", { name: "/review Review changes", exact: true })
+  await expect(suggestion).toBeVisible()
+  await suggestion.click()
+  await expect(editor).toHaveText("/review")
+  const pending = await submitPending(page, mock, "/review latest commit")
+  await draftFollowUp(page)
+
+  mock.worktree.resolve({ status: 200, json: { directory: workspace } })
+
+  await expect
+    .poll(() => commands)
+    .toEqual([
+      {
+        sessionID: pending.sessionID,
+        body: { command: "review", text: "latest commit", files: [], agents: [], skills: [], delivery: "steer" },
+      },
+    ])
+  await expect(pending.shimmer).toHaveCount(0)
+  await expect(page.locator('[data-slot="user-message-text"]')).toHaveText(expanded)
+  await expect(editor).toHaveText(followUp)
+  await expect(page.locator('[data-action="composer-submit"]')).toBeEnabled()
+  expect(mock.creates).toEqual([expect.objectContaining({ id: pending.sessionID, location: { directory: workspace } })])
+  expect(mock.prompts).toEqual([])
+  await testInfo.attach("expanded-worktree-command", {
+    body: await page.screenshot({ path: testInfo.outputPath("expanded-worktree-command.png") }),
+    contentType: "image/png",
+  })
+})
+
 async function draftFollowUp(page: Page) {
  const editor = page.locator('[data-component="composer-editor"]')
  await editor.pressSequentially("!")
@@ -346,12 +449,16 @@ async function draftFollowUp(page: Page) {
  await expect(editor).toHaveText(followUp)
 }

-async function openDraft(page: Page, options?: { failSessionCreate?: boolean }) {
+async function openDraft(
+  page: Page,
+  options?: { failSessionCreate?: boolean; untitled?: boolean; command?: boolean; events?: () => OpenCodeEvent[] },
+) {
  const worktree = Promise.withResolvers<{ status: number; json: { directory?: string; message?: string } }>()
  const calls: string[] = []
  const worktreeRequests: Record<string, unknown>[] = []
  const creates: Record<string, unknown>[] = []
  const prompts: { sessionID: string; body: Record<string, unknown> }[] = []
+  const events: OpenCodeEvent[] = []
  const project = {
    id: projectID,
    worktree: directory,
@@ -378,24 +485,29 @@ async function openDraft(page: Page, options?: { failSessionCreate?: boolean })
    sessions,
    pageMessages: () => ({ items: [] }),
    onPrompt: (input) => prompts.push(input),
+    events: options?.events ?? (() => events.splice(0)),
  })
  page.on("request", (request) => {
    if (request.method() !== "POST") return
    const path = new URL(request.url()).pathname
-    if (path === `/api/worktree/${projectID}`) {
+    if (path === "/api/worktree") {
+      expect(new URL(request.url()).searchParams.get("location[directory]")).toBe(directory)
      calls.push("worktree")
      worktreeRequests.push(request.postDataJSON())
    }
    if (path === "/api/session") calls.push("session")
    if (/^\/api\/session\/[^/]+\/prompt$/.test(path)) calls.push("prompt")
  })
-  await page.route(`**/api/worktree/${projectID}`, async (route) => {
-    if (route.request().method() !== "POST") return route.fallback()
-    // Keep the real HTTP response pending until the test has checked the preview.
-    const response = await worktree.promise
-    if (response.status === 200) project.sandboxes.push(workspace)
-    await route.fulfill({ ...response, headers })
-  })
+  await page.route(
+    (url) => url.pathname === "/api/worktree",
+    async (route) => {
+      if (route.request().method() !== "POST") return route.fallback()
+      // Keep the real HTTP response pending until the test has checked the preview.
+      const response = await worktree.promise
+      if (response.status === 200) project.sandboxes.push(workspace)
+      await route.fulfill({ ...response, headers })
+    },
+  )
  await page.route("**/api/session", async (route) => {
    if (route.request().method() !== "POST") return route.fallback()
    const body: Record<string, unknown> = route.request().postDataJSON()
@@ -404,7 +516,10 @@ async function openDraft(page: Page, options?: { failSessionCreate?: boolean })
      return route.fulfill({ status: 500, json: { message: "Session creation failed in the fixture" }, headers })
    }
    if (typeof body.id !== "string") throw new Error("Session creation must use the client-reserved ID")
-    const session = currentSession({ ...body, id: body.id, projectID, title: "Created workspace session" }, workspace)
+    const session = currentSession(
+      { ...body, id: body.id, projectID, title: options?.untitled ? "" : "Created workspace session" },
+      workspace,
+    )
    sessions.push(session)
    return route.fulfill({ json: { data: session }, headers })
  })
@@ -436,6 +551,17 @@ async function openDraft(page: Page, options?: { failSessionCreate?: boolean })
      headers,
    }),
  )
+  if (options?.command) {
+    await page.route("**/api/command?**", (route) =>
+      route.fulfill({
+        json: {
+          location: { directory: new URL(route.request().url()).searchParams.get("location[directory]") ?? directory },
+          data: [{ name: "review", description: "Review changes" }],
+        },
+        headers,
+      }),
+    )
+  }
  await page.addInitScript(
    ({ directory, draftID, otherID, server }) => {
      localStorage.setItem(
@@ -461,11 +587,11 @@ async function openDraft(page: Page, options?: { failSessionCreate?: boolean })
  await page.getByRole("menuitem", { name: "New worktree", exact: true }).click()
  await expect(page.getByRole("button", { name: "New worktree", exact: true })).toBeVisible()
  await expect(page.locator('[data-component="composer-editor"]')).toBeEditable()
-  return { worktree, worktreeRequests, calls, creates, prompts }
+  return { worktree, worktreeRequests, calls, creates, prompts, events }
 }

-async function submitPending(page: Page, mock: Awaited<ReturnType<typeof openDraft>>) {
-  await page.locator('[data-component="composer-editor"]').fill(text)
+async function submitPending(page: Page, mock: Awaited<ReturnType<typeof openDraft>>, prompt = text) {
+  await page.locator('[data-component="composer-editor"]').fill(prompt)
  await expect(page.locator('[data-action="composer-submit"]')).toBeEnabled()
  await page.locator('[data-action="composer-submit"]').click()
  await expect(page).toHaveURL((url) => url.pathname.startsWith(sessionPath) && /\/ses_[^/]+$/.test(url.pathname))
@@ -476,12 +602,12 @@ async function submitPending(page: Page, mock: Awaited<ReturnType<typeof openDra
  const shimmer = preparing.getByRole("status").locator('[data-component="text-shimmer"]')
  const title = preparing.getByRole("heading", { level: 1 })
  await expect(preparing).toBeVisible()
-  await expect(title).toHaveText("New session")
+  await expect(title).toHaveText("Session")
  await expect(page.locator('[data-component="composer-editor"]')).toBeEditable()
  await expect(page.locator('[data-action="composer-submit"]')).toBeDisabled()
  await expect(preparing.locator('[data-component="user-message"]')).toHaveCount(1)
  await expect(message).toHaveCount(1)
-  await expect(message.locator('[data-slot="user-message-text"]')).toHaveText(text)
+  await expect(message.locator('[data-slot="user-message-text"]')).toHaveText(prompt)
  await expect(message).toHaveAttribute("data-timeline-part-id", /^.+:text:0$/)
  const messageID = (await message.getAttribute("data-timeline-part-id"))!.replace(/:text:0$/, "")
  await expect(shimmer).toBeVisible()
--- a/Show More
+++ b/Show More
Author	SHA1	Message	Date
Brendan Allan	59e36f7120	feat(desktop): add SSH server connections	2026-09-07 15:30:39 +08:00
Aiden Cline	42bccc301e	refactor(ai): give named providers dedicated modules (#47726 )	2026-09-06 23:56:39 -05:00
David HillandLukeParkerDev	243e81041e	feat(app): refine timeline detail settings (#47226 ) Co-authored-by: LukeParkerDev <10430890+Hona@users.noreply.github.com>	2026-09-07 04:54:28 +00:00
Luke Parker	5d971e5272	perf(app): externalize large draft text into content-addressed chunks (#47706 )	2026-09-07 14:17:34 +10:00
Dax Raad	7ef9cb68d1	Revert "docs(go): document client session compatibility" This reverts commit `0ff9ccb227`.	2026-09-06 23:42:25 -04:00
opencode-agent[bot]	733f138dae	fix(app): remove bottom safe area while keyboard is open (#47712 )	2026-09-07 13:41:44 +10:00
Dax Raad	0ff9ccb227	docs(go): document client session compatibility	2026-09-06 23:35:26 -04:00
opencode-agent[bot]andHona	5ff8a3f5fa	fix(ui): align unread avatar badges with the corner (#47718 ) Co-authored-by: Hona <10430890+Hona@users.noreply.github.com>	2026-09-07 13:15:48 +10:00
Luke Parker	44513bd0b0	perf(app): serialize persisted stores on a schedule instead of per setter call (#47705 )	2026-09-07 13:01:52 +10:00
Luke Parker	dc46ecfc55	perf(app): cache storage namespaces and batch writes in the renderer (#47704 )	2026-09-07 12:33:09 +10:00
usrnk1andLukeParkerDev	f16acb1c95	feat(desktop): improve worktree ui (#47310 ) Co-authored-by: LukeParkerDev <10430890+Hona@users.noreply.github.com>	2026-09-07 01:26:11 +00:00
opencode-agent[bot]	d0d3504780	chore: update nix node_modules hashes	2026-09-07 01:05:55 +00:00
Luke Parker	5e3100a46a	fix(desktop): persist renderer state in SQLite instead of electron-store (#47695 )	2026-09-07 10:51:46 +10:00
Luke Parker	69b30acfd4	chore: stop trusting tree-sitter install scripts (#47696 )	2026-09-07 00:09:44 +00:00
Luke Parker	ecb3c88601	fix(app): give worktree creation a setup-length request deadline (#47694 )	2026-09-06 23:44:45 +00:00
opencode-agent[bot]andHona	a5ebf5bf2f	fix(www): regenerate OpenAPI documents (#47693 ) Co-authored-by: Hona <10430890+Hona@users.noreply.github.com>	2026-09-06 23:26:56 +00:00
Luke Parker	f334377c5b	fix(util): bound opencode.log by trimming its head in place (#47676 )	2026-09-07 09:16:09 +10:00
Luke Parker	873d437c12	fix(desktop): emit preload as .cjs so it loads under --no-sandbox (#47691 )	2026-09-07 09:14:38 +10:00
Aiden Cline	13286e883e	fix(ai): raise typed errors for failed finishes (#47688 )	2026-09-06 18:09:49 -05:00
Luke Parker	0d70595338	fix(desktop): export tail of oversized logs in debug bundle (#47672 )	2026-09-07 08:47:30 +10:00
Aiden Cline	a02415a3eb	fix(ai): restrict Bedrock cache checkpoints to Claude (#47660 )	2026-09-06 17:00:12 -05:00
Dax	beca5e8926	feat(project): allow updating canonical directory (#47642 )	2026-09-06 17:52:48 -04:00
Shoubhit Dash	e3c89654d1	feat(core): keep web search providers sticky per session (#47334 )	2026-09-07 03:09:47 +05:30
Dax	33f48f36c9	docs(www): add Console documentation (#47638 )	2026-09-06 12:06:07 -04:00
Dax	58f949d2d0	fix(core): log MCP OAuth and credential lifecycle (#47636 )	2026-09-06 11:56:42 -04:00
Kit Langton	2a895b9e03	fix(core): interrupt sessions before inactivity eviction (#47629 )	2026-09-06 11:11:42 -04:00
Dax Raad	fe506f201d	docs: explain supporting v1 and v2 plugins	2026-09-06 11:08:26 -04:00
Luke Parker	b2cecc6350	fix(desktop): add sidecar credentials from the main process (#47588 )	2026-09-06 07:18:12 +00:00
Luke Parker	63a1074c6c	fix(app): keep slow git reads from filling the request queue (#47564 )	2026-09-06 17:02:03 +10:00
Luke Parker	31ee07e3ae	fix(app): pace directory re-sync after reconnect (#47565 )	2026-09-06 17:01:24 +10:00
Aarav Sareen	370b9965d3	feat(app): cmd+F to search session with highlighting	2026-09-06 16:47:20 +10:00
Dax Raad	8ea99ef9ad	fix(tui): dismiss update notification only on skip	2026-09-06 02:43:21 -04:00
Aarav Sareen	e64b2bc137	feat(app): add horizontal file sidebar scrolling + right click menu	2026-09-06 16:35:14 +10:00
opencode-agent[bot]andrekram1-node	cf1923c238	fix(core): disable default GPT and Claude search filtering (#47586 ) Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>	2026-09-06 01:22:48 -05:00
Luke Parker	bfcb388dd7	feat(app): pulse the status dot while the event stream reconnects (#47574 )	2026-09-06 15:13:38 +10:00
usrnk1	ec46439ef5	feat(desktop): simplify move to background action	2026-09-06 14:45:16 +10:00
Luke Parker	1be3b32a47	fix(client): detect stalled event streams and resync on foreground (#47571 )	2026-09-06 04:43:50 +00:00
Aiden Cline	2823b886d7	feat(core): add independent GPT and Claude tool optimization (#47559 )	2026-09-05 23:43:23 -05:00
Luke Parker	99651b2d50	fix(app): refresh queued inputs when the connection returns (#47573 )	2026-09-06 04:42:27 +00:00
usrnk1	33ef66746b	feat(desktop): add third-party web search consent	2026-09-06 14:38:55 +10:00
Luke Parker	cf212a4235	fix(app): time out requests the server never answers (#47572 )	2026-09-06 14:28:43 +10:00
David Hill	768e3d45a2	fix(ui): lighten dark contrast icon buttons	2026-09-06 14:01:47 +10:00
Luke Parker	0674ddf043	fix(client): coalesce catalog refetches from event bursts (#47561 )	2026-09-06 03:48:49 +00:00
David Hill	41b66bc560	fix(app): scroll overflowing composer controls	2026-09-06 13:47:42 +10:00
opencode-agent[bot]	49a5baa059	chore: update nix node_modules hashes	2026-09-06 03:39:31 +00:00
Luke Parker	1cf576ab55	fix(desktop): keep server CORS headers so preflights cache (#47560 )	2026-09-06 03:26:12 +00:00
David Hill	f7d0f69e6c	feat(app): add subtle new-session logo shimmer	2026-09-06 13:22:15 +10:00
Aiden Cline	a66e6cd55e	feat(core): discover Bedrock credentials in the provider plugin (#47548 )	2026-09-05 22:20:48 -05:00
David Hill	93b3725b14	fix(app): move project icon setting to experimental	2026-09-06 13:07:37 +10:00
David Hill	371e344ba6	fix(app): widen new-session prompt beyond logo	2026-09-06 13:07:33 +10:00
David Hill	6559dcb01b	fix(tui): list OpenCode Go before Zen	2026-09-06 13:07:29 +10:00
David Hill	35dbdfe287	fix(app): simplify vertical tabs footer	2026-09-06 13:07:24 +10:00
opencode-agent[bot]	cd504dc66a	chore: update nix node_modules hashes	2026-09-06 00:05:05 +00:00
Aiden Cline	c4f2926ffb	feat(ai): resolve Bedrock credentials through the AWS default chain (#47436 )	2026-09-05 18:49:05 -05:00
opencode-agent[bot]andrekram1-node	aa48433fa0	fix(core): send session ID with Copilot requests (#47213 ) Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>	2026-09-05 18:42:46 -05:00
Kit Langton	8da7526616	fix(core): reload local plugin helpers without restarting Reuse the shared dependency-aware plugin loader for Core hot reloads. Scope watcher setup callbacks and subscriptions, recover missing external helpers, and preserve explicit local entrypoint invalidation without reloading package dependencies. Add regression coverage and portable RPC fixtures.	2026-09-05 17:00:41 -04:00
opencode-agent[bot]andrekram1-node	7a4ad68af6	fix(mcp): retry initialization without codemode on HTTP 400 (#47507 ) Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>	2026-09-05 15:49:25 +00:00
opencode-agent[bot]andnexxeln	2960c61f9c	fix(tui): preserve parent scroll position across subagents (#47464 ) Co-authored-by: nexxeln <95541290+nexxeln@users.noreply.github.com>	2026-09-05 15:19:16 +05:30
opencode-agent[bot]andnexxeln	7ad705225e	fix(tui): avoid disconnection error on terminal exit (#47272 ) Co-authored-by: nexxeln <95541290+nexxeln@users.noreply.github.com>	2026-09-05 14:35:40 +05:30
opencode-agent[bot]andHona	5045b3ce18	fix(ui): keep file accordion borders visible when scaled (#47461 ) Co-authored-by: Hona <10430890+Hona@users.noreply.github.com>	2026-09-05 08:47:54 +00:00
Aiden Cline	4306c07b34	feat(core): update GPT prompts and remove legacy Anthropic prompt (#47447 )	2026-09-05 01:39:18 -05:00
opencode-agent[bot]	7de1e86b5d	chore: update nix node_modules hashes	2026-09-05 06:38:20 +00:00
opencode-agent[bot]andrekram1-node	a6fd1f8b22	chore: bump Bun to 1.4.2 (#47446 ) Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>	2026-09-05 01:21:42 -05:00
Dax Raad	52685d4517	fix(tui): simplify home update notices	2026-09-05 02:17:09 -04:00
Luke Parker	30d1049942	fix(app): load worktree inventory on demand and cap concurrent server requests (#47441 )	2026-09-05 05:25:43 +00:00
Stefan Avramandslickstef11	21adcb4969	fix(console): route enterprise forms to Chatwoot (#47437 ) Co-authored-by: slickstef11 <stefanavram22@gmail.com>	2026-09-05 00:05:40 -04:00
Kit Langton	c72b535dee	fix(tui): reload local plugin dependency graphs (#47388 ) Reload statically discovered local plugin dependencies through native module caches while preserving shared package identities and best-effort last-good registrations.	2026-09-04 23:41:02 -04:00
opencode-agent[bot]andjlongster	8d1a9799f4	fix(tui): remove completion notice links (#47426 ) Co-authored-by: jlongster <17031+jlongster@users.noreply.github.com>	2026-09-04 22:46:28 -04:00
David Hill	86ba09c6e0	fix(app): increase vertical tabs minimum width (#47424 )	2026-09-05 12:10:12 +10:00
opencode-agent[bot]andHona	41cb354c3e	fix(app): preserve vertical navigation label and icon widths (#47418 ) Co-authored-by: Hona <10430890+Hona@users.noreply.github.com>	2026-09-05 11:30:12 +10:00
James Long	23f3f8b6ca	fix(tui): reference update command (#47415 )	2026-09-04 20:43:44 -04:00
Aiden Cline	baab05727d	fix(core): compare Codex GPT versions by major and minor (#47404 )	2026-09-04 19:22:12 -05:00
Aiden Cline	541937d124	fix(core): clarify how to call search in Code Mode guidance (#47360 )	2026-09-04 18:50:47 -05:00
David Hill	0da55f1cef	fix(app): list OpenCode Go before Zen (#47386 )	2026-09-05 09:38:59 +10:00
opencode-agent[bot]	ac76382211	chore: update nix node_modules hashes	2026-09-04 23:23:36 +00:00
James Long	c30285c148	feat(worktree): support configurable plugin strategies (#47358 )	2026-09-04 19:07:40 -04:00
opencode-agent[bot]	b274224af1	fix(app): keep pending worktree composer styling consistent (#47377 )	2026-09-04 22:59:21 +00:00
opencode-agent[bot]	7a050a19a1	fix(app): apply worktree message color during preparation (#47379 )	2026-09-04 22:53:43 +00:00
Luke Parker	16601775f1	fix(app): align desktop worktree location with TUI (#47370 )	2026-09-04 22:12:15 +00:00
James Long	0991e8b5a5	fix(tui): clarify update notifications (#47372 )	2026-09-04 18:03:12 -04:00
James Long	6d791dfe67	feat(tui): add plugin-owned session panels (#47150 )	2026-09-04 17:37:45 -04:00
Dax	f268956c75	fix(tui): preview always-allow patterns without extra confirmation (#47369 )	2026-09-04 21:34:58 +00:00
James Long	3290a39667	feat(tui): add manual update dialog (#47364 )	2026-09-04 17:11:35 -04:00
James Long	51d69b26a0	refactor(tui): support reactive theme contexts (#47148 )	2026-09-04 16:42:07 -04:00
James Long	8ff1ef1a62	fix(tui): anchor diff file menus to the screen (#47147 )	2026-09-04 16:23:16 -04:00
James Long	90cd910f52	feat(tui): add shared interactivity context (#47144 )	2026-09-04 16:07:11 -04:00
Dax Raad	c3342ca812	chore: empty commit	2026-09-04 14:23:12 -04:00
Kit Langton	7ca047b2b9	fix(core): prioritize compaction over pending steers (#47340 )	2026-09-04 13:23:55 -04:00
opencode-agent[bot]andnexxeln	b78c2ea7b4	fix(tui): undo pending prompts and active sessions (#47343 ) Co-authored-by: nexxeln <95541290+nexxeln@users.noreply.github.com>	2026-09-04 22:52:03 +05:30
James Long	0e143437c8	feat(cli): restore automatic update policy (#47161 )	2026-09-04 12:58:30 -04:00
opencode-agent[bot]andrekram1-node	bff58fc387	feat(tui): show retry countdown (#47145 ) Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>	2026-09-04 11:24:04 -05:00
Shoubhit Dash	a26a978051	fix(ai): tolerate out-of-band responses notifications (#47316 )	2026-09-04 21:36:29 +05:30
opencode-agent[bot]andnexxeln	218a0dde97	fix(tui): clear stale error styling on resumed subagents (#47295 ) Co-authored-by: nexxeln <95541290+nexxeln@users.noreply.github.com>	2026-09-04 21:29:27 +05:30
usrnk1	bd54dc508f	feat(desktop): add dropzone to new sessions (#47303 )	2026-09-04 17:19:05 +02:00
Shoubhit Dash	0b3c45c948	feat(ai): add streamed compaction triggers (#47082 )	2026-09-04 14:00:02 +00:00
Shoubhit Dash	689a354a43	fix(ai): preserve websocket replay baselines (#47291 )	2026-09-04 13:46:00 +00:00
Shoubhit Dash	acb462a173	fix(tui): restore agent and model selection parity (#47260 )	2026-09-04 18:16:04 +05:30
opencode-agent[bot]	e26cbb935e	chore: update nix node_modules hashes	2026-09-04 12:21:18 +00:00
Luke Parker	fa6dd9df8f	chore: upgrade Bun to 1.4.1 and re-enable bytecode (#47271 )	2026-09-04 22:01:12 +10:00
opencode-agent[bot]andBrendonovich	b52b79d28f	fix(app): align file tool disclosures with Patch (#47269 ) Co-authored-by: Brendonovich <14191578+Brendonovich@users.noreply.github.com>	2026-09-04 19:47:57 +08:00
Aarav Sareen	632b65854c	fix(app): adjust sidebar and settings spacing (#47263 )	2026-09-04 09:50:30 +00:00
opencode-agent[bot]andBrendonovich	ed877cfebc	fix(app): expose skills in slash commands (#47247 ) Co-authored-by: Brendonovich <14191578+Brendonovich@users.noreply.github.com>	2026-09-04 17:40:39 +08:00
opencode-agent[bot]andBrendonovich	3a08d95907	fix(app): preserve thinking trigger visibility invariant (#47256 ) Co-authored-by: Brendonovich <14191578+Brendonovich@users.noreply.github.com>	2026-09-04 09:34:18 +00:00
opencode-agent[bot]Brendonovich Brendan Allan	b09a74591c	fix(app): omit notices from Used summary (#47245 ) Co-authored-by: Brendonovich <14191578+Brendonovich@users.noreply.github.com> Co-authored-by: Brendan Allan <git@brendonovich.dev>	2026-09-04 09:13:00 +00:00
Brendan Allan	e8481973ce	refactor(app): make session composer ownership explicit (#47254 )	2026-09-04 09:07:47 +00:00
usrnk1	211cd73f1a	feat(desktop): update composer dropzone (#42312 )	2026-09-04 16:45:42 +08:00
opencode-agent[bot]andBrendonovich	2375e81bd6	fix(app): match optimistic session heading to tab label (#47244 ) Co-authored-by: Brendonovich <14191578+Brendonovich@users.noreply.github.com>	2026-09-04 08:07:24 +00:00
opencode-agent[bot]MrMushrooooom Jack	8352addf6c	fix(console): preserve usage reset boundaries (#44729 ) Co-authored-by: MrMushrooooom <MrMushrooooom@users.noreply.github.com> Co-authored-by: Jack <jack@anoma.ly>	2026-09-04 15:12:54 +08:00
Aiden Cline	32f89748af	feat(ai): add tool namespaces (#46548 )	2026-09-04 01:09:52 -05:00
Aiden Cline	89478b36f1	refactor(plugin): rename session request kind "session" to "primary" (#47221 )	2026-09-04 00:12:18 -05:00
Aiden Cline	46458f0753	feat(core): tag session http hooks with request kind (#47214 )	2026-09-03 23:51:26 -05:00
opencode-agent[bot]andrekram1-node	c907d2ba27	fix(tui): preserve model release ordering in search (#47183 ) Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>	2026-09-03 23:18:09 -05:00
Aiden Cline	ffac1c5b11	fix(core): classify GitHub Copilot requests on every route (#47160 )	2026-09-03 22:13:04 -05:00
opencode-agent[bot]andrekram1-node	c5dca2df37	feat(core): suggest moving sessions into task worktrees (#47202 ) Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>	2026-09-03 22:06:15 -05:00
Aiden Cline	8889447f5a	fix(core): scope compaction Completed section to finished work (#47203 )	2026-09-03 22:05:25 -05:00
opencode-agent[bot]andHona	a1cb005799	fix(app): remove timeline settings advanced divider (#47196 ) Co-authored-by: Hona <10430890+Hona@users.noreply.github.com>	2026-09-04 12:25:33 +10:00
Kit Langton	19833ad1fd	feat(simulation): expose and record real mouse input (#47194 )	2026-09-04 02:17:33 +00:00
Kit Langton	cb852434b1	fix(tui): preserve selection under a stationary pointer (#47138 )	2026-09-04 02:04:29 +00:00
opencode-agent[bot]andiamdavidhill	b3733e9517	fix(app): add bottom padding to settings pages (#47191 ) Co-authored-by: iamdavidhill <1879069+iamdavidhill@users.noreply.github.com>	2026-09-04 01:58:55 +00:00
opencode-agent[bot]andHona	331f4ecd2f	fix(app): contain preferences scroll overflow (#47190 ) Co-authored-by: Hona <10430890+Hona@users.noreply.github.com>	2026-09-04 01:54:52 +00:00
Luke Parker	ecaa914b79	fix(app): align Windows app icon in title bar (#47189 )	2026-09-04 01:54:22 +00:00
Aiden Cline	4f8dea674a	refactor(ai): drop converse tool deltas without an open block (#47182 )	2026-09-03 20:50:35 -05:00
Aiden Cline	4bf5269c4c	fix(core): retry transient compaction failures (#47159 )	2026-09-03 20:45:47 -05:00
Kit Langton	c9d240704d	fix(core): settle abandoned compactions before resuming sessions (#47178 )	2026-09-04 01:02:29 +00:00
Kit Langton	c9df4ba80d	fix(util): skip unused audits during package installs (#47176 )	2026-09-04 00:58:29 +00:00
David HillandLukeParkerDev	ac71a55294	fix(app): keep right panel controls aligned (#46996 ) Co-authored-by: LukeParkerDev <10430890+Hona@users.noreply.github.com>	2026-09-04 00:54:39 +00:00
Aiden Cline	f84d927e07	refactor(ai): drop reasoning items when they finish (#47100 )	2026-09-03 19:53:22 -05:00
opencode-agent[bot]Hona rekram1-node	d431fedce5	test(core): disable npm audits in the test preload (#47170 ) Co-authored-by: Hona <10430890+Hona@users.noreply.github.com> Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>	2026-09-03 19:52:18 -05:00
Kit Langton	e536b9627e	feat(tui): inspect live shell output (#47134 )	2026-09-04 00:47:58 +00:00
David HillandLukeParkerDev	6af46cc8a9	feat(app): add settings to vertical tabs (#47119 ) Co-authored-by: LukeParkerDev <10430890+Hona@users.noreply.github.com>	2026-09-04 10:45:09 +10:00
David HillandLukeParkerDev	c76e602ba4	fix(app): keep session tab labels stable during creation (#47099 ) Co-authored-by: LukeParkerDev <10430890+Hona@users.noreply.github.com>	2026-09-04 10:33:51 +10:00
opencode-agent[bot]andrekram1-node	726107729e	test(core): align Code Mode catalog scope assertions (#47169 ) Co-authored-by: rekram1-node <rekram1-node@users.noreply.github.com>	2026-09-03 19:30:25 -05:00
David Hill	0808ebc3c5	fix(app): refine build indicators in tab navigation (#47112 )	2026-09-04 09:57:32 +10:00
David HillandLukeParkerDev	17e5c5fbf1	feat(app): show vertical sidebar shortcut hints (#47122 ) Co-authored-by: LukeParkerDev <10430890+Hona@users.noreply.github.com>	2026-09-04 09:42:29 +10:00
David Hill	b0a13b810e	fix(app): allow new session shortcut from settings (#47123 )	2026-09-04 09:42:14 +10:00
Luke Parker	6575215ddf	fix(desktop): keep command palette responsive and scoped (#47164 )	2026-09-04 09:41:12 +10:00
David HillandLukeParkerDev	09157609b7	feat(app): add session project menu and polish header (#46977 ) Co-authored-by: LukeParkerDev <10430890+Hona@users.noreply.github.com>	2026-09-04 09:25:31 +10:00
Luke Parker	7ba5f3e5b2	fix(core): restore Ctrl+C in Windows terminals (#47163 )	2026-09-04 09:16:14 +10:00
David HillandLukeParkerDev	0ec1587948	feat(app): add about settings page (#45782 ) Co-authored-by: LukeParkerDev <10430890+Hona@users.noreply.github.com>	2026-09-03 23:07:41 +00:00
David Hill	2fda2c8141	feat(app): give experimental settings a dedicated page (#46973 )	2026-09-04 08:33:29 +10:00
Luke Parker	f06bf47e72	fix(app): preserve slash commands when creating worktrees (#47155 )	2026-09-03 22:25:42 +00:00
opencode-agent[bot]andHona	fa6e7e179d	fix(desktop): render before telemetry (#47102 ) Co-authored-by: Hona <10430890+Hona@users.noreply.github.com>	2026-09-04 08:21:16 +10:00
Dax Raad	4272bb83e0	feat(cli): support command aliases with update alias for upgrade	2026-09-03 18:10:01 -04:00
usrnk1andiamdavidhill	1ae2cf3567	feat(desktop): polish menus, file tabs, and empty states (#47095 ) Co-authored-by: iamdavidhill <1879069+iamdavidhill@users.noreply.github.com>	2026-09-04 07:55:27 +10:00
Dax Raad	2084c52952	fix(core): clarify code mode tool scope	2026-09-03 17:21:06 -04:00
Aiden Cline	5a9931280e	feat(core): add Modal live model backfill plugin (#47136 )	2026-09-03 16:03:21 -05:00
Aiden Cline	b2fb2c5e36	chore(core): bump GitHub Copilot API version to 2026-08-01 (#47108 )	2026-09-03 15:38:32 -05:00