fix(tui): guard streaming part appends with a part index

Every session.text.delta calls appendPart, which ran setRows(produce(...)) plus a hasPart full-tree scan even when the part already existed. On a root-array Solid store the no-op update still unwraps the whole rows tree, so per-delta cost scaled with loaded history (~580ms per 50-delta burst at 4000 messages). Maintain a lazily-rebuilt Set of part keys: rebuilt from the plain reduce() output on each reconcile, extended on append, and checked O(1) before touching the store. Reasoning completion still flows through to flip the group flag. Measured with a headless createTestRenderer probe (mixed markdown + tools, 50 deltas + settle): 578ms -> 34ms at 4000 messages, 133ms -> 34ms at 1000 messages.
feat(tui): configure session permission handling (#48545 )
2026-09-12 11:56:23 +00:00 · 2026-09-11 16:26:20 -05:00 · 2026-09-11 19:50:39 +00:00 · 2026-09-11 15:50:35 -04:00 · 2026-09-11 14:50:07 -05:00 · 2026-09-11 15:34:33 -04:00
331 changed files with 10174 additions and 6230 deletions
@@ -25,7 +25,7 @@ on:
        required: false
        type: string

-concurrency: ${{ github.workflow }}-${{ github.ref }}-${{ inputs.version || inputs.bump }}
+concurrency: ${{ github.workflow }}-${{ github.ref }}-${{ (github.ref_name == 'v2' && (inputs.version || inputs.bump) && 'release') || inputs.version || inputs.bump }}

 permissions:
  id-token: write
@@ -33,7 +33,7 @@ permissions:
  packages: write

 env:
-  OPENCODE_CHANNEL: ${{ (github.ref_name == 'v2' && 'dev') || '' }}
+  OPENCODE_CHANNEL: ${{ (github.ref_name == 'v2' && !inputs.bump && !inputs.version && 'dev') || '' }}

 jobs:
  version:
@@ -168,7 +168,7 @@ jobs:
          fi

          found=0
-          for file in packages/cli/dist/cli-darwin-*/bin/opencode2; do
+          for file in packages/cli/dist/cli-darwin-*/bin/opencode; do
            if [ ! -f "$file" ]; then
              continue
            fi
@@ -378,7 +378,7 @@ jobs:
    needs:
      - version
      - sign-cli-macos
-    if: github.repository == 'anomalyco/opencode' && github.ref_name != 'v2'
+    if: github.repository == 'anomalyco/opencode' && (github.ref_name != 'v2' || needs.version.outputs.release != '')
    continue-on-error: false
    env:
      AZURE_CLIENT_ID: ${{ secrets.AZURE_CLIENT_ID }}
@@ -8,9 +8,9 @@

 ## Live V2 TUI Testing

- Run `bun run dev:live` from a development worktree to test its TUI against the currently elected `opencode2` background server and live sessions.
+- Run `bun run dev:live` from a development worktree to test its TUI against the currently elected `opencode` background server and live sessions.
 - Pass a directory after the script when needed, for example `bun run dev:live /path/to/project`.
- The script discovers the server with `opencode2 service status`, injects its private local credential from `opencode2 service get password`, and uses the `dev` TUI storage channel so tabs and other client-local state match the installed client.
+- The script discovers the server with `opencode service status`, injects its private local credential from `opencode service get password`, and uses the `dev` TUI storage channel so tabs and other client-local state match the installed client.
 - Prefer `dev:live` over plain `bun run dev` for this workflow. An implicit managed-service connection may replace the live server when the worktree client version differs; explicit `--server` warns and continues without replacing it.

 ## V2 TUI Stories
@@ -31,7 +31,7 @@
    },
    "packages/ai": {
      "name": "@opencode/ai",
-      "version": "1.17.20",
+      "version": "2.0.0",
      "dependencies": {
        "@aws-sdk/credential-providers": "3.1057.0",
        "@opencode/schema": "workspace:*",
@@ -53,7 +53,7 @@
    },
    "packages/app": {
      "name": "@opencode/app",
-      "version": "1.18.15",
+      "version": "2.0.0",
      "dependencies": {
        "@corvu/drawer": "catalog:",
        "@dnd-kit/abstract": "0.5.0",
@@ -111,7 +111,7 @@
    },
    "packages/cli": {
      "name": "@opencode/cli",
-      "version": "1.18.4",
+      "version": "2.0.0",
      "bin": {
        "opencode2": "./bin/opencode2.cjs",
      },
@@ -175,7 +175,7 @@
    },
    "packages/client": {
      "name": "@opencode/client",
-      "version": "1.17.13",
+      "version": "2.0.0",
      "dependencies": {
        "@opencode/protocol": "workspace:*",
        "@opencode/schema": "workspace:*",
@@ -201,7 +201,7 @@
    },
    "packages/codemode": {
      "name": "@opencode/codemode",
-      "version": "1.18.4",
+      "version": "2.0.0",
      "dependencies": {
        "acorn": "8.15.0",
        "effect": "catalog:",
@@ -215,7 +215,7 @@
    },
    "packages/console/app": {
      "name": "@opencode/console-app",
-      "version": "1.18.15",
+      "version": "2.0.0",
      "dependencies": {
        "@cloudflare/vite-plugin": "1.15.2",
        "@ibm/plex": "6.4.1",
@@ -251,7 +251,7 @@
    },
    "packages/console/core": {
      "name": "@opencode/console-core",
-      "version": "1.18.15",
+      "version": "2.0.0",
      "dependencies": {
        "@aws-sdk/client-sts": "3.782.0",
        "@jsx-email/render": "1.1.1",
@@ -278,7 +278,7 @@
    },
    "packages/console/function": {
      "name": "@opencode/console-function",
-      "version": "1.18.15",
+      "version": "2.0.0",
      "dependencies": {
        "@openauthjs/openauth": "0.0.0-20250322224806",
        "@opencode/console-core": "workspace:*",
@@ -295,7 +295,7 @@
    },
    "packages/console/mail": {
      "name": "@opencode/console-mail",
-      "version": "1.18.15",
+      "version": "2.0.0",
      "dependencies": {
        "@jsx-email/all": "2.2.3",
        "@jsx-email/cli": "1.4.3",
@@ -319,7 +319,7 @@
    },
    "packages/console/support": {
      "name": "@opencode/console-support",
-      "version": "1.18.15",
+      "version": "2.0.0",
      "dependencies": {
        "@cloudflare/vite-plugin": "1.15.2",
        "@opencode/console-core": "workspace:*",
@@ -339,7 +339,7 @@
    },
    "packages/core": {
      "name": "@opencode/core",
-      "version": "1.18.4",
+      "version": "2.0.0",
      "dependencies": {
        "@ai-sdk/alibaba": "1.0.17",
        "@ai-sdk/anthropic": "3.0.82",
@@ -411,7 +411,7 @@
    },
    "packages/desktop": {
      "name": "@opencode/desktop",
-      "version": "1.18.15",
+      "version": "2.0.0",
      "dependencies": {
        "@zip.js/zip.js": "2.7.62",
        "electron-context-menu": "4.1.2",
@@ -463,7 +463,7 @@
    },
    "packages/enterprise": {
      "name": "@opencode/enterprise",
-      "version": "1.18.15",
+      "version": "2.0.0",
      "dependencies": {
        "@hono/standard-validator": "catalog:",
        "@opencode-ai/sdk": "1.18.21",
@@ -500,7 +500,7 @@
    },
    "packages/function": {
      "name": "@opencode/function",
-      "version": "1.18.15",
+      "version": "2.0.0",
      "dependencies": {
        "@octokit/auth-app": "8.0.1",
        "@octokit/rest": "catalog:",
@@ -516,7 +516,7 @@
    },
    "packages/http-recorder": {
      "name": "@opencode/http-recorder",
-      "version": "1.18.15",
+      "version": "2.0.0",
      "dependencies": {
        "@effect/platform-node-shared": "4.0.0-rc.112",
      },
@@ -535,7 +535,7 @@
    },
    "packages/httpapi-codegen": {
      "name": "@opencode/httpapi-codegen",
-      "version": "0.0.0",
+      "version": "2.0.0",
      "dependencies": {
        "effect": "catalog:",
        "prettier": "3.6.2",
@@ -548,7 +548,7 @@
    },
    "packages/latex": {
      "name": "@opencode/latex",
-      "version": "0.0.0",
+      "version": "2.0.0",
      "dependencies": {
        "@opencode/plugin": "workspace:*",
        "@opentui/core": "catalog:",
@@ -562,7 +562,7 @@
    },
    "packages/merman": {
      "name": "@opencode/merman",
-      "version": "0.0.0",
+      "version": "2.0.0",
      "dependencies": {
        "@opencode/plugin": "workspace:*",
        "@opentui/core": "catalog:",
@@ -577,7 +577,7 @@
    },
    "packages/plugin": {
      "name": "@opencode/plugin",
-      "version": "1.18.15",
+      "version": "2.0.0",
      "dependencies": {
        "@ai-sdk/provider": "3.0.8",
        "@opencode/ai": "workspace:*",
@@ -616,7 +616,7 @@
    },
    "packages/plugin-browser": {
      "name": "@opencode/plugin-browser",
-      "version": "0.0.0",
+      "version": "2.0.0",
      "dependencies": {
        "@opencode/plugin": "workspace:*",
        "@opencode/schema": "workspace:*",
@@ -646,7 +646,7 @@
    },
    "packages/protocol": {
      "name": "@opencode/protocol",
-      "version": "1.17.11",
+      "version": "2.0.0",
      "dependencies": {
        "@opencode/schema": "workspace:*",
        "effect": "catalog:",
@@ -661,7 +661,7 @@
    },
    "packages/schema": {
      "name": "@opencode/schema",
-      "version": "1.17.11",
+      "version": "2.0.0",
      "dependencies": {
        "@standard-schema/spec": "catalog:",
        "effect": "catalog:",
@@ -685,7 +685,7 @@
    },
    "packages/sdk": {
      "name": "@opencode/sdk",
-      "version": "1.18.4",
+      "version": "2.0.0",
      "dependencies": {
        "@opencode/client": "workspace:*",
        "@opencode/core": "workspace:*",
@@ -706,7 +706,7 @@
    },
    "packages/server": {
      "name": "@opencode/server",
-      "version": "1.18.4",
+      "version": "2.0.0",
      "dependencies": {
        "@effect/platform-node": "catalog:",
        "@effect/platform-node-shared": "catalog:",
@@ -728,7 +728,7 @@
    },
    "packages/session-ui": {
      "name": "@opencode/session-ui",
-      "version": "1.18.15",
+      "version": "2.0.0",
      "dependencies": {
        "@kobalte/core": "catalog:",
        "@opencode/client": "workspace:*",
@@ -763,7 +763,7 @@
    },
    "packages/simulation": {
      "name": "@opencode/simulation",
-      "version": "1.17.13",
+      "version": "2.0.0",
      "dependencies": {
        "@opencode/ai": "workspace:*",
        "@opencode/core": "workspace:*",
@@ -783,7 +783,7 @@
    },
    "packages/stats/app": {
      "name": "@opencode/stats-app",
-      "version": "1.18.15",
+      "version": "2.0.0",
      "dependencies": {
        "@ibm/plex": "6.4.1",
        "@kobalte/core": "catalog:",
@@ -817,7 +817,7 @@
    },
    "packages/stats/core": {
      "name": "@opencode/stats-core",
-      "version": "1.18.15",
+      "version": "2.0.0",
      "dependencies": {
        "@aws-sdk/client-athena": "3.933.0",
        "@planetscale/database": "1.19.0",
@@ -836,7 +836,7 @@
    },
    "packages/stats/server": {
      "name": "@opencode/stats-server",
-      "version": "1.18.15",
+      "version": "2.0.0",
      "dependencies": {
        "@aws-sdk/client-firehose": "3.933.0",
        "@effect/platform-node": "catalog:",
@@ -882,7 +882,7 @@
    },
    "packages/theme": {
      "name": "@opencode/theme",
-      "version": "0.0.0",
+      "version": "2.0.0",
      "dependencies": {
        "@opentui/core": "catalog:",
        "effect": "catalog:",
@@ -896,7 +896,7 @@
    },
    "packages/tui": {
      "name": "@opencode/tui",
-      "version": "1.18.4",
+      "version": "2.0.0",
      "dependencies": {
        "@opencode/client": "workspace:*",
        "@opencode/core": "workspace:*",
@@ -931,7 +931,7 @@
    },
    "packages/ui": {
      "name": "@opencode/ui",
-      "version": "1.18.15",
+      "version": "2.0.0",
      "dependencies": {
        "@kobalte/core": "catalog:",
        "@pierre/diffs": "catalog:",
@@ -966,7 +966,7 @@
    },
    "packages/util": {
      "name": "@opencode/util",
-      "version": "1.18.3",
+      "version": "2.0.0",
      "dependencies": {
        "@effect/opentelemetry": "catalog:",
        "@effect/platform-node": "catalog:",
@@ -999,7 +999,7 @@
    },
    "packages/web": {
      "name": "@opencode/web",
-      "version": "1.18.15",
+      "version": "2.0.0",
      "dependencies": {
        "@astrojs/cloudflare": "12.6.3",
        "@astrojs/markdown-remark": "6.3.1",
@@ -1040,7 +1040,7 @@
    },
    "services/update": {
      "name": "@opencode/update",
-      "version": "1.18.4",
+      "version": "2.0.0",
      "dependencies": {
        "jose": "6.0.11",
        "semver": "catalog:",
@@ -1,6 +1,7 @@
 #!/usr/bin/env bash
 set -euo pipefail
-APP=opencode2
+APP=opencode
+SOURCE_APP=opencode

 MUTED='\033[0;2m'
 RED='\033[0;31m'
@@ -22,7 +23,7 @@ Options:
 Examples:
    curl -fsSL https://opencode.ai/v2/install | bash
    curl -fsSL https://opencode.ai/v2/install | bash -s -- --version 0.0.0-beta-17236
-    ./install --binary /path/to/opencode2
+    ./install --binary /path/to/opencode
 EOF
 }

@@ -201,9 +202,9 @@ else

    filename="cli-$target-$specific_version.tgz"
    url="https://registry.npmjs.org/$package_name/-/$filename"
-    binary_name="$APP"
+    binary_name="$SOURCE_APP"
    if [ "$os" = "windows" ]; then
-        binary_name="$APP.exe"
+        binary_name="$SOURCE_APP.exe"
    fi
 fi

@@ -230,12 +231,7 @@ check_version() {
        installed_version="${installed_version##* }"
        installed_version="${installed_version#v}"

-        if [[ "$installed_version" != "$specific_version" ]]; then
-            print_message info "${MUTED}Installed version: ${NC}$installed_version."
-        else
-            print_message info "${MUTED}Version ${NC}$specific_version${MUTED} already installed"
-            exit 0
-        fi
+        print_message info "${MUTED}Installed version: ${NC}$installed_version."
    fi
 }

@@ -340,15 +336,47 @@ download_and_install() {
    fi

    tar -xzf "$tmp_dir/$filename" -C "$tmp_dir"
-    mv "$tmp_dir/package/bin/$binary_name" "$INSTALL_DIR"
-    chmod 755 "${INSTALL_DIR}/$binary_name"
+    local installed_binary="$APP"
+    if [ "$os" = "windows" ]; then
+        installed_binary="$APP.exe"
+    fi
+    mv "$tmp_dir/package/bin/$binary_name" "$INSTALL_DIR/$installed_binary"
+    chmod 755 "$INSTALL_DIR/$installed_binary"
    rm -rf "$tmp_dir"
 }

 install_from_binary() {
    print_message info "\n${MUTED}Installing ${NC}$APP ${MUTED}from: ${NC}$binary_path"
-    cp "$binary_path" "${INSTALL_DIR}/$APP"
-    chmod 755 "${INSTALL_DIR}/$APP"
+    local installed_binary="$APP"
+    case "$(uname -s)" in
+        MINGW*|MSYS*|CYGWIN*) installed_binary="$APP.exe" ;;
+    esac
+    cp "$binary_path" "$INSTALL_DIR/$installed_binary"
+    chmod 755 "$INSTALL_DIR/$installed_binary"
+}
+
+install_legacy_shim() {
+    local shim_os="${os:-}"
+    if [[ -z "$shim_os" ]]; then
+        case "$(uname -s)" in
+            MINGW*|MSYS*|CYGWIN*) shim_os="windows" ;;
+        esac
+    fi
+    rm -f "$INSTALL_DIR/opencode2" "$INSTALL_DIR/opencode2.exe" "$INSTALL_DIR/opencode2.cmd"
+    if [[ "$shim_os" == "windows" ]]; then
+        cat > "$INSTALL_DIR/opencode2.cmd" <<'EOF'
+@echo off
+echo opencode2 is now just opencode. run opencode
+exit /b 1
+EOF
+        return
+    fi
+    cat > "$INSTALL_DIR/opencode2" <<'EOF'
+#!/bin/sh
+echo 'opencode2 is now just opencode. run opencode'
+exit 1
+EOF
+    chmod 755 "$INSTALL_DIR/opencode2"
 }

 if [ -n "$binary_path" ]; then
@@ -357,6 +385,7 @@ else
    check_version
    download_and_install
 fi
+install_legacy_shim


 add_to_path() {
@@ -453,7 +482,7 @@ echo -e ""
 echo -e "${MUTED}OpenCode includes free models, to start:${NC}"
 echo -e ""
 echo -e "cd <project>  ${MUTED}# Open directory${NC}"
-echo -e "opencode2     ${MUTED}# Run command${NC}"
+echo -e "opencode      ${MUTED}# Run command${NC}"
 echo -e ""
 echo -e "${MUTED}For more information visit ${NC}https://opencode.ai/v2/docs"
 echo -e ""
@@ -2,15 +2,15 @@
  "$schema": "https://json.schemastore.org/package.json",
  "name": "opencode",
  "description": "AI-powered development tool",
-  "version": "0.0.0",
+  "version": "2.0.0",
  "private": true,
  "type": "module",
  "packageManager": "bun@1.4.2",
  "scripts": {
    "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 \"$(opencode2 service status)\"' --",
+    "dev:live": "sh -c 'OPENCODE_TUI_CHANNEL=dev OPENCODE_PASSWORD=\"$(opencode service get password)\" exec bun run dev \"$@\" --server \"$(opencode service status)\"' --",
    "dev:vite": "bun run --cwd packages/cli --conditions=browser dev/vite.ts",
-    "dev:vite:live": "sh -c 'OPENCODE_TUI_CHANNEL=dev OPENCODE_PASSWORD=\"$(opencode2 service get password)\" exec bun run dev:vite \"$@\" --server \"$(opencode2 service status)\"' --",
+    "dev:vite:live": "sh -c 'OPENCODE_TUI_CHANNEL=dev OPENCODE_PASSWORD=\"$(opencode service get password)\" exec bun run dev:vite \"$@\" --server \"$(opencode service status)\"' --",
    "dev:desktop": "bun --cwd packages/desktop dev",
    "dev:web": "bun --cwd packages/app dev",
    "dev:console": "ulimit -n 10240 2>/dev/null; bun run --cwd packages/console/app dev",
@@ -122,13 +122,14 @@ Keep provider facades small and explicit:

 ### Provider Package Entrypoints

-Catalog-selected native providers use package-like export paths from `@opencode/ai`. They are internal entrypoints in one npm package, not separately published provider packages. Every entrypoint implements `ProviderPackage.Definition` and exposes `model(modelID, settings)`, where settings are serializable provider configuration plus common `headers`, `body`, and `limits` overlays.
+Catalog-selected native providers use package-like export paths from `@opencode/ai`. They are internal entrypoints in one npm package, not separately published provider packages. Every entrypoint implements `ProviderPackage.Definition` and exposes `model(modelID, settings)`, where settings are one flat serializable object: the connection keys the entrypoint declares (`apiKey`, `baseURL`, `region`, …), the common `headers` and `body` overlays, and the protocol's request options (`reasoningEffort`, `thinking`, …) side by side. Each entrypoint destructures its own connection keys and passes the rest to the route as `providerOptions`; there is no nested `providerOptions` at the entrypoint.

 ```ts
 import { model } from "@opencode/ai/providers/openai/responses"

 const selected = model("gpt-5", {
  apiKey,
+  reasoningEffort: "high",
 })
 ```

@@ -1,6 +1,6 @@
 {
  "$schema": "https://json.schemastore.org/package.json",
-  "version": "1.17.20",
+  "version": "2.0.0",
  "name": "@opencode/ai",
  "type": "module",
  "license": "MIT",
@@ -1708,7 +1708,10 @@ export const transport = <
 }

 function requiredBetaHeaders(body: Pick<AnthropicMessagesBody, "messages" | "context_management" | "thinking">) {
-  const betas: string[] = []
+  // Always request interleaved thinking. The API accepts the header on any
+  // model and ignores it where unsupported, while manual-thinking models need
+  // it for thinking between tool calls.
+  const betas: string[] = ["interleaved-thinking-2025-05-14"]
  const requestsCompaction = (body.context_management?.edits.length ?? 0) > 0
  const replaysCompaction = body.messages.some((message) =>
    message.content.some((block) => block.type === "compaction"),
@@ -18,7 +18,6 @@ const WebSocketResponseCreate = Schema.StructWithRest(Schema.Struct({ type: Sche
 ])
 const decodeMessage = ProviderShared.validateWith(Schema.decodeUnknownEffect(WebSocketResponseCreate))
 const encodeMessage = Schema.encodeSync(Schema.fromJsonString(WebSocketResponseCreate))
-const decodeEvent = Schema.decodeUnknownEffect(OpenResponses.protocol.stream.event)

 export interface Options {
  readonly id: string
@@ -27,6 +26,7 @@ export interface Options {
  readonly enabled?: (url: string) => boolean
  readonly url?: (url: string) => string
  readonly headers?: (headers: Headers.Headers) => Headers.Headers
+  readonly continuation?: OpenResponsesContinuation.Shape
 }

 export interface Prepared {
@@ -60,7 +60,7 @@ const driver = (options: Options, body: string): WebSocketChannelDriver => {
      }),
    observe: (_create, frame) =>
      Effect.gen(function* () {
-        const event = yield* decodeEvent(frame).pipe(
+        const event = yield* OpenResponses.decodeChannelEvent(frame).pipe(
          Effect.mapError((cause) =>
            ProviderShared.eventError(options.id, `Invalid ${options.name} WebSocket event`, frame, cause),
          ),
@@ -163,6 +163,7 @@ export const transport = <Body>(options: Options): Transport<Body, Prepared, str
                    request: create.request,
                    message: create.message,
                    base,
+                    continuation: options.continuation,
                  }),
                }
              })
@@ -6,7 +6,6 @@ import { OpenResponses } from "./open-responses.js"

 const PROTOCOL = "open-responses.websocket.v1"
 const VERSION = 1
-const decodeEvent = Schema.decodeUnknownEffect(OpenResponses.protocol.stream.event)

 interface CheckpointValue {
  readonly version: typeof VERSION
@@ -15,12 +14,19 @@ interface CheckpointValue {
  readonly output: ReadonlyArray<unknown>
 }

+/**
+ * Fields to send next to `previous_response_id` on an incremental step, or undefined to send the step in full.
+ * Whether omitted fields carry over from the continued response is provider behavior the route must know.
+ */
+export type Shape = (request: Readonly<Record<string, unknown>>) => Readonly<Record<string, unknown>> | undefined
+
 export interface DriverInput {
  readonly id: string
  readonly name: string
  readonly request: Readonly<Record<string, unknown>>
  readonly message: string
  readonly base: WebSocketChannelDriver
+  readonly continuation?: Shape
 }

 const checkpointValue = (checkpoint: ChannelCheckpoint | undefined): CheckpointValue | undefined => {
@@ -127,22 +133,26 @@ const rejected = (

 export const driver = (input: DriverInput): WebSocketChannelDriver => {
  const { previous_response_id: _previousResponseID, ...request } = input.request
+  const shape = input.continuation ?? ((fields: Readonly<Record<string, unknown>>) => fields)
  let output: OpenResponses.StreamItem[] = []
  return {
    create: (checkpoint) =>
      Effect.sync(() => {
        output = []
        const previous = checkpointValue(checkpoint)
-        const delta = previous ? incremental(request, previous) : undefined
-        if (!previous || !delta) return { message: ProviderShared.encodeJson(request), mode: "full" as const }
+        // Ask the route first: diffing the whole history is wasted when it declines the continuation.
+        const fields = previous ? shape(request) : undefined
+        const delta = previous && fields ? incremental(request, previous) : undefined
+        if (!previous || !fields || !delta)
+          return { message: ProviderShared.encodeJson(request), mode: "full" as const }
        return {
-          message: ProviderShared.encodeJson({ ...request, input: delta, previous_response_id: previous.responseID }),
+          message: ProviderShared.encodeJson({ ...fields, input: delta, previous_response_id: previous.responseID }),
          mode: "incremental" as const,
        }
      }),
    observe: (create, frame) =>
      Effect.gen(function* () {
-        const event = yield* decodeEvent(frame).pipe(
+        const event = yield* OpenResponses.decodeChannelEvent(frame).pipe(
          Effect.mapError((cause) =>
            ProviderShared.eventError(input.id, `Invalid ${input.name} WebSocket event`, frame, cause),
          ),
@@ -195,4 +205,4 @@ export const driver = (input: DriverInput): WebSocketChannelDriver => {
  }
 }

-export const OpenResponsesContinuation = { driver } as const
+export * as OpenResponsesContinuation from "./open-responses-continuation.js"
@@ -1,4 +1,4 @@
-import { Effect, Option, Schema } from "effect"
+import { Effect, Option, Schema, SchemaGetter } from "effect"
 import type { Content } from "@opencode/schema/tool"
 import { HttpTransport } from "../route/transport/index.js"
 import { Protocol } from "../route/protocol.js"
@@ -325,9 +325,8 @@ export const StreamItem = Schema.StructWithRest(
 export type StreamItem = Schema.Schema.Type<typeof StreamItem>
 export type OutputItem = StreamItem & { readonly id: string }

-// The Responses schema puts streaming error details at the top level and
-// response failures under `response.error`. WebSocket failures use an
-// event-level `error` envelope, so accept all three shapes here.
+// Responses-compatible providers put streaming error details at the top level or
+// under `error`, and response failures under `response.error`. Accept all three shapes.
 // https://www.openresponses.org/specification
 const OpenResponsesErrorPayload = Schema.Struct({
  type: optionalNull(Schema.String),
@@ -401,10 +400,39 @@ export const Event = Schema.StructWithRest(
    headers: Schema.optional(Schema.Unknown),
  }),
  [Schema.Record(Schema.String, Schema.Unknown)],
+).pipe(
+  Schema.decode({
+    decode: SchemaGetter.transform((event) => {
+      if (event.type !== "error" || event.error != null) return event
+      const { code, message, param, ...rest } = event
+      if (code === undefined && message === undefined && param === undefined) return event
+      // Flat errors (for example, Meta's) can also arrive through generic Responses endpoints.
+      return { ...rest, error: { code, message, param } }
+    }),
+    encode: SchemaGetter.passthrough(),
+  }),
 )
 export type Event = Schema.Schema.Type<typeof Event>
 export type NormalizedEvent = Event & { readonly item?: OutputItem | null }

+const decodeEventValue = Schema.decodeUnknownEffect(Event)
+const decodeFrame = Schema.decodeUnknownEffect(ProviderShared.Json)
+
+/**
+ * Decodes one WebSocket frame. xAI answers a rejected `response.create` with `{ "error": { "message", "type" } }` and no
+ * event type; that envelope reads as an error event so the failure classifies instead of failing decoding.
+ */
+export const decodeChannelEvent = (frame: string) =>
+  decodeFrame(frame).pipe(
+    Effect.flatMap((value) =>
+      decodeEventValue(
+        ProviderShared.isRecord(value) && value.type === undefined && ProviderShared.isRecord(value.error)
+          ? { ...value, type: "error" }
+          : value,
+      ),
+    ),
+  )
+
 export interface ProviderAdapter {
  readonly id: string
  readonly name: string
@@ -1,6 +1,10 @@
 import type { LanguageModel, ProviderOptions } from "./schema/index.js"
 import type { CompactionOperations } from "./route/client.js"

+/**
+ * Flat, serializable settings for `model(modelID, settings)`. Each entrypoint declares the connection keys it
+ * reads; every other key is a request option for the route's protocol.
+ */
 export interface Settings extends Readonly<Record<string, unknown>> {
  readonly baseURL?: string
  readonly headers?: Readonly<Record<string, string>>
@@ -1,3 +1,4 @@
+import { Struct } from "effect"
 import type { ProviderPackage } from "../provider-package.js"
 import { AlibabaChat } from "../protocols/alibaba-chat.js"
 import { AlibabaMessages } from "../protocols/alibaba-messages.js"
@@ -6,7 +7,7 @@ import { AuthOptions, type AtLeastOne, type ProviderAuthOption } from "../route/
 import { Route, type RouteDefaultsInput } from "../route/client.js"
 import { Endpoint } from "../route/endpoint.js"
 import { Framing } from "../route/framing.js"
-import { ProviderID, ToolDefinition, type ModelID } from "../schema/index.js"
+import { ProviderConfigurationError, ProviderID, ToolDefinition, type ModelID } from "../schema/index.js"

 export const id = ProviderID.make("alibaba")

@@ -34,9 +35,9 @@ export type Config = Location &
    readonly providerOptions?: ChatOptionsInput | MessagesOptionsInput | ResponsesOptionsInput
  }
 export type Settings<Options = ChatOptionsInput> = Location &
-  ProviderPackage.Settings & {
+  ProviderPackage.Settings &
+  Options & {
    readonly apiKey?: string
-    readonly providerOptions?: Options
  }

 const hosts = new Map<string, string>([
@@ -82,8 +83,13 @@ export const configure = (input: Config) => {
        ? hosts.get(region)
        : `${workspaceID}.${region}.maas.aliyuncs.com`
  if (baseURL === undefined) {
-    if (region === undefined) throw new Error("Alibaba requires region or baseURL")
-    if (host === undefined) throw new Error(`Alibaba region ${region} requires workspaceID or baseURL`)
+    if (region === undefined)
+      throw new ProviderConfigurationError({ provider: id, message: "Alibaba requires region or baseURL" })
+    if (host === undefined)
+      throw new ProviderConfigurationError({
+        provider: id,
+        message: `Alibaba region ${region} requires workspaceID or baseURL`,
+      })
  }
  const opts = { ...rest, auth: AuthOptions.bearer(input, ["DASHSCOPE_API_KEY", "ALIBABA_API_KEY"]) }
  const common = { ...opts, endpoint: { baseURL: baseURL ?? `https://${host}/compatible-mode/v1` } }
@@ -115,7 +121,11 @@ export const responsesModel: ProviderPackage.Definition<

 function fromSettings(input: Settings<Config["providerOptions"]>) {
  const { body, ...rest } = input
-  return configure({ ...rest, http: body === undefined ? undefined : { body } })
+  return configure({
+    ...rest,
+    http: body === undefined ? undefined : { body },
+    providerOptions: Struct.omit(rest, ["apiKey", "baseURL", "headers", "region", "workspaceID"]),
+  })
 }

 export const webSearch = () => hostedTool("web_search", "Search the web with Alibaba's hosted search tool.")
@@ -4,7 +4,7 @@ import type { ProviderPackage } from "../provider-package.js"
 import { OpenAIChat } from "../protocols/openai-chat.js"
 import { OpenResponses } from "../protocols/open-responses.js"
 import { BedrockAuth, type Credentials } from "../protocols/utils/bedrock-auth.js"
-import { ProviderID, type ModelID } from "../schema/index.js"
+import { ProviderConfigurationError, ProviderID, type ModelID } from "../schema/index.js"
 import { withOpenAIOptions, type OpenAIProviderOptionsInput } from "./openai-options.js"

 export const id = ProviderID.make("amazon-bedrock")
@@ -23,16 +23,16 @@ export type Config = RouteDefaultsInput & {
  readonly providerOptions?: OpenAIProviderOptionsInput
 }

-export interface Settings extends ProviderPackage.Settings {
-  readonly apiKey?: string
-  readonly auth?: "bearer" | "sigv4"
-  readonly baseURL?: string
-  readonly credentials?: Credentials
-  readonly profile?: string
-  readonly region?: string
-  readonly topP?: number
-  readonly providerOptions?: OpenAIProviderOptionsInput
-}
+export type Settings = ProviderPackage.Settings &
+  OpenAIProviderOptionsInput & {
+    readonly apiKey?: string
+    readonly auth?: "bearer" | "sigv4"
+    readonly baseURL?: string
+    readonly credentials?: Credentials
+    readonly profile?: string
+    readonly region?: string
+    readonly topP?: number
+  }

 const responsesRoute = Route.make({
  id: "bedrock-mantle-responses",
@@ -79,9 +79,12 @@ const defaults = (input: Config) => {

 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")
+    throw new ProviderConfigurationError({ provider: id, message: "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")
+    throw new ProviderConfigurationError({
+      provider: id,
+      message: "Amazon Bedrock Mantle SigV4 auth does not accept apiKey",
+    })
  const configuredResponsesRoute = configuredRoute(responsesRoute, input)
  const configuredChatRoute = configuredRoute(chatRoute, input)
  const modelDefaults = defaults(input)
@@ -105,18 +108,29 @@ export const configure = (input: Config = {}) => {

 export const provider = configure()

-const fromSettings = (settings: Settings) =>
+const fromSettings = ({
+  apiKey,
+  auth,
+  baseURL,
+  body,
+  credentials,
+  headers,
+  profile,
+  region,
+  topP,
+  ...providerOptions
+}: 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,
+    apiKey,
+    auth,
+    baseURL,
+    credentials,
+    generation: topP === undefined ? undefined : { topP },
+    headers: headers === undefined ? undefined : { ...headers },
+    http: body === undefined ? undefined : { body: { ...body } },
+    profile,
+    providerOptions,
+    region,
  })

 export const chatModel: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (
@@ -1,6 +1,6 @@
 import type { RouteDefaultsInput } from "../route/client.js"
 import type { ProviderPackage } from "../provider-package.js"
-import { ProviderID, type ModelID } from "../schema/index.js"
+import { ProviderConfigurationError, 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"
@@ -39,8 +39,9 @@ const bedrockBaseURL = (region: string) => `https://bedrock-runtime.${region}.am
 const configuredRoute = (input: Config) => {
  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")
+    throw new ProviderConfigurationError({ provider: id, message: "Amazon Bedrock bearer auth requires apiKey" })
+  if (auth === "sigv4" && apiKey !== undefined)
+    throw new ProviderConfigurationError({ provider: id, message: "Amazon Bedrock SigV4 auth does not accept apiKey" })
  const resolvedRegion = BedrockAuth.resolveRegion(input)
  return BedrockConverse.route.with({
    ...rest,
@@ -3,7 +3,7 @@ import { AnthropicMessages } from "../protocols/anthropic-messages.js"
 import { Auth } from "../route/auth.js"
 import type { ProviderAuthOption } from "../route/auth-options.js"
 import type { RouteDefaultsInput } from "../route/client.js"
-import { ProviderID, type ModelID } from "../schema/index.js"
+import { ProviderConfigurationError, ProviderID, type ModelID } from "../schema/index.js"

 export type AnthropicOptionsInput = AnthropicMessages.OptionsInput
 export type AnthropicProviderOptionsInput = AnthropicMessages.ProviderOptionsInput
@@ -19,13 +19,13 @@ export type Config = RouteDefaultsInput &
  }

 export type Settings = ProviderPackage.Settings &
+  AnthropicMessages.ProviderOptionsInput &
  (
    | { readonly apiKey?: string; readonly authToken?: never }
    | { readonly apiKey?: never; readonly authToken?: string }
  ) & {
    readonly baseURL: string
    readonly provider?: string
-    readonly providerOptions?: AnthropicMessages.ProviderOptionsInput
  }

 export const routes = [AnthropicMessages.route]
@@ -36,8 +36,12 @@ const auth = (input: ProviderAuthOption<"optional">) => {
 }

 export const configure = (input: Config) => {
-  if (!input.baseURL) throw new Error("Anthropic-compatible providers require a baseURL")
  const provider = input.provider ?? "anthropic-compatible"
+  if (!input.baseURL)
+    throw new ProviderConfigurationError({
+      provider: ProviderID.make(provider),
+      message: "Anthropic-compatible providers require a baseURL",
+    })
  const { provider: _, baseURL, apiKey: _apiKey, auth: _auth, ...rest } = input
  const route = AnthropicMessages.route.with({
    ...rest,
@@ -59,17 +63,20 @@ export const provider = {

 export const model: ProviderPackage.Definition<Settings, AnthropicMessages.ProviderOptionsInput>["model"] = (
  modelID,
-  settings,
+  { apiKey, authToken, baseURL, body, headers, provider, ...providerOptions },
 ) => {
-  if (settings.apiKey !== undefined && settings.authToken !== undefined)
-    throw new Error("Anthropic-compatible apiKey cannot be combined with authToken")
+  if (apiKey !== undefined && authToken !== undefined)
+    throw new ProviderConfigurationError({
+      provider: ProviderID.make(provider ?? id),
+      message: "Anthropic-compatible apiKey cannot be combined with authToken",
+    })
  return configure({
-    ...(settings.authToken === undefined ? { apiKey: settings.apiKey } : { auth: Auth.bearer(settings.authToken) }),
-    baseURL: settings.baseURL,
-    headers: settings.headers === undefined ? undefined : { ...settings.headers },
-    http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    provider: settings.provider,
-    providerOptions: settings.providerOptions,
+    ...(authToken === undefined ? { apiKey: apiKey } : { auth: Auth.bearer(authToken) }),
+    baseURL,
+    headers: headers === undefined ? undefined : { ...headers },
+    http: body === undefined ? undefined : { body: { ...body } },
+    provider,
+    providerOptions,
  }).model(modelID)
 }

@@ -2,7 +2,7 @@ import type { RouteDefaultsInput } from "../route/client.js"
 import { Auth } from "../route/auth.js"
 import type { ProviderAuthOption } from "../route/auth-options.js"
 import type { ProviderPackage } from "../provider-package.js"
-import { ProviderID, type ModelID } from "../schema/index.js"
+import { ProviderConfigurationError, ProviderID, type ModelID } from "../schema/index.js"
 import { AnthropicMessages } from "../protocols/anthropic-messages.js"
 import { AnthropicCompatible } from "./anthropic-compatible.js"

@@ -21,12 +21,12 @@ export type Config = RouteDefaultsInput &
  }

 export type Settings = ProviderPackage.Settings &
+  AnthropicMessages.ProviderOptionsInput &
  (
    | { readonly apiKey?: string; readonly authToken?: never }
    | { readonly apiKey?: never; readonly authToken?: string }
  ) & {
    readonly baseURL?: string
-    readonly providerOptions?: AnthropicMessages.ProviderOptionsInput
  }

 const auth = (options: ProviderAuthOption<"optional">) => {
@@ -54,15 +54,18 @@ export const configure = (input: Config = {}) => {
 export const provider = configure()
 export const model: ProviderPackage.Definition<Settings, AnthropicMessages.ProviderOptionsInput>["model"] = (
  modelID,
-  settings,
+  { apiKey, authToken, baseURL, body, headers, ...providerOptions },
 ) => {
-  if (settings.apiKey !== undefined && settings.authToken !== undefined)
-    throw new Error("Anthropic apiKey cannot be combined with authToken")
+  if (apiKey !== undefined && authToken !== undefined)
+    throw new ProviderConfigurationError({
+      provider: id,
+      message: "Anthropic apiKey cannot be combined with authToken",
+    })
  return configure({
-    ...(settings.authToken === undefined ? { apiKey: settings.apiKey } : { auth: Auth.bearer(settings.authToken) }),
-    baseURL: settings.baseURL,
-    headers: settings.headers === undefined ? undefined : { ...settings.headers },
-    http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    providerOptions: settings.providerOptions,
+    ...(authToken === undefined ? { apiKey: apiKey } : { auth: Auth.bearer(authToken) }),
+    baseURL,
+    headers: headers === undefined ? undefined : { ...headers },
+    http: body === undefined ? undefined : { body: { ...body } },
+    providerOptions,
  }).model(modelID)
 }
@@ -3,7 +3,7 @@ import { Auth } from "../route/auth.js"
 import { type AtLeastOne, type ProviderAuthOption } from "../route/auth-options.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 { ProviderConfigurationError, ProviderID, type ModelID } from "../schema/index.js"
 import * as OpenAIChat from "../protocols/openai-chat.js"
 import * as OpenAIResponses from "../protocols/openai-responses.js"
 import { ProviderShared } from "../protocols/shared.js"
@@ -28,12 +28,12 @@ export type LanguageModelOptions = AzureURL &
 export type Config = LanguageModelOptions

 export type Settings = ProviderPackage.Settings &
+  OpenAIProviderOptionsInput &
  AzureURL & {
    readonly apiKey?: string
    readonly apiVersion?: string
    readonly queryParams?: Readonly<Record<string, string>>
    readonly useDeploymentBasedUrls?: boolean
-    readonly providerOptions?: OpenAIProviderOptionsInput
  }

 const resourceBaseURL = (resourceName: string) => `https://${resourceName.trim()}.openai.azure.com/openai`
@@ -151,19 +151,29 @@ export const provider = {
  configure,
 }

-const config = (settings: Settings): Config => {
+const config = ({
+  apiKey,
+  apiVersion,
+  baseURL,
+  body,
+  headers,
+  queryParams,
+  resourceName,
+  useDeploymentBasedUrls,
+  ...providerOptions
+}: Settings): Config => {
  const common = {
-    apiKey: settings.apiKey,
-    apiVersion: settings.apiVersion,
-    headers: settings.headers === undefined ? undefined : { ...settings.headers },
-    http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    providerOptions: settings.providerOptions,
-    queryParams: settings.queryParams === undefined ? undefined : { ...settings.queryParams },
-    useDeploymentBasedUrls: settings.useDeploymentBasedUrls,
+    apiKey,
+    apiVersion,
+    headers: headers === undefined ? undefined : { ...headers },
+    http: body === undefined ? undefined : { body: { ...body } },
+    providerOptions,
+    queryParams: queryParams === undefined ? undefined : { ...queryParams },
+    useDeploymentBasedUrls,
  }
-  if (settings.baseURL !== undefined) return { ...common, baseURL: settings.baseURL }
-  if (settings.resourceName !== undefined) return { ...common, resourceName: settings.resourceName }
-  throw new Error("Azure requires resourceName or baseURL")
+  if (baseURL !== undefined) return { ...common, baseURL }
+  if (resourceName !== undefined) return { ...common, resourceName }
+  throw new ProviderConfigurationError({ provider: id, message: "Azure requires resourceName or baseURL" })
 }

 export const responsesModel: ProviderPackage.Definition<
@@ -15,11 +15,11 @@ export type LanguageModelOptions = Omit<RouteDefaultsInput, "providerOptions"> &
    readonly providerOptions?: OpenAIProviderOptionsInput
  }

-export interface Settings extends ProviderPackage.Settings {
-  readonly apiKey?: string
-  readonly baseURL?: string
-  readonly providerOptions?: OpenAIProviderOptionsInput
-}
+export type Settings = ProviderPackage.Settings &
+  OpenAIProviderOptionsInput & {
+    readonly apiKey?: string
+    readonly baseURL?: string
+  }

 export const route = Route.make({
  id: "baseten-chat",
@@ -48,13 +48,16 @@ export const configure = (input: LanguageModelOptions = {}) => {

 export const provider = configure()

-export const model: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (modelID, settings) =>
+export const model: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (
+  modelID,
+  { apiKey, baseURL, body, headers, ...providerOptions },
+) =>
  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,
+    apiKey,
+    baseURL,
+    headers: headers === undefined ? undefined : { ...headers },
+    http: body === undefined ? undefined : { body: { ...body } },
+    providerOptions,
  }).model(modelID)

 export * as Baseten from "./baseten.js"
@@ -15,11 +15,11 @@ export type LanguageModelOptions = Omit<RouteDefaultsInput, "providerOptions"> &
    readonly providerOptions?: OpenAIProviderOptionsInput
  }

-export interface Settings extends ProviderPackage.Settings {
-  readonly apiKey?: string
-  readonly baseURL?: string
-  readonly providerOptions?: OpenAIProviderOptionsInput
-}
+export type Settings = ProviderPackage.Settings &
+  OpenAIProviderOptionsInput & {
+    readonly apiKey?: string
+    readonly baseURL?: string
+  }

 export const route = Route.make({
  id: "cerebras-chat",
@@ -52,11 +52,14 @@ export const configure = (input: LanguageModelOptions = {}) => {

 export const provider = configure()

-export const model: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (modelID, settings) =>
+export const model: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (
+  modelID,
+  { apiKey, baseURL, body, headers, ...providerOptions },
+) =>
  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,
+    apiKey,
+    baseURL,
+    headers: headers === undefined ? undefined : { ...headers },
+    http: body === undefined ? undefined : { body: { ...body } },
+    providerOptions,
  }).model(modelID)
@@ -5,7 +5,7 @@ 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 { ProviderConfigurationError, ProviderID, type ModelID } from "../schema/index.js"
 import type { OpenAIProviderOptionsInput } from "./openai-options.js"

 export const id = ProviderID.make("cloudflare-ai-gateway")
@@ -27,15 +27,19 @@ export type LanguageModelOptions = GatewayURL &
  }

 export type Settings = ProviderPackage.Settings &
+  OpenAIProviderOptionsInput &
  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")
+  if (!input.accountId)
+    throw new ProviderConfigurationError({
+      provider: id,
+      message: "CloudflareAIGateway.configure requires accountId unless baseURL is supplied",
+    })
  return `https://gateway.ai.cloudflare.com/v1/${encodeURIComponent(input.accountId)}/${encodeURIComponent(input.gatewayId?.trim() || "default")}/compat`
 }

@@ -85,14 +89,25 @@ export const configure = (input: LanguageModelOptions) => {

 export const provider = { id, configure }

-export const model: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (modelID, settings) =>
-  configure({
-    apiKey: settings.apiKey,
-    gatewayApiKey: settings.gatewayApiKey,
+export const model: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (modelID, settings) => {
+  const {
+    accountId: _,
+    apiKey,
+    baseURL: _url,
+    body,
+    gatewayApiKey,
+    gatewayId: _id,
+    headers,
+    ...providerOptions
+  } = settings
+  return configure({
+    apiKey,
+    gatewayApiKey,
    baseURL: baseURL(settings),
-    headers: settings.headers === undefined ? undefined : { ...settings.headers },
-    http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    providerOptions: settings.providerOptions,
+    headers: headers === undefined ? undefined : { ...headers },
+    http: body === undefined ? undefined : { body: { ...body } },
+    providerOptions,
  }).model(modelID)
+}

 export * as CloudflareAIGateway from "./cloudflare-ai-gateway.js"
@@ -3,7 +3,7 @@ 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 { ProviderConfigurationError, ProviderID, type ModelID } from "../schema/index.js"
 import type { OpenAIProviderOptionsInput } from "./openai-options.js"

 export const id = ProviderID.make("cloudflare-workers-ai")
@@ -21,14 +21,18 @@ export type LanguageModelOptions = WorkersAIURL &
  }

 export type Settings = ProviderPackage.Settings &
+  OpenAIProviderOptionsInput &
  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")
+  if (!input.accountId)
+    throw new ProviderConfigurationError({
+      provider: id,
+      message: "CloudflareWorkersAI.configure requires accountId unless baseURL is supplied",
+    })
  return `https://api.cloudflare.com/client/v4/accounts/${encodeURIComponent(input.accountId)}/ai/v1`
 }

@@ -59,13 +63,15 @@ export const configure = (input: LanguageModelOptions) => {

 export const provider = { id, configure }

-export const model: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (modelID, settings) =>
-  configure({
-    apiKey: settings.apiKey,
+export const model: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (modelID, settings) => {
+  const { accountId: _, apiKey, baseURL: _url, body, headers, ...providerOptions } = settings
+  return configure({
+    apiKey,
    baseURL: baseURL(settings),
-    headers: settings.headers === undefined ? undefined : { ...settings.headers },
-    http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    providerOptions: settings.providerOptions,
+    headers: headers === undefined ? undefined : { ...headers },
+    http: body === undefined ? undefined : { body: { ...body } },
+    providerOptions,
  }).model(modelID)
+}

 export * as CloudflareWorkersAI from "./cloudflare-workers-ai.js"
@@ -15,11 +15,11 @@ export type LanguageModelOptions = Omit<RouteDefaultsInput, "providerOptions"> &
    readonly providerOptions?: OpenAIProviderOptionsInput
  }

-export interface Settings extends ProviderPackage.Settings {
-  readonly apiKey?: string
-  readonly baseURL?: string
-  readonly providerOptions?: OpenAIProviderOptionsInput
-}
+export type Settings = ProviderPackage.Settings &
+  OpenAIProviderOptionsInput & {
+    readonly apiKey?: string
+    readonly baseURL?: string
+  }

 export const route = Route.make({
  id: "deepinfra-chat",
@@ -55,11 +55,14 @@ export const configure = (input: LanguageModelOptions = {}) => {

 export const provider = configure()

-export const model: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (modelID, settings) =>
+export const model: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (
+  modelID,
+  { apiKey, baseURL, body, headers, ...providerOptions },
+) =>
  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,
+    apiKey,
+    baseURL,
+    headers: headers === undefined ? undefined : { ...headers },
+    http: body === undefined ? undefined : { body: { ...body } },
+    providerOptions,
  }).model(modelID)
@@ -15,11 +15,11 @@ export type LanguageModelOptions = Omit<RouteDefaultsInput, "providerOptions"> &
    readonly providerOptions?: OpenAIProviderOptionsInput
  }

-export interface Settings extends ProviderPackage.Settings {
-  readonly apiKey?: string
-  readonly baseURL?: string
-  readonly providerOptions?: OpenAIProviderOptionsInput
-}
+export type Settings = ProviderPackage.Settings &
+  OpenAIProviderOptionsInput & {
+    readonly apiKey?: string
+    readonly baseURL?: string
+  }

 export const route = Route.make({
  id: "deepseek-chat",
@@ -52,13 +52,16 @@ export const configure = (input: LanguageModelOptions = {}) => {

 export const provider = configure()

-export const model: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (modelID, settings) =>
+export const model: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (
+  modelID,
+  { apiKey, baseURL, body, headers, ...providerOptions },
+) =>
  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,
+    apiKey,
+    baseURL,
+    headers: headers === undefined ? undefined : { ...headers },
+    http: body === undefined ? undefined : { body: { ...body } },
+    providerOptions,
  }).model(modelID)

 export * as DeepSeek from "./deepseek.js"
@@ -15,11 +15,11 @@ export type LanguageModelOptions = Omit<RouteDefaultsInput, "providerOptions"> &
    readonly providerOptions?: OpenAIProviderOptionsInput
  }

-export interface Settings extends ProviderPackage.Settings {
-  readonly apiKey?: string
-  readonly baseURL?: string
-  readonly providerOptions?: OpenAIProviderOptionsInput
-}
+export type Settings = ProviderPackage.Settings &
+  OpenAIProviderOptionsInput & {
+    readonly apiKey?: string
+    readonly baseURL?: string
+  }

 export const route = Route.make({
  id: "fireworks-chat",
@@ -48,13 +48,16 @@ export const configure = (input: LanguageModelOptions = {}) => {

 export const provider = configure()

-export const model: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (modelID, settings) =>
+export const model: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (
+  modelID,
+  { apiKey, baseURL, body, headers, ...providerOptions },
+) =>
  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,
+    apiKey,
+    baseURL,
+    headers: headers === undefined ? undefined : { ...headers },
+    http: body === undefined ? undefined : { body: { ...body } },
+    providerOptions,
  }).model(modelID)

 export * as Fireworks from "./fireworks.js"
@@ -2,7 +2,7 @@ import type { ProviderPackage } from "../provider-package.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 { ProviderConfigurationError, ProviderID, type ModelID } from "../schema/index.js"
 import { GoogleVertexShared } from "./google-vertex-shared.js"
 import type { OpenAIProviderOptionsInput } from "./openai-options.js"

@@ -16,14 +16,14 @@ export type Config = RouteDefaultsInput &
    readonly providerOptions?: OpenAIProviderOptionsInput
  }

-export interface Settings extends ProviderPackage.Settings {
-  readonly accessToken?: string
-  readonly apiKey?: never
-  readonly baseURL?: string
-  readonly location?: string
-  readonly project?: string
-  readonly providerOptions?: OpenAIProviderOptionsInput
-}
+export type Settings = ProviderPackage.Settings &
+  OpenAIProviderOptionsInput & {
+    readonly accessToken?: string
+    readonly apiKey?: never
+    readonly baseURL?: string
+    readonly location?: string
+    readonly project?: string
+  }

 const route = Route.make({
  id: "google-vertex-chat",
@@ -37,7 +37,8 @@ const route = Route.make({
 export const routes = [route]

 const configuredRoute = (input: Config) => {
-  if ("apiKey" in input && input.apiKey !== undefined) throw new Error("Google Vertex Chat does not support API keys")
+  if ("apiKey" in input && input.apiKey !== undefined)
+    throw new ProviderConfigurationError({ provider: id, message: "Google Vertex Chat does not support API keys" })
  const {
    accessToken: _accessToken,
    auth: _auth,
@@ -73,15 +74,19 @@ export const provider = {
  configure,
 }

-export const model: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (modelID, settings) => {
-  if (settings.apiKey !== undefined) throw new Error("Google Vertex Chat does not support API keys")
+export const model: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (
+  modelID,
+  { accessToken, apiKey, baseURL, body, headers, location, project, ...providerOptions },
+) => {
+  if (apiKey !== undefined)
+    throw new ProviderConfigurationError({ provider: id, message: "Google Vertex Chat does not support API keys" })
  return configure({
-    accessToken: settings.accessToken,
-    baseURL: settings.baseURL,
-    headers: settings.headers === undefined ? undefined : { ...settings.headers },
-    http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    location: settings.location,
-    project: settings.project,
-    providerOptions: settings.providerOptions,
+    accessToken,
+    baseURL,
+    headers: headers === undefined ? undefined : { ...headers },
+    http: body === undefined ? undefined : { body: { ...body } },
+    location,
+    project,
+    providerOptions,
  }).model(modelID)
 }
@@ -5,7 +5,7 @@ import { Auth } from "../route/auth.js"
 import { Route, type RouteDefaultsInput } from "../route/client.js"
 import { Endpoint } from "../route/endpoint.js"
 import { Protocol } from "../route/protocol.js"
-import { ProviderID, type ModelID } from "../schema/index.js"
+import { ProviderConfigurationError, ProviderID, type ModelID } from "../schema/index.js"
 import { GoogleVertexShared } from "./google-vertex-shared.js"

 export type AnthropicOptionsInput = AnthropicMessages.OptionsInput
@@ -25,14 +25,14 @@ export type Config = RouteDefaultsInput &
    readonly providerOptions?: AnthropicMessages.ProviderOptionsInput
  }

-export interface Settings extends ProviderPackage.Settings {
-  readonly accessToken?: string
-  readonly apiKey?: never
-  readonly baseURL?: string
-  readonly location?: string
-  readonly project?: string
-  readonly providerOptions?: AnthropicMessages.ProviderOptionsInput
-}
+export type Settings = ProviderPackage.Settings &
+  AnthropicMessages.ProviderOptionsInput & {
+    readonly accessToken?: string
+    readonly apiKey?: never
+    readonly baseURL?: string
+    readonly location?: string
+    readonly project?: string
+  }

 const route = Route.make({
  id: "google-vertex-messages",
@@ -67,7 +67,7 @@ export const routes = [route]

 const configuredRoute = (input: Config) => {
  if ("apiKey" in input && input.apiKey !== undefined)
-    throw new Error("Google Vertex Messages does not support API keys")
+    throw new ProviderConfigurationError({ provider: id, message: "Google Vertex Messages does not support API keys" })
  const {
    accessToken: _accessToken,
    auth: _auth,
@@ -105,16 +105,17 @@ export const provider = {

 export const model: ProviderPackage.Definition<Settings, AnthropicMessages.ProviderOptionsInput>["model"] = (
  modelID,
-  settings,
+  { accessToken, apiKey, baseURL, body, headers, location, project, ...providerOptions },
 ) => {
-  if (settings.apiKey !== undefined) throw new Error("Google Vertex Messages does not support API keys")
+  if (apiKey !== undefined)
+    throw new ProviderConfigurationError({ provider: id, message: "Google Vertex Messages does not support API keys" })
  return configure({
-    accessToken: settings.accessToken,
-    baseURL: settings.baseURL,
-    headers: settings.headers === undefined ? undefined : { ...settings.headers },
-    http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    location: settings.location,
-    project: settings.project,
-    providerOptions: settings.providerOptions,
+    accessToken,
+    baseURL,
+    headers: headers === undefined ? undefined : { ...headers },
+    http: body === undefined ? undefined : { body: { ...body } },
+    location,
+    project,
+    providerOptions,
  }).model(modelID)
 }
@@ -2,7 +2,7 @@ import type { ProviderPackage } from "../provider-package.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 { ProviderConfigurationError, ProviderID, type ModelID } from "../schema/index.js"
 import { GoogleVertexShared } from "./google-vertex-shared.js"
 import type { OpenResponsesProviderOptionsInput } from "./open-responses-options.js"

@@ -16,14 +16,14 @@ export type Config = RouteDefaultsInput &
    readonly providerOptions?: OpenResponsesProviderOptionsInput
  }

-export interface Settings extends ProviderPackage.Settings {
-  readonly accessToken?: string
-  readonly apiKey?: never
-  readonly baseURL?: string
-  readonly location?: string
-  readonly project?: string
-  readonly providerOptions?: OpenResponsesProviderOptionsInput
-}
+export type Settings = ProviderPackage.Settings &
+  OpenResponsesProviderOptionsInput & {
+    readonly accessToken?: string
+    readonly apiKey?: never
+    readonly baseURL?: string
+    readonly location?: string
+    readonly project?: string
+  }

 const route = Route.make({
  id: "google-vertex-responses",
@@ -39,7 +39,7 @@ export const routes = [route]

 const configuredRoute = (input: Config) => {
  if ("apiKey" in input && input.apiKey !== undefined)
-    throw new Error("Google Vertex Responses does not support API keys")
+    throw new ProviderConfigurationError({ provider: id, message: "Google Vertex Responses does not support API keys" })
  const {
    accessToken: _accessToken,
    auth: _auth,
@@ -77,16 +77,17 @@ export const provider = {

 export const model: ProviderPackage.Definition<Settings, OpenResponsesProviderOptionsInput>["model"] = (
  modelID,
-  settings,
+  { accessToken, apiKey, baseURL, body, headers, location, project, ...providerOptions },
 ) => {
-  if (settings.apiKey !== undefined) throw new Error("Google Vertex Responses does not support API keys")
+  if (apiKey !== undefined)
+    throw new ProviderConfigurationError({ provider: id, message: "Google Vertex Responses does not support API keys" })
  return configure({
-    accessToken: settings.accessToken,
-    baseURL: settings.baseURL,
-    headers: settings.headers === undefined ? undefined : { ...settings.headers },
-    http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    location: settings.location,
-    project: settings.project,
-    providerOptions: settings.providerOptions,
+    accessToken,
+    baseURL,
+    headers: headers === undefined ? undefined : { ...headers },
+    http: body === undefined ? undefined : { body: { ...body } },
+    location,
+    project,
+    providerOptions,
  }).model(modelID)
 }
@@ -1,8 +1,10 @@
 import type { AnyAuthClient } from "google-auth-library"
 import { Effect, Redacted } from "effect"
 import { Auth, MissingCredentialError } from "../route/auth.js"
+import { ProviderConfigurationError, ProviderID } from "../schema/index.js"

 const SCOPE = "https://www.googleapis.com/auth/cloud-platform"
+const id = ProviderID.make("google-vertex")

 export type OAuthOptions =
  | { readonly accessToken?: string; readonly auth?: never }
@@ -35,12 +37,18 @@ export const host = (location: string) => {

 export const requireProject = (value: string | undefined) => {
  if (value) return value
-  throw new Error("Google Vertex requires a project when baseURL is not configured")
+  throw new ProviderConfigurationError({
+    provider: id,
+    message: "Google Vertex requires a project when baseURL is not configured",
+  })
 }

 export const apiKey = (input: ApiKeyOptions) => {
  if (input.apiKey !== undefined && (input.accessToken !== undefined || input.auth !== undefined))
-    throw new Error("Google Vertex apiKey cannot be combined with accessToken or auth")
+    throw new ProviderConfigurationError({
+      provider: id,
+      message: "Google Vertex apiKey cannot be combined with accessToken or auth",
+    })
  if (input.accessToken !== undefined || input.auth !== undefined) return undefined
  return input.apiKey ?? process.env.GOOGLE_VERTEX_API_KEY
 }
@@ -68,7 +76,10 @@ const adc = (project?: string) => {

 export const oauth = (input: OAuthOptions, project?: string) => {
  if (input.accessToken !== undefined && input.auth !== undefined)
-    throw new Error("Google Vertex accessToken cannot be combined with auth")
+    throw new ProviderConfigurationError({
+      provider: id,
+      message: "Google Vertex accessToken cannot be combined with auth",
+    })
  if (input.auth) return input.auth
  if (input.accessToken !== undefined) return Auth.bearer(input.accessToken)
  return adc(project)
@@ -6,7 +6,7 @@ import { Auth } from "../route/auth.js"
 import { Route, type RouteDefaultsInput } from "../route/client.js"
 import { Endpoint } from "../route/endpoint.js"
 import { Framing } from "../route/framing.js"
-import { ProviderID, type LLMRequest, type ModelID } from "../schema/index.js"
+import { ProviderConfigurationError, ProviderID, type LLMRequest, type ModelID } from "../schema/index.js"
 import { GoogleVertexShared } from "./google-vertex-shared.js"

 export interface GeminiOptionsInput extends Gemini.OptionsInput {
@@ -26,6 +26,7 @@ export type Config = RouteDefaultsInput &
  }

 export type Settings = ProviderPackage.Settings &
+  GeminiProviderOptionsInput &
  (
    | { readonly accessToken?: string; readonly apiKey?: never }
    | { readonly accessToken?: never; readonly apiKey?: string }
@@ -33,7 +34,6 @@ export type Settings = ProviderPackage.Settings &
    readonly baseURL?: string
    readonly location?: string
    readonly project?: string
-    readonly providerOptions?: GeminiProviderOptionsInput
  }

 const fromRequest = Effect.fn("GoogleVertex.fromRequest")(function* (request: LLMRequest) {
@@ -93,7 +93,10 @@ const configuredRoute = (input: Config, modelID: string | ModelID) => {
  const apiKey = GoogleVertexShared.apiKey(input)
  const endpointModel = String(modelID).startsWith("endpoints/")
  if (apiKey !== undefined && endpointModel)
-    throw new Error("Google Vertex tuned models do not support Express Mode API keys")
+    throw new ProviderConfigurationError({
+      provider: id,
+      message: "Google Vertex tuned models do not support Express Mode API keys",
+    })
  const location = GoogleVertexShared.location(inputLocation, "us-central1")
  const project = GoogleVertexShared.project(inputProject)
  const endpoint =
@@ -121,16 +124,22 @@ export const provider = {
  id,
  configure,
 }
-export const model: ProviderPackage.Definition<Settings, GeminiProviderOptionsInput>["model"] = (modelID, settings) => {
-  if (settings.apiKey !== undefined && settings.accessToken !== undefined)
-    throw new Error("Google Vertex apiKey cannot be combined with accessToken or auth")
+export const model: ProviderPackage.Definition<Settings, GeminiProviderOptionsInput>["model"] = (
+  modelID,
+  { accessToken, apiKey, baseURL, body, headers, location, project, ...providerOptions },
+) => {
+  if (apiKey !== undefined && accessToken !== undefined)
+    throw new ProviderConfigurationError({
+      provider: id,
+      message: "Google Vertex apiKey cannot be combined with accessToken or auth",
+    })
  return configure({
-    ...(settings.apiKey === undefined ? { accessToken: settings.accessToken } : { apiKey: settings.apiKey }),
-    baseURL: settings.baseURL,
-    headers: settings.headers === undefined ? undefined : { ...settings.headers },
-    http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    location: settings.location,
-    project: settings.project,
-    providerOptions: settings.providerOptions,
+    ...(apiKey === undefined ? { accessToken: accessToken } : { apiKey: apiKey }),
+    baseURL,
+    headers: headers === undefined ? undefined : { ...headers },
+    http: body === undefined ? undefined : { body: { ...body } },
+    location,
+    project,
+    providerOptions,
  }).model(modelID)
 }
@@ -20,11 +20,11 @@ export type Config = RouteDefaultsInput &
    readonly providerOptions?: Gemini.ProviderOptionsInput
  }

-export interface Settings extends ProviderPackage.Settings {
-  readonly apiKey?: string
-  readonly baseURL?: string
-  readonly providerOptions?: Gemini.ProviderOptionsInput
-}
+export type Settings = ProviderPackage.Settings &
+  Gemini.ProviderOptionsInput & {
+    readonly apiKey?: string
+    readonly baseURL?: string
+  }

 const auth = (options: ProviderAuthOption<"optional">) => {
  if ("auth" in options && options.auth) return options.auth
@@ -57,13 +57,16 @@ export const configure = (input: Config = {}) => {
 }

 export const provider = configure()
-export const model: ProviderPackage.Definition<Settings, Gemini.ProviderOptionsInput>["model"] = (modelID, settings) =>
+export const model: ProviderPackage.Definition<Settings, Gemini.ProviderOptionsInput>["model"] = (
+  modelID,
+  { apiKey, baseURL, body, headers, ...providerOptions },
+) =>
  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,
+    apiKey,
+    baseURL,
+    headers: headers === undefined ? undefined : { ...headers },
+    http: body === undefined ? undefined : { body: { ...body } },
+    providerOptions,
  }).model(modelID)

 export const image = provider.image
@@ -26,11 +26,11 @@ export type LanguageModelOptions = Omit<RouteDefaultsInput, "providerOptions"> &
    readonly providerOptions?: ProviderOptions
  }

-export interface Settings extends ProviderPackage.Settings {
-  readonly apiKey?: string
-  readonly baseURL?: string
-  readonly providerOptions?: ProviderOptions
-}
+export type Settings = ProviderPackage.Settings &
+  ProviderOptions & {
+    readonly apiKey?: string
+    readonly baseURL?: string
+  }

 const Options = Schema.Struct({
  includeReasoning: Schema.optional(Schema.Boolean),
@@ -103,13 +103,16 @@ export const configure = (input: LanguageModelOptions = {}) => {

 export const provider = configure()

-export const model: ProviderPackage.Definition<Settings, ProviderOptions>["model"] = (modelID, settings) =>
+export const model: ProviderPackage.Definition<Settings, ProviderOptions>["model"] = (
+  modelID,
+  { apiKey, baseURL, body, headers, ...providerOptions },
+) =>
  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,
+    apiKey,
+    baseURL,
+    headers: headers === undefined ? undefined : { ...headers },
+    http: body === undefined ? undefined : { body: { ...body } },
+    providerOptions,
  }).model(modelID)

 export * as Groq from "./groq.js"
@@ -79,11 +79,11 @@ export type LanguageModelOptions = Omit<RouteDefaultsInput, "providerOptions"> &
    readonly providerOptions?: ProviderOptionsInput
  }

-export interface Settings extends ProviderPackage.Settings {
-  readonly apiKey?: string
-  readonly baseURL?: string
-  readonly providerOptions?: ProviderOptionsInput
-}
+export type Settings = ProviderPackage.Settings &
+  ProviderOptionsInput & {
+    readonly apiKey?: string
+    readonly baseURL?: string
+  }

 const responsesRoute = Route.make({
  id: "meta-responses",
@@ -169,13 +169,13 @@ export const chatModel: ProviderPackage.Definition<Settings, OpenResponsesProvid
 export const messagesModel: ProviderPackage.Definition<Settings, MessagesOptionsInput>["model"] = (modelID, settings) =>
  fromSettings(settings).messages(modelID)

-function fromSettings(settings: Settings) {
+function fromSettings({ apiKey, baseURL, body, headers, ...providerOptions }: Settings) {
  return configure({
-    apiKey: settings.apiKey,
-    baseURL: settings.baseURL,
-    headers: settings.headers,
-    http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    providerOptions: settings.providerOptions,
+    apiKey,
+    baseURL,
+    headers,
+    http: body === undefined ? undefined : { body: { ...body } },
+    providerOptions,
  })
 }

@@ -40,11 +40,11 @@ export type Config = Omit<RouteDefaultsInput, "providerOptions"> &
    readonly providerOptions?: ProviderOptionsInput
  }

-export interface Settings<Options = MessagesOptionsInput> extends ProviderPackage.Settings {
-  readonly apiKey?: string
-  readonly baseURL?: string
-  readonly providerOptions?: Options
-}
+export type Settings<Options = MessagesOptionsInput> = ProviderPackage.Settings &
+  Options & {
+    readonly apiKey?: string
+    readonly baseURL?: string
+  }

 const ChatOptions = Schema.Struct({
  thinking: Schema.optional(Schema.Struct({ type: Schema.String })),
@@ -127,14 +127,14 @@ export const provider = configure()

 export const model: ProviderPackage.Definition<Settings<MessagesOptionsInput>, MessagesOptionsInput>["model"] = (
  modelID,
-  settings,
+  { apiKey, baseURL, body, headers, ...providerOptions },
 ) =>
  configure({
-    apiKey: settings.apiKey,
-    baseURL: settings.baseURL,
-    headers: settings.headers,
-    http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    providerOptions: settings.providerOptions,
+    apiKey,
+    baseURL,
+    headers,
+    http: body === undefined ? undefined : { body: { ...body } },
+    providerOptions,
  }).model(modelID)

 export const messages = provider.messages
@@ -3,11 +3,14 @@ import { MiniMax } from "../minimax.js"

 export type Settings = MiniMax.Settings<MiniMax.ChatOptionsInput>

-export const model: ProviderPackage.Definition<Settings, MiniMax.ChatOptionsInput>["model"] = (modelID, settings) =>
+export const model: ProviderPackage.Definition<Settings, MiniMax.ChatOptionsInput>["model"] = (
+  modelID,
+  { apiKey, baseURL, body, headers, ...providerOptions },
+) =>
  MiniMax.configure({
-    apiKey: settings.apiKey,
-    baseURL: settings.baseURL,
-    headers: settings.headers,
-    http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    providerOptions: settings.providerOptions,
+    apiKey,
+    baseURL,
+    headers,
+    http: body === undefined ? undefined : { body: { ...body } },
+    providerOptions,
  }).chat(modelID)
@@ -5,12 +5,12 @@ export type Settings = MiniMax.Settings<MiniMax.ResponsesOptionsInput>

 export const model: ProviderPackage.Definition<Settings, MiniMax.ResponsesOptionsInput>["model"] = (
  modelID,
-  settings,
+  { apiKey, baseURL, body, headers, ...providerOptions },
 ) =>
  MiniMax.configure({
-    apiKey: settings.apiKey,
-    baseURL: settings.baseURL,
-    headers: settings.headers,
-    http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    providerOptions: settings.providerOptions,
+    apiKey,
+    baseURL,
+    headers,
+    http: body === undefined ? undefined : { body: { ...body } },
+    providerOptions,
  }).responses(modelID)
@@ -14,11 +14,11 @@ export type LanguageModelOptions = Omit<RouteDefaultsInput, "providerOptions"> &
    readonly providerOptions?: ProviderOptions
  }

-export interface Settings extends ProviderPackage.Settings {
-  readonly apiKey?: string
-  readonly baseURL?: string
-  readonly providerOptions?: ProviderOptions
-}
+export type Settings = ProviderPackage.Settings &
+  ProviderOptions & {
+    readonly apiKey?: string
+    readonly baseURL?: string
+  }

 export const route = MistralChat.route
 export const routes = [route]
@@ -39,13 +39,16 @@ export const configure = (input: LanguageModelOptions = {}) => {

 export const provider = configure()

-export const model: ProviderPackage.Definition<Settings, ProviderOptions>["model"] = (modelID, settings) =>
+export const model: ProviderPackage.Definition<Settings, ProviderOptions>["model"] = (
+  modelID,
+  { apiKey, baseURL, body, headers, ...providerOptions },
+) =>
  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,
+    apiKey,
+    baseURL,
+    headers: headers === undefined ? undefined : { ...headers },
+    http: body === undefined ? undefined : { body: { ...body } },
+    providerOptions,
  }).model(modelID)

 export * as Mistral from "./mistral.js"
@@ -42,11 +42,11 @@ export type Config = Omit<RouteDefaultsInput, "providerOptions"> &
    readonly providerOptions?: ChatOptionsInput | MessagesOptionsInput | ResponsesOptionsInput
  }

-export interface Settings<Options = ChatOptionsInput> extends ProviderPackage.Settings {
-  readonly apiKey?: string
-  readonly baseURL?: string
-  readonly providerOptions?: Options
-}
+export type Settings<Options = ChatOptionsInput> = ProviderPackage.Settings &
+  Options & {
+    readonly apiKey?: string
+    readonly baseURL?: string
+  }

 const ChatOptions = Schema.Struct({
  reasoningEffort: Schema.optional(Schema.String),
@@ -133,13 +133,16 @@ export const chat = provider.chat
 export const messages = provider.messages
 export const responses = provider.responses

-export const model: ProviderPackage.Definition<Settings, ChatOptionsInput>["model"] = (modelID, settings) =>
+export const model: ProviderPackage.Definition<Settings, ChatOptionsInput>["model"] = (
+  modelID,
+  { apiKey, baseURL, body, headers, ...providerOptions },
+) =>
  configure({
-    apiKey: settings.apiKey,
-    baseURL: settings.baseURL,
-    headers: settings.headers,
-    http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    providerOptions: settings.providerOptions,
+    apiKey,
+    baseURL,
+    headers,
+    http: body === undefined ? undefined : { body: { ...body } },
+    providerOptions,
  }).model(modelID)

 export * as Moonshot from "./moonshot.js"
@@ -5,12 +5,12 @@ export type Settings = Moonshot.Settings<Moonshot.MessagesOptionsInput>

 export const model: ProviderPackage.Definition<Settings, Moonshot.MessagesOptionsInput>["model"] = (
  modelID,
-  settings,
+  { apiKey, baseURL, body, headers, ...providerOptions },
 ) =>
  Moonshot.configure({
-    apiKey: settings.apiKey,
-    baseURL: settings.baseURL,
-    headers: settings.headers,
-    http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    providerOptions: settings.providerOptions,
+    apiKey,
+    baseURL,
+    headers,
+    http: body === undefined ? undefined : { body: { ...body } },
+    providerOptions,
  }).messages(modelID)
@@ -5,12 +5,12 @@ export type Settings = Moonshot.Settings<Moonshot.ResponsesOptionsInput>

 export const model: ProviderPackage.Definition<Settings, Moonshot.ResponsesOptionsInput>["model"] = (
  modelID,
-  settings,
+  { apiKey, baseURL, body, headers, ...providerOptions },
 ) =>
  Moonshot.configure({
-    apiKey: settings.apiKey,
-    baseURL: settings.baseURL,
-    headers: settings.headers,
-    http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    providerOptions: settings.providerOptions,
+    apiKey,
+    baseURL,
+    headers,
+    http: body === undefined ? undefined : { body: { ...body } },
+    providerOptions,
  }).responses(modelID)
@@ -16,12 +16,12 @@ export type Config = RouteDefaultsInput &
    readonly providerOptions?: OpenResponsesProviderOptionsInput
  }

-export interface Settings extends ProviderPackage.Settings {
-  readonly apiKey?: string
-  readonly baseURL: string
-  readonly provider?: string
-  readonly providerOptions?: OpenResponsesProviderOptionsInput
-}
+export type Settings = ProviderPackage.Settings &
+  OpenResponsesProviderOptionsInput & {
+    readonly apiKey?: string
+    readonly baseURL: string
+    readonly provider?: string
+  }

 export const routes = [OpenAICompatibleResponses.route]

@@ -48,13 +48,13 @@ export const provider = {

 export const model: ProviderPackage.Definition<Settings, OpenResponsesProviderOptionsInput>["model"] = (
  modelID,
-  settings,
+  { apiKey, baseURL, body, headers, provider, ...providerOptions },
 ) =>
  configure({
-    apiKey: settings.apiKey,
-    baseURL: settings.baseURL,
-    headers: settings.headers === undefined ? undefined : { ...settings.headers },
-    http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    provider: settings.provider,
-    providerOptions: settings.providerOptions,
+    apiKey,
+    baseURL,
+    headers: headers === undefined ? undefined : { ...headers },
+    http: body === undefined ? undefined : { body: { ...body } },
+    provider,
+    providerOptions,
  }).model(modelID)
@@ -14,12 +14,12 @@ type GenericModelOptions = Omit<RouteDefaultsInput, "providerOptions"> &
    readonly providerOptions?: OpenAIProviderOptionsInput
  }

-export interface Settings extends ProviderPackage.Settings {
-  readonly apiKey?: string
-  readonly baseURL: string
-  readonly provider?: string
-  readonly providerOptions?: OpenAIProviderOptionsInput
-}
+export type Settings = ProviderPackage.Settings &
+  OpenAIProviderOptionsInput & {
+    readonly apiKey?: string
+    readonly baseURL: string
+    readonly provider?: string
+  }

 export const routes = [OpenAICompatibleChat.route]

@@ -45,14 +45,17 @@ export const provider = {
  configure,
 }

-export const model: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (modelID, settings) =>
+export const model: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (
+  modelID,
+  { apiKey, baseURL, body, headers, provider, ...providerOptions },
+) =>
  configure({
-    apiKey: settings.apiKey,
-    baseURL: settings.baseURL,
-    headers: settings.headers === undefined ? undefined : { ...settings.headers },
-    http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    provider: settings.provider,
-    providerOptions: settings.providerOptions,
+    apiKey,
+    baseURL,
+    headers: headers === undefined ? undefined : { ...headers },
+    http: body === undefined ? undefined : { body: { ...body } },
+    provider,
+    providerOptions,
  }).model(modelID)

 export * as OpenAICompatible from "./openai-compatible.js"
@@ -57,14 +57,14 @@ export const imageGeneration = (options: ImageGenerationOptions = {}) =>
    },
  })

-export interface Settings extends ProviderPackage.Settings {
-  readonly apiKey?: string
-  readonly baseURL?: string
-  readonly organization?: string
-  readonly project?: string
-  readonly queryParams?: Readonly<Record<string, string>>
-  readonly providerOptions?: OpenAIProviderOptionsInput
-}
+export type Settings = ProviderPackage.Settings &
+  OpenAIProviderOptionsInput & {
+    readonly apiKey?: string
+    readonly baseURL?: string
+    readonly organization?: string
+    readonly project?: string
+    readonly queryParams?: Readonly<Record<string, string>>
+  }

 const auth = (options: ProviderAuthOption<"optional">) => AuthOptions.bearer(options, "OPENAI_API_KEY")

@@ -116,19 +116,28 @@ export const configure = (input: Config = {}) => {

 export const provider = configure()

-const config = (settings: Settings): Config => {
+const config = ({
+  apiKey,
+  baseURL,
+  body,
+  headers: given,
+  organization,
+  project,
+  queryParams,
+  ...providerOptions
+}: Settings): Config => {
  const headers = {
-    ...(settings.organization === undefined ? {} : { "OpenAI-Organization": settings.organization }),
-    ...(settings.project === undefined ? {} : { "OpenAI-Project": settings.project }),
-    ...settings.headers,
+    ...(organization === undefined ? {} : { "OpenAI-Organization": organization }),
+    ...(project === undefined ? {} : { "OpenAI-Project": project }),
+    ...given,
  }
  return {
-    apiKey: settings.apiKey,
-    baseURL: settings.baseURL,
+    apiKey,
+    baseURL,
    headers: Object.keys(headers).length === 0 ? undefined : headers,
-    http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    providerOptions: settings.providerOptions,
-    queryParams: settings.queryParams === undefined ? undefined : { ...settings.queryParams },
+    http: body === undefined ? undefined : { body: { ...body } },
+    providerOptions,
+    queryParams: queryParams === undefined ? undefined : { ...queryParams },
  }
 }

@@ -77,11 +77,11 @@ export type LanguageModelOptions = Omit<RouteDefaultsInput, "providerOptions"> &
    readonly providerOptions?: OpenRouterProviderOptionsInput
  }

-export interface Settings extends ProviderPackage.Settings {
-  readonly apiKey?: string
-  readonly baseURL?: string
-  readonly providerOptions?: OpenRouterProviderOptionsInput
-}
+export type Settings = ProviderPackage.Settings &
+  OpenRouterProviderOptionsInput & {
+    readonly apiKey?: string
+    readonly baseURL?: string
+  }

 const OpenRouterBody = Schema.StructWithRest(Schema.Struct(OpenAIChat.bodyFields), [
  Schema.Record(Schema.String, Schema.Any),
@@ -191,12 +191,12 @@ export const configure = (input: LanguageModelOptions = {}) => {
 export const provider = configure()
 export const model: ProviderPackage.Definition<Settings, OpenRouterProviderOptionsInput>["model"] = (
  modelID,
-  settings,
+  { apiKey, baseURL, body, headers, ...providerOptions },
 ) =>
  configure({
-    apiKey: settings.apiKey,
-    baseURL: settings.baseURL,
-    headers: settings.headers,
-    http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    providerOptions: settings.providerOptions,
+    apiKey,
+    baseURL,
+    headers,
+    http: body === undefined ? undefined : { body: { ...body } },
+    providerOptions,
  }).model(modelID)
@@ -15,11 +15,11 @@ export type LanguageModelOptions = Omit<RouteDefaultsInput, "providerOptions"> &
    readonly providerOptions?: OpenAIProviderOptionsInput
  }

-export interface Settings extends ProviderPackage.Settings {
-  readonly apiKey?: string
-  readonly baseURL?: string
-  readonly providerOptions?: OpenAIProviderOptionsInput
-}
+export type Settings = ProviderPackage.Settings &
+  OpenAIProviderOptionsInput & {
+    readonly apiKey?: string
+    readonly baseURL?: string
+  }

 export const route = Route.make({
  id: "togetherai-chat",
@@ -52,11 +52,14 @@ export const configure = (input: LanguageModelOptions = {}) => {

 export const provider = configure()

-export const model: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (modelID, settings) =>
+export const model: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (
+  modelID,
+  { apiKey, baseURL, body, headers, ...providerOptions },
+) =>
  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,
+    apiKey,
+    baseURL,
+    headers: headers === undefined ? undefined : { ...headers },
+    http: body === undefined ? undefined : { body: { ...body } },
+    providerOptions,
  }).model(modelID)
@@ -20,11 +20,11 @@ export type LanguageModelOptions = Omit<RouteDefaultsInput, "providerOptions"> &
    readonly providerOptions?: XAIProviderOptionsInput
  }

-export interface Settings extends ProviderPackage.Settings {
-  readonly apiKey?: string
-  readonly baseURL?: string
-  readonly providerOptions?: XAIProviderOptionsInput
-}
+export type Settings = ProviderPackage.Settings &
+  XAIProviderOptionsInput & {
+    readonly apiKey?: string
+    readonly baseURL?: string
+  }

 export type { XAIImageOptions } from "../protocols/xai-images.js"

@@ -41,6 +41,10 @@ const responsesRoute = Route.make({
    id: "openai-responses",
    name: "xAI Responses",
    rotateAfterMs: RESPONSES_WEBSOCKET_ROTATE_AFTER_MS,
+    // xAI continues a chain only from stored responses: with `store: false` (the route default) `previous_response_id`
+    // fails with "Response with id=… not found", so those steps are sent in full over the reused connection. It also
+    // rejects `instructions` next to `previous_response_id` and keeps the instructions of the response it continues.
+    continuation: ({ instructions: _instructions, ...request }) => (request.store === false ? undefined : request),
  }),
  defaults: { providerOptions: { store: false, include: ["reasoning.encrypted_content"] } },
 })
@@ -106,13 +110,13 @@ export const model: ProviderPackage.Definition<
  Settings,
  XAIProviderOptionsInput,
  typeof responsesRoute.compact
->["model"] = (modelID, settings) =>
+>["model"] = (modelID, { apiKey, baseURL, body, headers, ...providerOptions }) =>
  configure({
-    apiKey: settings.apiKey,
-    baseURL: settings.baseURL,
-    headers: settings.headers,
-    http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    providerOptions: settings.providerOptions,
+    apiKey,
+    baseURL,
+    headers,
+    http: body === undefined ? undefined : { body: { ...body } },
+    providerOptions,
  }).model(modelID)
 export const responses = provider.responses
 export const chat = provider.chat
@@ -23,11 +23,11 @@ export type Config = Omit<RouteDefaultsInput, "providerOptions"> &
    readonly providerOptions?: ChatOptionsInput | MessagesOptionsInput | ResponsesOptionsInput
  }

-export interface Settings<Options = ChatOptionsInput> extends ProviderPackage.Settings {
-  readonly apiKey?: string
-  readonly baseURL?: string
-  readonly providerOptions?: Options
-}
+export type Settings<Options = ChatOptionsInput> = ProviderPackage.Settings &
+  Options & {
+    readonly apiKey?: string
+    readonly baseURL?: string
+  }

 const chatRoute = Route.make({
  id: "zai-coding-chat",
@@ -80,13 +80,16 @@ export const chat = provider.chat
 export const messages = provider.messages
 export const responses = provider.responses

-export const model: ProviderPackage.Definition<Settings, ChatOptionsInput>["model"] = (modelID, settings) =>
+export const model: ProviderPackage.Definition<Settings, ChatOptionsInput>["model"] = (
+  modelID,
+  { apiKey, baseURL, body, headers, ...providerOptions },
+) =>
  configure({
-    apiKey: settings.apiKey,
-    baseURL: settings.baseURL,
-    headers: settings.headers,
-    http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    providerOptions: settings.providerOptions,
+    apiKey,
+    baseURL,
+    headers,
+    http: body === undefined ? undefined : { body: { ...body } },
+    providerOptions,
  }).model(modelID)

 export * as ZAICodingPlan from "./zai-coding-plan.js"
@@ -5,12 +5,12 @@ export type Settings = ZAICodingPlan.Settings<ZAICodingPlan.MessagesOptionsInput

 export const model: ProviderPackage.Definition<Settings, ZAICodingPlan.MessagesOptionsInput>["model"] = (
  modelID,
-  settings,
+  { apiKey, baseURL, body, headers, ...providerOptions },
 ) =>
  ZAICodingPlan.configure({
-    apiKey: settings.apiKey,
-    baseURL: settings.baseURL,
-    headers: settings.headers,
-    http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    providerOptions: settings.providerOptions,
+    apiKey,
+    baseURL,
+    headers,
+    http: body === undefined ? undefined : { body: { ...body } },
+    providerOptions,
  }).messages(modelID)
@@ -5,12 +5,12 @@ export type Settings = ZAICodingPlan.Settings<ZAICodingPlan.ResponsesOptionsInpu

 export const model: ProviderPackage.Definition<Settings, ZAICodingPlan.ResponsesOptionsInput>["model"] = (
  modelID,
-  settings,
+  { apiKey, baseURL, body, headers, ...providerOptions },
 ) =>
  ZAICodingPlan.configure({
-    apiKey: settings.apiKey,
-    baseURL: settings.baseURL,
-    headers: settings.headers,
-    http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    providerOptions: settings.providerOptions,
+    apiKey,
+    baseURL,
+    headers,
+    http: body === undefined ? undefined : { body: { ...body } },
+    providerOptions,
  }).responses(modelID)
@@ -17,11 +17,11 @@ export type Config = Omit<RouteDefaultsInput, "providerOptions"> &
    readonly providerOptions?: ChatOptionsInput
  }

-export interface Settings extends ProviderPackage.Settings {
-  readonly apiKey?: string
-  readonly baseURL?: string
-  readonly providerOptions?: ChatOptionsInput
-}
+export type Settings = ProviderPackage.Settings &
+  ChatOptionsInput & {
+    readonly apiKey?: string
+    readonly baseURL?: string
+  }

 export type { ZAIImageOptions } from "../protocols/zai-images.js"

@@ -70,13 +70,16 @@ export const provider = configure()
 export const image = provider.image
 export const chat = provider.chat

-export const model: ProviderPackage.Definition<Settings, ChatOptionsInput>["model"] = (modelID, settings) =>
+export const model: ProviderPackage.Definition<Settings, ChatOptionsInput>["model"] = (
+  modelID,
+  { apiKey, baseURL, body, headers, ...providerOptions },
+) =>
  configure({
-    apiKey: settings.apiKey,
-    baseURL: settings.baseURL,
-    headers: settings.headers,
-    http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    providerOptions: settings.providerOptions,
+    apiKey,
+    baseURL,
+    headers,
+    http: body === undefined ? undefined : { body: { ...body } },
+    providerOptions,
  }).model(modelID)

 export * as ZAI from "./zai.js"
@@ -23,6 +23,7 @@ import {
  LanguageModel,
  LLMEvent,
  InvalidProviderOutputError,
+  ProviderConfigurationError,
  ProviderID,
  mergeGenerationOptions,
  mergeHttpOptions,
@@ -128,7 +129,10 @@ const makeRouteLanguageModel = <Options extends ProviderOptions, Compact extends
  const provider = route.provider ?? ("provider" in mapped ? mapped.provider : undefined)
  if (!provider) throw new Error(`Route.model(${route.id}) requires a provider`)
  if (!endpointBaseURL(route.endpoint))
-    throw new Error(`Route.model(${route.id}) requires an endpoint baseURL — configure it on the route first`)
+    throw new ProviderConfigurationError({
+      provider: ProviderID.make(provider),
+      message: `Route.model(${route.id}) requires an endpoint baseURL — configure it on the route first`,
+    })
  return LanguageModel.make<Options, Compact>({
    ...mapped,
    provider,
@@ -50,6 +50,19 @@ export class UnsupportedOperationError extends Schema.TaggedError<UnsupportedOpe
  route: Schema.optional(RouteID),
 }) {}

+/**
+ * Provider settings that are missing, conflicting, or unsupported, such as
+ * Azure without `resourceName` or `baseURL`. Thrown synchronously while a
+ * provider facade or package entrypoint configures a model, before any
+ * request exists, so it is not an `AIError` reason.
+ */
+export class ProviderConfigurationError extends Schema.TaggedError<ProviderConfigurationError>(
+  "AI.Error.ProviderConfiguration",
+)("ProviderConfiguration", {
+  provider: ProviderID,
+  message: Schema.String,
+}) {}
+
 export class NoRouteError extends Schema.TaggedError<NoRouteError>("AI.Error.NoRoute")("NoRoute", {
  ...ReasonFields,
  route: RouteID,
@@ -3,6 +3,9 @@ import { model } from "@opencode/ai/providers/openai"
 import { LLM } from "../src/index.js"
 import { Endpoint } from "../src/route/endpoint.js"

+const configuration = (provider: string, message: string) =>
+  expect.objectContaining({ _tag: "ProviderConfiguration", provider, message })
+
 describe("provider package entrypoints", () => {
  test("semantic API aliases expose the same contract", async () => {
    const modules = await Promise.all([
@@ -185,13 +188,13 @@ describe("provider package entrypoints", () => {
      baseURL: "https://provider.example.test/v1/",
      headers: { "x-application": "opencode" },
      body: { service_tier: "priority" },
-      providerOptions: { reasoningEffort: "high" as const },
+      reasoningEffort: "high" as const,
    }
    const deepinfra = DeepInfra.model("google/gemma-3-27b-it", settings)

    expect(deepinfra.route.id).toBe("deepinfra-chat")
    expect(deepinfra.route.endpoint.baseURL).toBe("https://provider.example.test/v1/openai")
-    expect(deepinfra.route.defaults.providerOptions).toEqual(settings.providerOptions)
+    expect(deepinfra.route.defaults.providerOptions).toEqual({ reasoningEffort: "high" })
    expect(deepinfra.route.defaults.headers).toEqual(settings.headers)
    expect(deepinfra.route.defaults.http?.body).toEqual(settings.body)
  })
@@ -207,7 +210,7 @@ describe("provider package entrypoints", () => {
        apiKey: "fixture",
        headers: { "x-application": "opencode" },
        body: { custom: true },
-        providerOptions: { reasoningEffort: "high" },
+        reasoningEffort: "high",
      })
      expect(selected.provider).toBe(provider.id)
      expect(selected.route.endpoint.baseURL).toBe(provider.baseURL({ accountId: "account" }))
@@ -228,11 +231,11 @@ describe("provider package entrypoints", () => {
    }
    const openrouter = OpenRouter.model("anthropic/claude-sonnet-4", {
      ...settings,
-      providerOptions: { usage: true },
+      usage: true,
    })
    const xai = XAI.model("grok-4", {
      ...settings,
-      providerOptions: { reasoningEffort: "high" },
+      reasoningEffort: "high",
    })

    for (const selected of [openrouter, xai]) {
@@ -266,7 +269,8 @@ describe("provider package entrypoints", () => {
      provider: "example",
      headers: { "x-application": "opencode" },
      body: { service_tier: "priority" },
-      providerOptions: { reasoningEffort: "low", store: true },
+      reasoningEffort: "low",
+      store: true,
    })

    expect(String(selected.provider)).toBe("example")
@@ -292,7 +296,7 @@ describe("provider package entrypoints", () => {
      provider: "example",
      headers: { "x-application": "opencode" },
      body: { metadata: { user_id: "user_1" } },
-      providerOptions: { effort: "low" },
+      effort: "low",
    })

    expect(String(selected.provider)).toBe("example")
@@ -312,7 +316,7 @@ describe("provider package entrypoints", () => {
    const Anthropic = await import("@opencode/ai/providers/anthropic")
    const selected = Anthropic.model("claude-sonnet-4-6", {
      apiKey: "fixture",
-      providerOptions: { thinking: { type: "adaptive" } },
+      thinking: { type: "adaptive" },
    })

    expect(selected.route.defaults.providerOptions).toEqual({ thinking: { type: "adaptive" } })
@@ -322,7 +326,7 @@ describe("provider package entrypoints", () => {
    const AnthropicCompatible = await import("@opencode/ai/providers/anthropic-compatible")
    expect(() =>
      Reflect.apply(AnthropicCompatible.model, undefined, ["compatible-model", { apiKey: "fixture" }]),
-    ).toThrow("Anthropic-compatible providers require a baseURL")
+    ).toThrow(configuration("anthropic-compatible", "Anthropic-compatible providers require a baseURL"))
  })

  test("rejects conflicting Anthropic-compatible auth settings at runtime", async () => {
@@ -337,10 +341,10 @@ describe("provider package entrypoints", () => {
          baseURL: "https://messages.example.test/v1",
        },
      ]),
-    ).toThrow("Anthropic-compatible apiKey cannot be combined with authToken")
+    ).toThrow(configuration("anthropic-compatible", "Anthropic-compatible apiKey cannot be combined with authToken"))
    expect(() =>
      Reflect.apply(Anthropic.model, undefined, ["claude-sonnet-4-6", { apiKey: "fixture", authToken: "token" }]),
-    ).toThrow("Anthropic apiKey cannot be combined with authToken")
+    ).toThrow(configuration("anthropic", "Anthropic apiKey cannot be combined with authToken"))
  })

  test("maps legacy OpenAI organization and project settings to headers", () => {
@@ -406,7 +410,7 @@ describe("provider package entrypoints", () => {
      baseURL: "https://generativelanguage.test/v1beta",
      headers: { "x-application": "opencode" },
      body: { safetySettings: [] },
-      providerOptions: { thinkingConfig: { thinkingBudget: 1_024 } },
+      thinkingConfig: { thinkingBudget: 1_024 },
    })

    expect(selected.route.id).toBe("gemini")
@@ -490,43 +494,45 @@ describe("provider package entrypoints", () => {
        "gemini-3.5-flash",
        { accessToken: "token", apiKey: "fixture", project: "vertex-project" },
      ]),
-    ).toThrow("Google Vertex apiKey cannot be combined with accessToken or auth")
+    ).toThrow(configuration("google-vertex", "Google Vertex apiKey cannot be combined with accessToken or auth"))
    const configured = Reflect.apply(GoogleVertex.configure, undefined, [
      { accessToken: "token", auth: {}, project: "vertex-project" },
    ])
-    expect(() => configured.model("gemini-3.5-flash")).toThrow("Google Vertex accessToken cannot be combined with auth")
+    expect(() => configured.model("gemini-3.5-flash")).toThrow(
+      configuration("google-vertex", "Google Vertex accessToken cannot be combined with auth"),
+    )
    expect(() =>
      Reflect.apply(GoogleVertexMessages.model, undefined, [
        "claude-sonnet-4-6",
        { apiKey: "fixture", project: "vertex-project" },
      ]),
-    ).toThrow("Google Vertex Messages does not support API keys")
+    ).toThrow(configuration("google-vertex", "Google Vertex Messages does not support API keys"))
    expect(() =>
      Reflect.apply(Providers.GoogleVertexMessages.configure, undefined, [
        { apiKey: "fixture", project: "vertex-project" },
      ]),
-    ).toThrow("Google Vertex Messages does not support API keys")
+    ).toThrow(configuration("google-vertex", "Google Vertex Messages does not support API keys"))
    expect(() =>
      Reflect.apply(GoogleVertexChat.model, undefined, [
        "deepseek-ai/deepseek-v3.2-maas",
        { apiKey: "fixture", project: "vertex-project" },
      ]),
-    ).toThrow("Google Vertex Chat does not support API keys")
+    ).toThrow(configuration("google-vertex", "Google Vertex Chat does not support API keys"))
    expect(() =>
      Reflect.apply(Providers.GoogleVertexChat.configure, undefined, [
        { apiKey: "fixture", project: "vertex-project" },
      ]),
-    ).toThrow("Google Vertex Chat does not support API keys")
+    ).toThrow(configuration("google-vertex", "Google Vertex Chat does not support API keys"))
    expect(() =>
      Reflect.apply(GoogleVertexResponses.model, undefined, [
        "xai/grok-4.20-reasoning",
        { apiKey: "fixture", project: "vertex-project" },
      ]),
-    ).toThrow("Google Vertex Responses does not support API keys")
+    ).toThrow(configuration("google-vertex", "Google Vertex Responses does not support API keys"))
    expect(() =>
      Reflect.apply(Providers.GoogleVertexResponses.configure, undefined, [
        { apiKey: "fixture", project: "vertex-project" },
      ]),
-    ).toThrow("Google Vertex Responses does not support API keys")
+    ).toThrow(configuration("google-vertex", "Google Vertex Responses does not support API keys"))
  })
 })
@@ -76,7 +76,13 @@ it.effect("Alibaba owns regional shared and workspace-specific endpoints", () =>

 test("Alibaba requires explicit placement and supports complete base URL overrides", () => {
  for (const region of ["eu-central-1", "ap-northeast-1", "future-region"])
-    expect(() => Alibaba.configure({ region })).toThrow("requires workspaceID or baseURL")
+    expect(() => Alibaba.configure({ region })).toThrow(
+      expect.objectContaining({
+        _tag: "ProviderConfiguration",
+        provider: "alibaba",
+        message: `Alibaba region ${region} requires workspaceID or baseURL`,
+      }),
+    )
  for (const config of [
    { baseURL: "https://gateway.example/prefix" },
    { region: "future-region", workspaceID: "ignored", baseURL: "https://gateway.example/prefix" },
@@ -69,7 +69,9 @@ for (const model of [
      dynamicResponse(({ request, text, respond }) =>
        Effect.sync(() => {
          const body = JSON.parse(text)
-          expect(request.headers["anthropic-beta"]).toBe("existing-beta,compact-2026-01-12")
+          expect(request.headers["anthropic-beta"]).toBe(
+            "existing-beta,interleaved-thinking-2025-05-14,compact-2026-01-12",
+          )
          if (body.messages.length === 1) {
            expect(body.context_management.edits).toEqual([
              {
@@ -32,7 +32,9 @@ for (const [id, enabled] of [
        enabled ? { type: "adaptive", block_binding: { prefix_mismatch_behavior: "drop_block" } } : undefined,
      )
      expect(prepared.request.headers["anthropic-beta"]).toBe(
-        enabled ? "existing-beta,thinking-binding-controls-2026-08-01" : "existing-beta",
+        enabled
+          ? "existing-beta,interleaved-thinking-2025-05-14,thinking-binding-controls-2026-08-01"
+          : "existing-beta,interleaved-thinking-2025-05-14",
      )
    }),
  )
@@ -53,7 +55,9 @@ it.effect("preserves explicit thinking settings and combines required beta heade
      const prepared = yield* AnthropicMessages.route.prepareTransport(compiled.body, request)
      expect(compiled.body.thinking).toEqual(thinking)
      expect(prepared.request.headers["anthropic-beta"]).toBe(
-        thinking.type === "disabled" ? "compact-2026-01-12" : "compact-2026-01-12,thinking-binding-controls-2026-08-01",
+        thinking.type === "disabled"
+          ? "interleaved-thinking-2025-05-14,compact-2026-01-12"
+          : "interleaved-thinking-2025-05-14,compact-2026-01-12,thinking-binding-controls-2026-08-01",
      )
    }
  }),
@@ -1458,7 +1458,13 @@ describe("Bedrock Converse route", () => {
        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")
+      expect(() => AmazonBedrock.configure({ auth: "sigv4", apiKey: "k" })).toThrow(
+        expect.objectContaining({
+          _tag: "ProviderConfiguration",
+          provider: "amazon-bedrock",
+          message: "Amazon Bedrock SigV4 auth does not accept apiKey",
+        }),
+      )
    }).pipe(
      withProcessEnv({
        ...noAmbientAWS,
@@ -378,7 +378,11 @@ describe("Google Vertex providers", () => {

  test("rejects tuned Gemini models in express mode", () => {
    expect(() => GoogleVertex.configure({ apiKey: "fixture" }).model("endpoints/1234567890")).toThrow(
-      "Google Vertex tuned models do not support Express Mode API keys",
+      expect.objectContaining({
+        _tag: "ProviderConfiguration",
+        provider: "google-vertex",
+        message: "Google Vertex tuned models do not support Express Mode API keys",
+      }),
    )
  })
 })
@@ -34,12 +34,10 @@ it.effect("Groq lowers its own options for custom catalog identities and endpoin
        baseURL: "https://gateway.example/v1",
        headers: { "x-client": "test" },
        body: { custom: "value" },
-        providerOptions: {
-          reasoningEffort: "default",
-          parallelToolCalls: true,
-          serviceTier: "flex",
-          user: "test-user",
-        },
+        reasoningEffort: "default",
+        parallelToolCalls: true,
+        serviceTier: "flex",
+        user: "test-user",
      }),
      { provider: "custom-groq" },
    )
@@ -87,7 +87,7 @@ it.effect("Meta package selectors preserve overrides and Chat token policy on cu
        baseURL: "https://gateway.example/v1",
        headers: { "x-client": "test" },
        body: { custom: "value" },
-        providerOptions: { reasoningEffort: "future-effort" },
+        reasoningEffort: "future-effort",
      })
      expect(model.route.endpoint.baseURL).toBe("https://gateway.example/v1")
      expect(model.route.defaults.headers).toEqual({ "x-client": "test" })
@@ -99,7 +99,7 @@ describe("native OpenAI-compatible providers", () => {
        const settings = {
          apiKey: "fixture",
          baseURL: "https://gateway.example/v1",
-          providerOptions: { reasoningEffort: "high" },
+          reasoningEffort: "high",
        }
        const selected = provider.configure(settings).model("test-model")
        expect(selected.provider).toBe(provider.id)
@@ -143,7 +143,7 @@ describe("native OpenAI-compatible providers", () => {
          tools: [
            ToolDefinition.make({ name: "lookup", description: "Look up data", inputSchema: { type: "object" } }),
          ],
-          providerOptions: { store: true },
+          store: true,
        }),
      )

@@ -168,7 +168,7 @@ describe("native OpenAI-compatible providers", () => {
            ]),
            Message.user("Continue."),
          ],
-          providerOptions: { store: true },
+          store: true,
        }),
      )

@@ -206,7 +206,7 @@ describe("native OpenAI-compatible providers", () => {
        baseURL: "https://gateway.example/v1",
        headers: { "x-application": "opencode" },
        body: { service_tier: "priority" },
-        providerOptions: { reasoningEffort: "high" },
+        reasoningEffort: "high",
      })

      expect(selected.route.endpoint.baseURL).toBe("https://gateway.example/v1")
@@ -0,0 +1,78 @@
+import { expect } from "bun:test"
+import { Effect, Schema } from "effect"
+import { LLM, LLMClient } from "../../src/index.js"
+import { OpenResponses } from "../../src/protocols/open-responses.js"
+import { Meta } from "../../src/providers/index.js"
+import { configure } from "../../src/providers/openai-compatible-responses.js"
+import { it } from "../lib/effect.js"
+import { fixedResponse } from "../lib/http.js"
+import { sseEvents } from "../lib/sse.js"
+
+const decodeEvent = Schema.decodeUnknownEffect(OpenResponses.protocol.stream.event)
+
+it.effect("normalizes flat errors in shared SSE and WebSocket decoding", () =>
+  Effect.gen(function* () {
+    const frame = {
+      type: "error",
+      sequence_number: 4,
+      code: "server_shutting_down",
+      message: "Server is shutting down. Please retry your request.",
+      param: null,
+    }
+    for (const decode of [decodeEvent, OpenResponses.decodeChannelEvent]) {
+      const event = yield* decode(JSON.stringify(frame))
+      expect(event).toEqual({
+        type: "error",
+        sequence_number: 4,
+        error: { code: frame.code, message: frame.message, param: null },
+      })
+
+      for (const unchanged of [
+        event,
+        { type: "error" },
+        {
+          type: "response.failed",
+          response: { id: "resp_failed", error: { code: "server_error", message: "Internal server error" } },
+        },
+        { type: "response.output_text.delta", item_id: "msg_text", delta: "Hello" },
+      ]) {
+        expect(yield* decode(JSON.stringify(unchanged))).toEqual(unchanged)
+      }
+    }
+  }),
+)
+
+it.effect("continues to normalize untyped xAI WebSocket errors", () =>
+  Effect.gen(function* () {
+    const frame = { error: { type: "api_error", message: "gRPC error: Response with id=resp_missing not found" } }
+    expect(yield* OpenResponses.decodeChannelEvent(JSON.stringify(frame))).toEqual({ ...frame, type: "error" })
+  }),
+)
+
+it.effect("retains classification and original error bodies through Meta and generic Responses routes", () =>
+  Effect.gen(function* () {
+    const raw = `{
+  "type": "error",
+  "sequence_number": 4,
+  "code": "server_shutting_down",
+  "message": "Server is shutting down. Please retry your request.",
+  "param": null,
+  "diagnostic": "retain-original-frame"
+}`
+    for (const model of [
+      Meta.configure({ apiKey: "fixture" }).responses("muse-spark-1.3"),
+      configure({ apiKey: "fixture", provider: "gateway", baseURL: "https://responses.example.test/v1" }).model(
+        "example-model",
+      ),
+    ]) {
+      const error = yield* LLMClient.generate(LLM.request({ model, prompt: "Hello" })).pipe(
+        Effect.provide(fixedResponse(sseEvents(raw.replaceAll("\n", "\ndata: ")))),
+        Effect.flip,
+      )
+      expect(error.reason._tag).toBe("ProviderInternal")
+      expect(error.message).toBe("server_shutting_down: Server is shutting down. Please retry your request.")
+      expect(error.reason.body).toBe(raw)
+      expect(error.reason.http?.status).toBe(200)
+    }
+  }),
+)
@@ -90,7 +90,11 @@ const classifyingChannelDriver = (message: string): WebSocketChannelDriver => {
  }
 }

-const continuationDriver = (request: Readonly<Record<string, unknown>>, base = baseChannelDriver) => {
+const continuationDriver = (
+  request: Readonly<Record<string, unknown>>,
+  base = baseChannelDriver,
+  continuation?: OpenResponsesContinuation.Shape,
+) => {
  const message = ProviderShared.encodeJson(request)
  return OpenResponsesContinuation.driver({
    id: "openai-responses",
@@ -98,6 +102,7 @@ const continuationDriver = (request: Readonly<Record<string, unknown>>, base = b
    request,
    message,
    base: base(message),
+    continuation,
  })
 }

@@ -921,6 +926,58 @@ describe("OpenAI Responses route", () => {
        type: "provider-failure",
        error: { reason: { _tag: "InvalidRequest", classification: "context-overflow" } },
      })
+
+      // A retryable failure stays one: the runner retries it, and the transport has already dropped the
+      // checkpoint, so that retry is a full send. xAI reports every rejection this way.
+      const internal = ProviderShared.encodeJson({
+        type: "error",
+        error: { type: "api_error", message: "gRPC error: Response with id=resp_1 not found" },
+      })
+      expect(yield* second.observe(yield* second.create(saved), internal)).toMatchObject({
+        type: "provider-failure",
+        error: { reason: { _tag: "ProviderInternal" } },
+      })
+    }),
+  )
+
+  it.effect("shapes the incremental send with the route continuation", () =>
+    Effect.gen(function* () {
+      const firstRequest = {
+        type: "response.create",
+        model: "grok-4.6",
+        store: true,
+        instructions: "You are terse.",
+        input: [{ role: "user", content: [{ type: "input_text", text: "First" }] }],
+      }
+      const secondRequest = {
+        ...firstRequest,
+        input: [...firstRequest.input, { role: "user", content: [{ type: "input_text", text: "Second" }] }],
+      }
+      const saved = checkpoint(
+        yield* continuationDriver(firstRequest).observe(
+          yield* continuationDriver(firstRequest).create(undefined),
+          ProviderShared.encodeJson({ type: "response.completed", response: { id: "resp_1" } }),
+        ),
+      )
+
+      const trimmed = yield* continuationDriver(
+        secondRequest,
+        baseChannelDriver,
+        ({ instructions: _, ...rest }) => rest,
+      ).create(saved)
+      expect(trimmed.mode).toBe("incremental")
+      expect(JSON.parse(trimmed.message)).toEqual({
+        type: "response.create",
+        model: "grok-4.6",
+        store: true,
+        previous_response_id: "resp_1",
+        input: [{ role: "user", content: [{ type: "input_text", text: "Second" }] }],
+      })
+
+      // Declining the continuation sends the step in full and never sends a previous_response_id.
+      const declined = yield* continuationDriver(secondRequest, baseChannelDriver, () => undefined).create(saved)
+      expect(declined.mode).toBe("full")
+      expect(JSON.parse(declined.message)).toEqual(secondRequest)
    }),
  )

@@ -1,11 +1,18 @@
 import { describe, expect } from "bun:test"
-import { Effect } from "effect"
+import { Effect, Layer, Stream } from "effect"
 import { LLM, LLMEvent, Message } from "../../src/index.js"
 import { XAI } from "../../src/providers.js"
 import { OpenResponses } from "../../src/protocols/open-responses.js"
 import { OpenAIResponses } from "../../src/protocols/openai-responses.js"
+import * as ProviderShared from "../../src/protocols/shared.js"
 import { XAIResponses } from "../../src/protocols/xai-responses.js"
-import { LLMClient } from "../../src/route.js"
+import {
+  LLMClient,
+  RequestExecutor,
+  WebSocketTransport,
+  type ChannelCheckpoint,
+  type WebSocketChannelDriver,
+} from "../../src/route.js"
 import { compileRequest } from "../../src/route/client.js"
 import { it } from "../lib/effect.js"
 import { fixedResponse } from "../lib/http.js"
@@ -13,6 +20,35 @@ import { sseEvents } from "../lib/sse.js"

 const model = XAI.configure({ apiKey: "test", baseURL: "https://api.x.ai/v1" }).responses("grok-4.6")

+/** Runs a request through the WebSocket transport and hands back its channel driver; the HTTP fallback answers. */
+const channelDriver = (request: ReturnType<typeof LLM.request>) =>
+  Effect.gen(function* () {
+    let driver: WebSocketChannelDriver | undefined
+    yield* LLMClient.generate(request, {
+      webSocket: {
+        execute: (exchange) =>
+          Effect.sync(() => {
+            driver = exchange.driver
+            return { frames: exchange.fallback(), complete: Effect.void }
+          }),
+      },
+    }).pipe(Effect.provide(fixedResponse(sseEvents({ type: "response.completed", response: { id: "http" } }))))
+    if (!driver) throw new Error("Expected a WebSocket channel driver")
+    return driver
+  })
+
+const completed = (driver: WebSocketChannelDriver, id: string) =>
+  Effect.gen(function* () {
+    const create = yield* driver.create(undefined)
+    yield* driver.observe(create, ProviderShared.encodeJson({ type: "response.created", response: { id } }))
+    const observation = yield* driver.observe(
+      create,
+      ProviderShared.encodeJson({ type: "response.completed", response: { id } }),
+    )
+    if (observation.type !== "completed" || !observation.checkpoint) throw new Error("Expected a checkpoint")
+    return observation.checkpoint
+  })
+
 describe("xAI Responses route", () => {
  it.effect("composes the Open Responses baseline with xAI extensions", () =>
    Effect.gen(function* () {
@@ -162,6 +198,78 @@ describe("xAI Responses route", () => {
    }),
  )

+  it.effect("classifies xAI's untyped WebSocket error envelope", () =>
+    Effect.gen(function* () {
+      // xAI answers a rejected response.create with an error envelope that carries no event type.
+      const envelope = ProviderShared.encodeJson({
+        error: {
+          message:
+            'Request validation error: {"code":"400","error":"Argument not supported: instructions and previous_response_id together"}',
+          type: "api_error",
+        },
+      })
+      const webSocket = WebSocketTransport.makeDirect({
+        open: () =>
+          Effect.succeed({ sendText: () => Effect.void, messages: Stream.make(envelope), close: Effect.void }),
+      })
+      const error = yield* LLMClient.generate(LLM.request({ model, prompt: "Hello" }), { webSocket }).pipe(
+        Effect.provide(
+          LLMClient.layer.pipe(
+            Layer.provide(
+              Layer.succeed(
+                RequestExecutor.Service,
+                RequestExecutor.Service.of({ execute: () => Effect.die("unexpected HTTP request") }),
+              ),
+            ),
+          ),
+        ),
+        Effect.flip,
+      )
+
+      expect(error.reason._tag).toBe("ProviderInternal")
+      expect(error.message).toContain("Argument not supported: instructions and previous_response_id together")
+      expect(error.reason.body).toBe(envelope)
+    }),
+  )
+
+  it.effect("continues stored responses without instructions and sends unstored steps in full", () =>
+    Effect.gen(function* () {
+      const step = (store: boolean, ...prompts: string[]) =>
+        LLM.request({
+          model,
+          system: "You are terse.",
+          messages: prompts.map((prompt) => Message.user(prompt)),
+          providerOptions: { store },
+        })
+      const send = (store: boolean, checkpoint: ChannelCheckpoint) =>
+        channelDriver(step(store, "First", "Second")).pipe(Effect.flatMap((driver) => driver.create(checkpoint)))
+
+      const stored = yield* send(true, yield* completed(yield* channelDriver(step(true, "First")), "resp_1"))
+      expect(stored.mode).toBe("incremental")
+      expect(JSON.parse(stored.message)).toEqual({
+        type: "response.create",
+        model: "grok-4.6",
+        store: true,
+        include: ["reasoning.encrypted_content"],
+        previous_response_id: "resp_1",
+        input: [{ role: "user", content: [{ type: "input_text", text: "Second" }] }],
+      })
+
+      // The connection cache only serves stored responses, so the default store: false never chains.
+      const unstored = yield* send(false, yield* completed(yield* channelDriver(step(false, "First")), "resp_1"))
+      expect(unstored.mode).toBe("full")
+      expect(JSON.parse(unstored.message)).toMatchObject({
+        instructions: "You are terse.",
+        store: false,
+        input: [
+          { role: "user", content: [{ type: "input_text", text: "First" }] },
+          { role: "user", content: [{ type: "input_text", text: "Second" }] },
+        ],
+      })
+      expect(JSON.parse(unstored.message).previous_response_id).toBeUndefined()
+    }),
+  )
+
  it.effect("parses xAI hosted tool items", () =>
    Effect.gen(function* () {
      const item = { type: "x_search_call", id: "x_search_1", status: "completed", action: { query: "news" } }
@@ -19,11 +19,13 @@ story("cancelling a version mismatch permits reconnecting again", async ({ mount
  const component = await mount("app-dialog-ssh--incompatible-session")
  await component.getByRole("button", { name: "Reconnect", exact: true }).click()
  const dialog = page.getByRole("dialog")
-  await expect(dialog.getByRole("alert")).toBeVisible()
+  await expect(dialog.getByRole("status")).toContainText("Server update required")
+  await expect(dialog.getByRole("textbox")).toHaveCount(0)
  await dialog.getByRole("button", { name: "Cancel", exact: true }).click()
  await expect(dialog).toHaveCount(0)
  await component.getByRole("button", { name: "Reconnect", exact: true }).click()
-  await expect(dialog.getByRole("alert")).toBeVisible()
+  await expect(dialog.getByRole("status")).toContainText("Server update required")
+  await expect(dialog.getByRole("textbox")).toHaveCount(0)
  await dialog.getByRole("button", { name: "Cancel", exact: true }).click()
  await expect(dialog).toHaveCount(0)
 })
@@ -43,6 +45,17 @@ story("adding a server keeps all SSH challenges in the original connection dialo
  await expect(dialog).toHaveCount(0)
 })

+story("adding an incompatible server advances to a dedicated update step", async ({ mount, page }) => {
+  await mount("app-dialog-ssh--incompatible-host")
+  const dialog = page.getByRole("dialog")
+  await dialog.getByRole("textbox", { name: "Host or SSH command" }).fill("ssh devbox")
+  await dialog.getByRole("button", { name: "Add server", exact: true }).click()
+  await expect(dialog.getByRole("status")).toContainText("Server update required")
+  await expect(dialog.getByRole("textbox")).toHaveCount(0)
+  await expect(dialog.getByRole("alert")).toHaveCount(0)
+  await expect(dialog.getByRole("button", { name: "Update and reconnect", exact: true })).toBeVisible()
+})
+
 story("updating an incompatible connection continues authentication in the same dialog", async ({ mount, page }) => {
  const component = await mount("app-dialog-ssh--incompatible-session")
  await component.getByRole("button", { name: "Reconnect", exact: true }).click()
@@ -115,7 +115,7 @@ test("combines follow-up patches into one three-file stack inside Used", async (
  })
  const group = page.locator('[data-component="collapsed-tool-group"]')
  await group.getByRole("button", { name: "Used 2 Shell, Patch", exact: true }).click()
-  await expect(group.getByText("2 files", { exact: true })).toBeVisible()
+  await expect(group.locator('[data-slot="apply-patch-filename"]')).toHaveText(["a.ts", "b.ts"])
  await timeline.send(
    partUpdated(
      toolPart(
@@ -134,7 +134,6 @@ test("combines follow-up patches into one three-file stack inside Used", async (
    "true",
  )
  await expect(group.locator('[data-component="apply-patch-tool"]')).toHaveCount(1)
-  await expect(group.getByText("3 files", { exact: true })).toBeVisible()
  await expect(group.locator('[data-slot="apply-patch-filename"]')).toHaveText(["a.ts", "b.ts", "c.ts"])
 })

@@ -1,6 +1,6 @@
 {
  "name": "@opencode/app",
-  "version": "1.18.15",
+  "version": "2.0.0",
  "description": "",
  "type": "module",
  "exports": {
@@ -126,15 +126,9 @@ export function createComposerSubmit(input: ComposerSubmitInput) {

      if (command) {
        clearSubmission(input, submission)
-        // Commands always steer: the server applies a command's configured
-        // agent and model immediately at admission, so queueing one would
-        // reconfigure the turn it is supposed to wait behind.
-        void sendCommand(
-          session,
-          { ...value, delivery: "steer" },
-          command,
-          input.adapter.controls().model.selection.trackSessionCommit,
-        ).catch((error) => failSubmission(input, session, "command", error, restore, value.id))
+        void sendCommand(session, value, command, input.adapter.controls().model.selection.trackSessionCommit).catch(
+          (error) => failSubmission(input, session, "command", error, restore, value.id),
+        )
        return
      }
    } finally {
@@ -326,7 +320,8 @@ async function sendCommand(
  track?: ModelSelection["trackSessionCommit"],
 ) {
  const request = await buildSubmissionRequest(session, value)
-  await applySelection(session, value.selection, track)
+  // Like queued prompts, queued commands must not apply the composer's selection to active work.
+  if (value.delivery === "steer") await applySelection(session, value.selection, track)
  await session.api.command({
    sessionID: session.id,
    command: command.command,
@@ -285,7 +285,7 @@ export const DESKTOP_NATIVE_ENGLISH = {
  "desktop.updater.dialog.later": "Later",

  "desktop.cli.installed.title": "CLI Installed",
-  "desktop.cli.installed.message": "CLI installed to {{path}}\n\nRestart your terminal to use the 'opencode2' command.",
+  "desktop.cli.installed.message": "CLI installed to {{path}}\n\nRestart your terminal to use the 'opencode' command.",
  "desktop.cli.failed.title": "Installation Failed",
  "desktop.cli.failed.message": "Failed to install CLI: {{error}}",

@@ -54,8 +54,12 @@ export function createWebPlatform(version: string) {

 function getCurrentServerUrl() {
  if (import.meta.env.VITE_OPENCODE_SERVER_MODE === "none") return undefined
-  if (import.meta.env.DEV)
-    return `http://${import.meta.env.VITE_OPENCODE_SERVER_HOST ?? "localhost"}:${import.meta.env.VITE_OPENCODE_SERVER_PORT ?? "4096"}`
+  if (import.meta.env.DEV) {
+    const loopback =
+      location.hostname === "localhost" || location.hostname === "[::1]" || location.hostname.startsWith("127.")
+    const host = import.meta.env.VITE_OPENCODE_SERVER_HOST ?? (loopback ? location.hostname : "localhost")
+    return `http://${host}:${import.meta.env.VITE_OPENCODE_SERVER_PORT ?? "4096"}`
+  }
  return location.origin
 }

@@ -10,6 +10,7 @@ import { createData } from "@opencode/client/solid"
 import type { ServerScope } from "@/runtime/server/scope"
 import { createPermissionAutoApprover } from "@/session/requests/auto-approve"
 import { createServerNotificationState } from "@/shell/notifications/notification"
+import { createNotificationCoordinator } from "@/shell/notifications/coordinator"
 import { Persist, persisted } from "@/runtime/persistence/storage"
 import { createDesktopData } from "./data"
 import { ModelState } from "./persistence"
@@ -33,6 +34,7 @@ export const { use: useGlobal, provider: GlobalProvider } = createSimpleContext(
      },
    })
    const models = createGlobalModels()
+    const notificationCoordinator = createNotificationCoordinator()

    const settingsServer = createMemo(() => {
      const list = server.list
@@ -57,7 +59,7 @@ export const { use: useGlobal, provider: GlobalProvider } = createSimpleContext(
      if (existing) return existing
      const serverCtx = createRoot((dispose) => {
        serverCtxDisposers.set(key, dispose)
-        return createServerController(conn, server.scope(key), server.projects.forServer(key))
+        return createServerController(conn, server.scope(key), server.projects.forServer(key), notificationCoordinator)
      }, owner)
      serverCtxs.set(key, serverCtx)
      return serverCtx
@@ -131,6 +133,7 @@ function createServerController(
  conn: ServerConnection.Any,
  scope: ServerScope,
  projects: ReturnType<typeof createServerProjects>,
+  notificationCoordinator: ReturnType<typeof createNotificationCoordinator>,
 ) {
  const language = useLanguage()
  const settings = useSettings()
@@ -159,7 +162,7 @@ function createServerController(
  })
  const sync = createServerSyncContext(sdk, data)
  createPermissionAutoApprover({ sdk, data })
-  const notification = createServerNotificationState({ sdk, data, key: connKey })
+  const notification = createServerNotificationState({ sdk, data, key: connKey, coordinator: notificationCoordinator })

  function enrich(project: { worktree: string; expanded: boolean }) {
    const [childStore] = sync.child(project.worktree, { bootstrap: false })
@@ -278,6 +278,7 @@ function Open(props: { initial?: string }) {
 export default { title: "App/Dialogs/SSH", id: "app-dialog-ssh" }
 export const AuthenticationRequired = { render: () => <Fixture initial="required" /> }
 export const SettingsReconnect = { render: () => <Fixture initial="required" settings connectionDelay={200} /> }
+export const IncompatibleHost = { render: () => <Fixture incompatible /> }
 export const IncompatibleSession = { render: () => <Fixture initial="required" session incompatible /> }
 export const InactiveSession = { render: () => <Fixture initial="required" session connectionDelay={3000} /> }
 export const KeyReconnect = { render: () => <Fixture initial="required" session keyOnly connectionDelay={3000} /> }
@@ -120,7 +120,13 @@ export function DialogSsh(props: {
      <Divider />
      <DialogBody class="flex w-full min-w-0 flex-1 flex-col px-4 pt-4 pb-2">
        <div class="flex w-full min-w-0 flex-col gap-6">
-          <Show when={!props.promptOnly && (!state.prompted || (!!error() && !prompt()))}>
+          <Show
+            when={
+              !props.promptOnly &&
+              item()?.stage !== "incompatible" &&
+              (!state.prompted || (!!error() && !prompt()))
+            }
+          >
            <div class="flex w-full min-w-0 flex-col gap-2">
              <label class="settings-server-dialog-label" for="ssh-target">
                {language.t("ssh.target")}
@@ -160,6 +166,12 @@ export function DialogSsh(props: {
              />
            </div>
          </Show>
+          <Show when={item()?.stage === "incompatible"}>
+            <div class="flex w-full min-w-0 flex-col gap-2" role="status" aria-live="polite">
+              <span class="text-14-medium text-v2-text-text-base">{language.t("ssh.stage.incompatible")}</span>
+              <span class="text-13-regular text-v2-text-text-muted">{language.t("ssh.error.version")}</span>
+            </div>
+          </Show>
          <Show when={prompt()} keyed>
            {(prompt) => (
              <div class="flex w-full min-w-0 flex-col gap-2">
@@ -195,7 +207,7 @@ export function DialogSsh(props: {
              </div>
            )}
          </Show>
-          <Show when={error()}>
+          <Show when={item()?.stage !== "incompatible" && error()}>
            {(error) => (
              <span class="settings-server-dialog-error !leading-[var(--line-height-compact)]" role="alert">
                {error()}
@@ -1,6 +1,6 @@
 import { createMemo, createUniqueId, Show } from "solid-js"
 import { createStore } from "solid-js/store"
-import { createQuery } from "@tanstack/solid-query"
+import { createQuery, keepPreviousData } from "@tanstack/solid-query"
 import { Icon } from "@opencode/ui/icon"
 import { SessionFilePanelV2, SessionFilePanelV2Empty } from "@opencode/session-ui/v2/session-file-panel-v2"
 import { SessionReviewV2Sidebar } from "@opencode/session-ui/v2/session-review-v2"
@@ -56,6 +56,7 @@ export function SessionFileBrowserTab(props: {
      queryKey: [serverSDK.scope, "session-open-file", workspaceKey(), value] as const,
      enabled: serverSDK.connection.status() === "connected" && value.length > 0,
      queryFn: ({ signal }) => file.searchFiles(value, { limit: 200, signal }),
+      placeholderData: keepPreviousData,
    }
  })
  const files = createMemo(() => {
@@ -0,0 +1,93 @@
+import { onCleanup } from "solid-js"
+
+const FOCUS_LOCK = "opencode:notification-focus"
+const MAX_CLAIMED = 500
+
+export function createNotificationCoordinator() {
+  const locks = typeof navigator === "undefined" ? undefined : navigator.locks
+  const claimed = new Set<string>()
+  const focus = { pending: false, release: undefined as (() => void) | undefined }
+
+  const updateFocus = () => {
+    if (typeof document === "undefined" || !document.hasFocus()) {
+      focus.release?.()
+      return
+    }
+    if (!locks || focus.pending || focus.release) return
+
+    focus.pending = true
+    void locks
+      .request(FOCUS_LOCK, { mode: "shared" }, async () => {
+        focus.pending = false
+        if (!document.hasFocus()) return
+        await new Promise<void>((resolve) => {
+          focus.release = resolve
+        })
+        focus.release = undefined
+      })
+      .catch(() => {
+        focus.pending = false
+      })
+  }
+
+  if (typeof window !== "undefined") {
+    window.addEventListener("focus", updateFocus)
+    window.addEventListener("blur", updateFocus)
+    document.addEventListener("visibilitychange", updateFocus)
+    updateFocus()
+    onCleanup(() => {
+      window.removeEventListener("focus", updateFocus)
+      window.removeEventListener("blur", updateFocus)
+      document.removeEventListener("visibilitychange", updateFocus)
+      focus.release?.()
+    })
+  }
+
+  const once = async (kind: "sound" | "system", eventID: string, run: () => Promise<unknown> | void) => {
+    const key = `${kind}:${eventID}`
+    const execute = async () => {
+      if (!claim(kind, key, claimed)) return
+      await run()
+    }
+    if (!locks) return execute()
+    await locks.request(`opencode:notification:${key}`, execute)
+  }
+
+  return {
+    sound(eventID: string, run: () => Promise<unknown> | void) {
+      return once("sound", eventID, run)
+    },
+    system(eventID: string, run: () => Promise<unknown> | void) {
+      return once("system", eventID, async () => {
+        if (typeof document !== "undefined" && document.hasFocus()) return
+        if (!locks) return run()
+        await locks.request(FOCUS_LOCK, { mode: "exclusive", ifAvailable: true }, async (lock) => {
+          if (!lock) return
+          await run()
+        })
+      })
+    },
+  }
+}
+
+function claim(kind: "sound" | "system", eventID: string, claimed: Set<string>) {
+  if (claimed.has(eventID)) return false
+
+  if (typeof localStorage !== "undefined") {
+    try {
+      const storageKey = `opencode:notification-${kind}`
+      const value: unknown = JSON.parse(localStorage.getItem(storageKey) ?? "[]")
+      const events = Array.isArray(value) ? value.filter((item): item is string => typeof item === "string") : []
+      if (events.includes(eventID)) {
+        claimed.add(eventID)
+        return false
+      }
+      localStorage.setItem(storageKey, JSON.stringify([...events, eventID].slice(-MAX_CLAIMED)))
+    } catch {
+      // The in-memory claim still prevents duplicates in this renderer when storage is unavailable.
+    }
+  }
+
+  claimed.add(eventID)
+  return true
+}
@@ -11,7 +11,8 @@ import { useSettings } from "@/settings/model"
 import { decode64 } from "@/runtime/persistence/base64"
 import { Persist, persisted } from "@/runtime/persistence/storage"
 import { Persistence } from "@/runtime/persistence/schema"
-import { playSoundByIdOnce } from "@/shell/notifications/sound"
+import { playSoundById } from "@/shell/notifications/sound"
+import type { createNotificationCoordinator } from "@/shell/notifications/coordinator"
 import { useGlobal } from "@/runtime/server/runtime"
 import { ServerConnection, useServers } from "@/runtime/server/registry"
 import { sessionIDHasOpenTab, useTabs } from "@/shell/tabs/tabs"
@@ -114,7 +115,12 @@ function buildNotificationIndex(list: Notification[]) {
  return index
 }

-export function createServerNotificationState(input: { sdk: ServerSDK; data: Data; key: ServerConnection.Key }) {
+export function createServerNotificationState(input: {
+  sdk: ServerSDK
+  data: Data
+  key: ServerConnection.Key
+  coordinator: ReturnType<typeof createNotificationCoordinator>
+}) {
  const platform = usePlatform()
  const settings = useSettings()
  const language = useLanguage()
@@ -223,7 +229,7 @@ export function createServerNotificationState(input: { sdk: ServerSDK; data: Dat
      if (session.parentID) return

      if (sessionIDHasOpenTab(tabs.store, input.key, sessionID) && settings.sounds.agentEnabled()) {
-        void playSoundByIdOnce(settings.sounds.agent(), `${input.key}\0${eventID}`)
+        void input.coordinator.sound(`${input.key}\0${eventID}`, () => playSoundById(settings.sounds.agent()))
      }

      append({
@@ -235,8 +241,10 @@ export function createServerNotificationState(input: { sdk: ServerSDK; data: Dat
      })

      if (settings.notifications.agent()) {
-        void platform.notify(language.t("notification.session.responseReady.title"), session.title ?? sessionID, () =>
-          openNotificationSession(tabs, input.key, sessionID),
+        void input.coordinator.system(`${input.key}\0${eventID}`, () =>
+          platform.notify(language.t("notification.session.responseReady.title"), session.title ?? sessionID, () =>
+            openNotificationSession(tabs, input.key, sessionID),
+          ),
        )
      }
    })
@@ -248,7 +256,7 @@ export function createServerNotificationState(input: { sdk: ServerSDK; data: Dat
      if (session?.parentID) return

      if (sessionIDHasOpenTab(tabs.store, input.key, sessionID) && settings.sounds.errorsEnabled()) {
-        void playSoundByIdOnce(settings.sounds.errors(), `${input.key}\0${eventID}`)
+        void input.coordinator.sound(`${input.key}\0${eventID}`, () => playSoundById(settings.sounds.errors()))
      }

      append({
@@ -263,8 +271,10 @@ export function createServerNotificationState(input: { sdk: ServerSDK; data: Dat
        session?.title ??
        (typeof error === "string" ? error : language.t("notification.session.error.fallbackDescription"))
      if (settings.notifications.errors()) {
-        void platform.notify(language.t("notification.session.error.title"), description, () =>
-          openNotificationSession(tabs, input.key, sessionID),
+        void input.coordinator.system(`${input.key}\0${eventID}`, () =>
+          platform.notify(language.t("notification.session.error.title"), description, () =>
+            openNotificationSession(tabs, input.key, sessionID),
+          ),
        )
      }
    })
@@ -74,9 +74,6 @@ function getLoads() {
 }

 const cache = new Map<SoundID, Promise<string | undefined>>()
-const claimed = new Set<string>()
-const CLAIMED_STORAGE_KEY = "opencode:notification-sounds"
-const MAX_CLAIMED = 500

 export function soundSrc(id: string | undefined) {
  const loads = getLoads()
@@ -103,34 +100,3 @@ export function playSound(src: string | undefined) {
 export function playSoundById(id: string | undefined) {
  return soundSrc(id).then((src) => playSound(src))
 }
-
-export async function playSoundByIdOnce(id: string | undefined, eventID: string) {
-  const play = async () => {
-    if (!claim(eventID)) return
-    await playSoundById(id)
-  }
-
-  if (typeof navigator === "undefined" || !navigator.locks) return play()
-  await navigator.locks.request(`${CLAIMED_STORAGE_KEY}:${eventID}`, play)
-}
-
-function claim(eventID: string) {
-  if (claimed.has(eventID)) return false
-
-  if (typeof localStorage !== "undefined") {
-    try {
-      const value: unknown = JSON.parse(localStorage.getItem(CLAIMED_STORAGE_KEY) ?? "[]")
-      const events = Array.isArray(value) ? value.filter((item): item is string => typeof item === "string") : []
-      if (events.includes(eventID)) {
-        claimed.add(eventID)
-        return false
-      }
-      localStorage.setItem(CLAIMED_STORAGE_KEY, JSON.stringify([...events, eventID].slice(-MAX_CLAIMED)))
-    } catch {
-      // The in-memory claim still prevents duplicates in this renderer when storage is unavailable.
-    }
-  }
-
-  claimed.add(eventID)
-  return true
-}
@@ -0,0 +1,16 @@
+FROM alpine AS base
+
+ARG BUN_RUNTIME_TRANSPILER_CACHE_PATH=0
+ENV BUN_RUNTIME_TRANSPILER_CACHE_PATH=${BUN_RUNTIME_TRANSPILER_CACHE_PATH}
+RUN apk add --no-cache libgcc libstdc++ ripgrep
+
+FROM base AS build-amd64
+COPY dist/opencode-linux-x64-baseline-musl/bin/opencode /usr/local/bin/opencode
+
+FROM base AS build-arm64
+COPY dist/opencode-linux-arm64-musl/bin/opencode /usr/local/bin/opencode
+
+ARG TARGETARCH
+FROM build-${TARGETARCH}
+RUN opencode --version
+ENTRYPOINT ["opencode"]
@@ -0,0 +1,135 @@
+#!/usr/bin/env node
+
+const childProcess = require("child_process")
+const fs = require("fs")
+const path = require("path")
+const os = require("os")
+
+const forwardedSignals =
+  process.platform === "win32" ? ["SIGINT", "SIGTERM", "SIGHUP"] : ["SIGINT", "SIGTERM", "SIGHUP", "SIGUSR1"]
+
+function run(target) {
+  const child = childProcess.spawn(target, process.argv.slice(2), { stdio: "inherit" })
+  child.on("error", (error) => {
+    console.error(error.message)
+    process.exit(1)
+  })
+  const forwarders = {}
+  for (const signal of forwardedSignals) {
+    forwarders[signal] = () => {
+      try {
+        child.kill(signal)
+      } catch {}
+    }
+    process.on(signal, forwarders[signal])
+  }
+  child.on("exit", (code, signal) => {
+    for (const forwardedSignal of forwardedSignals) process.removeListener(forwardedSignal, forwarders[forwardedSignal])
+    if (signal) return process.kill(process.pid, signal)
+    process.exit(typeof code === "number" ? code : 0)
+  })
+}
+
+const envPath = process.env.OPENCODE_BIN_PATH
+const scriptDir = path.dirname(fs.realpathSync(__filename))
+const command = path.basename(__filename).replace(/\.cjs$/, "")
+const nodeBuild = command === "opencode-node"
+const sourceCommand = nodeBuild ? "opencode2-node" : "opencode"
+const cached = path.join(scriptDir, `.${command}`)
+const platform = { darwin: "darwin", linux: "linux", win32: "windows" }[os.platform()] || os.platform()
+const arch = { x64: "x64", arm64: "arm64", arm: "arm" }[os.arch()] || os.arch()
+const base = `@opencode/cli${nodeBuild ? "-node" : ""}-` + platform + "-" + arch
+const binary = platform === "windows" ? `${sourceCommand}.exe` : sourceCommand
+
+function supportsAvx2() {
+  if (arch !== "x64") return false
+  if (platform === "linux") {
+    try {
+      return /(^|\s)avx2(\s|$)/i.test(fs.readFileSync("/proc/cpuinfo", "utf8"))
+    } catch {
+      return false
+    }
+  }
+  if (platform === "darwin") {
+    try {
+      const result = childProcess.spawnSync("sysctl", ["-n", "hw.optional.avx2_0"], { encoding: "utf8", timeout: 1500 })
+      return result.status === 0 && (result.stdout || "").trim() === "1"
+    } catch {
+      return false
+    }
+  }
+  if (platform === "windows") {
+    const command =
+      '(Add-Type -MemberDefinition "[DllImport(""kernel32.dll"")] public static extern bool IsProcessorFeaturePresent(int ProcessorFeature);" -Name Kernel32 -Namespace Win32 -PassThru)::IsProcessorFeaturePresent(40)'
+    for (const executable of ["powershell.exe", "pwsh.exe", "pwsh", "powershell"]) {
+      try {
+        const result = childProcess.spawnSync(executable, ["-NoProfile", "-NonInteractive", "-Command", command], {
+          encoding: "utf8",
+          timeout: 3000,
+          windowsHide: true,
+        })
+        if (result.status !== 0) continue
+        const output = (result.stdout || "").trim().toLowerCase()
+        if (output === "true" || output === "1") return true
+        if (output === "false" || output === "0") return false
+      } catch {
+        continue
+      }
+    }
+  }
+  return false
+}
+
+const names = (() => {
+  if (nodeBuild) return [base]
+  const baseline = arch === "x64" && !supportsAvx2()
+  if (platform === "linux") {
+    const musl = (() => {
+      try {
+        if (fs.existsSync("/etc/alpine-release")) return true
+        const result = childProcess.spawnSync("ldd", ["--version"], { encoding: "utf8" })
+        return ((result.stdout || "") + (result.stderr || "")).toLowerCase().includes("musl")
+      } catch {
+        return false
+      }
+    })()
+    if (musl)
+      return arch === "x64"
+        ? baseline
+          ? [`${base}-baseline-musl`, `${base}-musl`, `${base}-baseline`, base]
+          : [`${base}-musl`, `${base}-baseline-musl`, base, `${base}-baseline`]
+        : [`${base}-musl`, base]
+    return arch === "x64"
+      ? baseline
+        ? [`${base}-baseline`, base, `${base}-baseline-musl`, `${base}-musl`]
+        : [base, `${base}-baseline`, `${base}-musl`, `${base}-baseline-musl`]
+      : [base, `${base}-musl`]
+  }
+  return arch === "x64" ? (baseline ? [`${base}-baseline`, base] : [base, `${base}-baseline`]) : [base]
+})()
+
+function findBinary(startDir) {
+  let current = startDir
+  for (;;) {
+    const modules = path.join(current, "node_modules")
+    if (fs.existsSync(modules))
+      for (const name of names) {
+        const candidate = path.join(modules, name, "bin", binary)
+        if (fs.existsSync(candidate)) return candidate
+      }
+    const parent = path.dirname(current)
+    if (parent === current) return
+    current = parent
+  }
+}
+
+const resolved = envPath || (fs.existsSync(cached) ? cached : findBinary(scriptDir))
+if (!resolved) {
+  console.error(
+    `It seems that your package manager failed to install the right ${command} CLI package. Try manually installing ` +
+      names.map((name) => `"${name}"`).join(" or ") +
+      " package",
+  )
+  process.exit(1)
+}
+run(resolved)
@@ -1,134 +1,4 @@
 #!/usr/bin/env node

-const childProcess = require("child_process")
-const fs = require("fs")
-const path = require("path")
-const os = require("os")
-
-const forwardedSignals =
-  process.platform === "win32" ? ["SIGINT", "SIGTERM", "SIGHUP"] : ["SIGINT", "SIGTERM", "SIGHUP", "SIGUSR1"]
-
-function run(target) {
-  const child = childProcess.spawn(target, process.argv.slice(2), { stdio: "inherit" })
-  child.on("error", (error) => {
-    console.error(error.message)
-    process.exit(1)
-  })
-  const forwarders = {}
-  for (const signal of forwardedSignals) {
-    forwarders[signal] = () => {
-      try {
-        child.kill(signal)
-      } catch {}
-    }
-    process.on(signal, forwarders[signal])
-  }
-  child.on("exit", (code, signal) => {
-    for (const forwardedSignal of forwardedSignals) process.removeListener(forwardedSignal, forwarders[forwardedSignal])
-    if (signal) return process.kill(process.pid, signal)
-    process.exit(typeof code === "number" ? code : 0)
-  })
-}
-
-const envPath = process.env.OPENCODE_BIN_PATH
-const scriptDir = path.dirname(fs.realpathSync(__filename))
-const command = path.basename(__filename).replace(/\.cjs$/, "")
-const nodeBuild = command === "opencode2-node"
-const cached = path.join(scriptDir, `.${command}`)
-const platform = { darwin: "darwin", linux: "linux", win32: "windows" }[os.platform()] || os.platform()
-const arch = { x64: "x64", arm64: "arm64", arm: "arm" }[os.arch()] || os.arch()
-const base = `@opencode/cli${nodeBuild ? "-node" : ""}-` + platform + "-" + arch
-const binary = platform === "windows" ? `${command}.exe` : command
-
-function supportsAvx2() {
-  if (arch !== "x64") return false
-  if (platform === "linux") {
-    try {
-      return /(^|\s)avx2(\s|$)/i.test(fs.readFileSync("/proc/cpuinfo", "utf8"))
-    } catch {
-      return false
-    }
-  }
-  if (platform === "darwin") {
-    try {
-      const result = childProcess.spawnSync("sysctl", ["-n", "hw.optional.avx2_0"], { encoding: "utf8", timeout: 1500 })
-      return result.status === 0 && (result.stdout || "").trim() === "1"
-    } catch {
-      return false
-    }
-  }
-  if (platform === "windows") {
-    const command =
-      '(Add-Type -MemberDefinition "[DllImport(""kernel32.dll"")] public static extern bool IsProcessorFeaturePresent(int ProcessorFeature);" -Name Kernel32 -Namespace Win32 -PassThru)::IsProcessorFeaturePresent(40)'
-    for (const executable of ["powershell.exe", "pwsh.exe", "pwsh", "powershell"]) {
-      try {
-        const result = childProcess.spawnSync(executable, ["-NoProfile", "-NonInteractive", "-Command", command], {
-          encoding: "utf8",
-          timeout: 3000,
-          windowsHide: true,
-        })
-        if (result.status !== 0) continue
-        const output = (result.stdout || "").trim().toLowerCase()
-        if (output === "true" || output === "1") return true
-        if (output === "false" || output === "0") return false
-      } catch {
-        continue
-      }
-    }
-  }
-  return false
-}
-
-const names = (() => {
-  if (nodeBuild) return [base]
-  const baseline = arch === "x64" && !supportsAvx2()
-  if (platform === "linux") {
-    const musl = (() => {
-      try {
-        if (fs.existsSync("/etc/alpine-release")) return true
-        const result = childProcess.spawnSync("ldd", ["--version"], { encoding: "utf8" })
-        return ((result.stdout || "") + (result.stderr || "")).toLowerCase().includes("musl")
-      } catch {
-        return false
-      }
-    })()
-    if (musl)
-      return arch === "x64"
-        ? baseline
-          ? [`${base}-baseline-musl`, `${base}-musl`, `${base}-baseline`, base]
-          : [`${base}-musl`, `${base}-baseline-musl`, base, `${base}-baseline`]
-        : [`${base}-musl`, base]
-    return arch === "x64"
-      ? baseline
-        ? [`${base}-baseline`, base, `${base}-baseline-musl`, `${base}-musl`]
-        : [base, `${base}-baseline`, `${base}-musl`, `${base}-baseline-musl`]
-      : [base, `${base}-musl`]
-  }
-  return arch === "x64" ? (baseline ? [`${base}-baseline`, base] : [base, `${base}-baseline`]) : [base]
-})()
-
-function findBinary(startDir) {
-  let current = startDir
-  for (;;) {
-    const modules = path.join(current, "node_modules")
-    if (fs.existsSync(modules))
-      for (const name of names) {
-        const candidate = path.join(modules, name, "bin", binary)
-        if (fs.existsSync(candidate)) return candidate
-      }
-    const parent = path.dirname(current)
-    if (parent === current) return
-    current = parent
-  }
-}
-
-const resolved = envPath || (fs.existsSync(cached) ? cached : findBinary(scriptDir))
-if (!resolved) {
-  console.error(
-    `It seems that your package manager failed to install the right ${command} CLI package. Try manually installing ` +
-      names.map((name) => `"${name}"`).join(" or ") +
-      " package",
-  )
-  process.exit(1)
-}
-run(resolved)
+console.log("opencode2 is now just opencode. run opencode")
+process.exit(1)
@@ -1,10 +1,11 @@
 {
  "$schema": "https://json.schemastore.org/package.json",
  "name": "@opencode/cli",
-  "version": "1.18.4",
+  "version": "2.0.0",
  "type": "module",
  "license": "MIT",
  "bin": {
+    "opencode": "./bin/opencode.cjs",
    "opencode2": "./bin/opencode2.cjs"
  },
  "files": [
@@ -12,7 +12,7 @@ import { verifyArtifact, verifySimulationGraph } from "./verify-artifact"
 import { resolveOpencodePty } from "./opencode-pty"

 const dir = path.resolve(import.meta.dirname, "..")
-const binary = "opencode2"
+const binary = "opencode"
 const outdir = path.resolve(
  dir,
  process.argv.find((arg) => arg.startsWith("--outdir="))?.slice("--outdir=".length) ?? "dist",
@@ -149,7 +149,7 @@ export default { path: file, version: ${JSON.stringify(opencodePty.version)}, sh
    },
    define: {
      OPENCODE_VERSION: `'${Script.version}'`,
-      OPENCODE_CLI_NAME: `'${binary}'`,
+      OPENCODE_CLI_NAME: "'opencode'",
      OPENCODE_CHANNEL: `'${Script.channel}'`,
      OPENCODE_ARTIFACT: `'cli'`,
      OPENCODE_LIBC: item.os === "linux" ? `'${item.abi ?? "glibc"}'` : "undefined",
@@ -12,10 +12,11 @@ const require = createRequire(import.meta.url)
 const packageJson = JSON.parse(fs.readFileSync(path.join(directory, "package.json"), "utf8"))
 const command = Object.keys(packageJson.bin ?? {})[0]
 if (!command) throw new Error("OpenCode package does not declare a binary")
+const sourceCommand = packageJson.opencodeSourceBinary ?? command

 const platform = { darwin: "darwin", linux: "linux", win32: "windows" }[os.platform()] ?? os.platform()
 const arch = { x64: "x64", arm64: "arm64", arm: "arm" }[os.arch()] ?? os.arch()
-const sourceBinary = platform === "windows" ? `${command}.exe` : command
+const sourceBinary = platform === "windows" ? `${sourceCommand}.exe` : sourceCommand
 const targetBinary = path.resolve(directory, packageJson.bin[command])
 const dependencies = packageJson.optionalDependencies ?? {}
 const base = Object.keys(dependencies).find((name) => name.endsWith(`-${platform}-${arch}`))
@@ -6,15 +6,10 @@ import path from "node:path"
 import { fileURLToPath } from "node:url"
 import { UpdateArtifact } from "../../../script/update-artifact"

-if (Script.channel !== "beta" && Script.channel !== "latest") {
-  throw new Error("AUR publishing requires the beta or latest channel")
-}
-const beta = Script.channel === "beta"
-const name = beta ? "opencode-beta" : "opencode-bin"
-const command = beta ? "opencode2" : "opencode"
-if (!(beta ? /^\d+\.\d+\.\d+-beta[.-]\d+(?:\.\d+)?$/ : /^\d+\.\d+\.\d+$/).test(Script.version)) {
-  throw new Error(`Expected a ${Script.channel} release version`)
-}
+if (Script.channel !== "beta") throw new Error("AUR publishing requires the beta channel")
+const name = "opencode-beta"
+const command = "opencode"
+if (!/^\d+\.\d+\.\d+-beta[.-]\d+(?:\.\d+)?$/.test(Script.version)) throw new Error("Expected a beta release version")

 const dir = fileURLToPath(new URL("..", import.meta.url))
 const root = path.resolve(process.env.OPENCODE_CLI_DIST ?? path.join(dir, "dist"))
@@ -22,6 +17,7 @@ const outdir = path.join(root, `aur-${name}`)
 const dryRun = process.argv.includes("--dry-run")
 const pkgver = Script.version.replaceAll("-", ".")
 const license = Bun.file(path.join(dir, "..", "..", "LICENSE"))
+const shim = "#!/bin/sh\necho 'opencode2 is now just opencode. run opencode'\nexit 1\n"

 await rm(outdir, { recursive: true, force: true })
 await mkdir(path.dirname(outdir), { recursive: true })
@@ -48,6 +44,7 @@ const sources = await Promise.all(
 )

 await Bun.write(path.join(outdir, "LICENSE"), license)
+await Bun.write(path.join(outdir, "opencode2"), shim)
 await Bun.write(
  path.join(outdir, "PKGBUILD"),
  [
@@ -55,21 +52,22 @@ await Bun.write(
    `pkgname=${name}`,
    `pkgver=${pkgver}`,
    "pkgrel=1",
-    `pkgdesc='OpenCode${beta ? " V2 beta" : ""} - the AI coding agent for the terminal'`,
+    "pkgdesc='OpenCode beta - the AI coding agent for the terminal'",
    "url='https://github.com/anomalyco/opencode'",
    "arch=('x86_64' 'aarch64')",
    "license=('MIT')",
    "depends=('glibc' 'gcc-libs' 'ripgrep')",
-    `provides=('${command}')`,
-    `conflicts=('${command}')`,
+    `provides=('${command}' 'opencode2')`,
+    `conflicts=('${command}' 'opencode2')`,
    // Stripping a compiled Bun executable can damage its embedded application.
    "options=('!strip' '!debug')",
-    "source=('LICENSE')",
-    `sha256sums=('${new Bun.CryptoHasher("sha256").update(await license.arrayBuffer()).digest("hex")}')`,
+    "source=('LICENSE' 'opencode2')",
+    `sha256sums=('${new Bun.CryptoHasher("sha256").update(await license.arrayBuffer()).digest("hex")}' '${new Bun.CryptoHasher("sha256").update(shim).digest("hex")}')`,
    ...sources,
    "",
    "package() {",
-    `  install -Dm755 "$srcdir/package/bin/opencode2" "$pkgdir/usr/bin/${command}"`,
+    `  install -Dm755 "$srcdir/package/bin/opencode" "$pkgdir/usr/bin/${command}"`,
+    '  install -Dm755 "$srcdir/opencode2" "$pkgdir/usr/bin/opencode2"',
    '  install -Dm644 "$srcdir/LICENSE" "$pkgdir/usr/share/licenses/$pkgname/LICENSE"',
    "}",
    "",
@@ -79,7 +77,7 @@ await Bun.write(path.join(outdir, ".SRCINFO"), await $`makepkg --printsrcinfo`.c
 console.log(`Prepared ${name} ${pkgver} in ${outdir}`)
 if (dryRun) process.exit(0)

-await $`git add PKGBUILD .SRCINFO LICENSE`.cwd(outdir)
+await $`git add PKGBUILD .SRCINFO LICENSE opencode2`.cwd(outdir)
 if ((await $`git diff --cached --quiet`.cwd(outdir).nothrow()).exitCode !== 0) {
  await $`git commit -m ${`chore: update ${name} to ${pkgver}`}`.cwd(outdir)
  await $`git push origin master`.cwd(outdir)
@@ -29,6 +29,8 @@ async function publish(dir: string, name: string, version: string) {
 async function publishDistribution(input: {
  root: string
  name: string
+  command: string
+  legacyCommand?: string
  binary: string
  packagePrefix: string
  artifact: string
@@ -48,7 +50,7 @@ async function publishDistribution(input: {

  await $`mkdir -p ${input.root}/${input.name}/bin`
  await $`cp ./script/postinstall.mjs ${input.root}/${input.name}/postinstall.mjs`
-  await Bun.file(`${input.root}/${input.name}/bin/${input.binary}.exe`).write(
+  await Bun.file(`${input.root}/${input.name}/bin/${input.command}.exe`).write(
    [
      `echo "Error: ${input.name}'s postinstall script was not run." >&2`,
      'echo "" >&2',
@@ -58,11 +60,19 @@ async function publishDistribution(input: {
      "",
    ].join("\n"),
  )
+  if (input.legacyCommand)
+    await Bun.file(`${input.root}/${input.name}/bin/${input.legacyCommand}.cjs`).write(
+      `#!/usr/bin/env node\n\nconsole.log(${JSON.stringify(`opencode2 is now just opencode. run ${input.command}`)})\nprocess.exit(1)\n`,
+    )
  await Bun.file(`${input.root}/${input.name}/package.json`).write(
    JSON.stringify(
      {
        name: input.name,
-        bin: { [input.binary]: `./bin/${input.binary}.exe` },
+        bin: {
+          [input.command]: `./bin/${input.command}.exe`,
+          ...(input.legacyCommand ? { [input.legacyCommand]: `./bin/${input.legacyCommand}.cjs` } : {}),
+        },
+        ...(input.command !== input.binary ? { opencodeSourceBinary: input.binary } : {}),
        scripts: { postinstall: "node ./postinstall.mjs" },
        version,
        license: pkg.license,
@@ -120,7 +130,9 @@ async function publishDistribution(input: {
 await publishDistribution({
  root,
  name: pkg.name,
-  binary: "opencode2",
+  command: "opencode",
+  legacyCommand: "opencode2",
+  binary: "opencode",
  packagePrefix: "@opencode/cli-",
  artifact: "cli",
 })
@@ -128,13 +140,18 @@ if (existsSync(path.join(root, "node"))) {
  await publishDistribution({
    root: path.join(root, "node"),
    name: "@opencode/cli-node",
+    command: "opencode2-node",
    binary: "opencode2-node",
    packagePrefix: "@opencode/cli-node-",
    artifact: "cli-node",
  })
 }

-if ((Script.channel === "beta" || Script.channel === "latest") && Script.release) {
+if (Script.channel === "latest" && Script.release && !dryRun) {
+  await $`docker buildx build --platform linux/amd64,linux/arm64 --tag ghcr.io/anomalyco/opencode:${Script.version} --push .`
+}
+
+if (Script.channel === "beta" && Script.release) {
  await $`bun ./script/publish-aur.ts ${dryRun ? ["--dry-run"] : []}`.env({ ...process.env, OPENCODE_CLI_DIST: root })
 }

@@ -11,7 +11,10 @@ import path from "node:path"
 const nodeBuild = process.argv.includes("--node")
 const target = `cli${nodeBuild ? "-node" : ""}-${process.platform === "win32" ? "windows" : process.platform}-${process.arch}`
 const directory = path.join(import.meta.dir, "..", "dist", ...(nodeBuild ? ["node"] : []), target, "bin")
-const binary = path.join(directory, `opencode2${nodeBuild ? "-node" : ""}${process.platform === "win32" ? ".exe" : ""}`)
+const binary = path.join(
+  directory,
+  `${nodeBuild ? "opencode2-node" : "opencode"}${process.platform === "win32" ? ".exe" : ""}`,
+)
 if (!(await Bun.file(binary).exists())) throw new Error(`Missing compiled CLI in ${directory}`)

 const root = await fs.realpath(await fs.mkdtemp(path.join(os.tmpdir(), "opencode-service-smoke-")))
@@ -76,7 +76,10 @@ export function buildEffortSelectOption(input: {
    category: "thought_level",
    type: "select",
    currentValue: selectVariant(input.currentVariant, input.variants),
-    options: input.variants.map((variant) => ({ value: variant, name: formatVariantName(variant) })),
+    options: [...new Set([...input.variants, DEFAULT_VARIANT_VALUE])].map((variant) => ({
+      value: variant,
+      name: formatVariantName(variant),
+    })),
  }
 }

@@ -125,6 +128,7 @@ export function formatVariantName(variant: string) {
 }

 function selectVariant(variant: string | undefined, variants: readonly string[]) {
+  if (!variant || variant === DEFAULT_VARIANT_VALUE) return DEFAULT_VARIANT_VALUE
  if (variant && variants.includes(variant)) return variant
  if (variants.includes(DEFAULT_VARIANT_VALUE)) return DEFAULT_VARIANT_VALUE
  return variants[0] ?? DEFAULT_VARIANT_VALUE
@@ -201,7 +201,7 @@ export async function streamTurn(input: {
        if (!child) assistantMessageID = event.data.assistantMessageID
        await send({
          sessionUpdate: "agent_thought_chunk",
-          messageId: event.data.assistantMessageID,
+          messageId: `${event.data.assistantMessageID}:reasoning:${event.data.ordinal}`,
          content: { type: "text", text: event.data.delta },
        })
        continue
@@ -455,6 +455,8 @@ async function replayMessage(
    return
  }
  if (message.type !== "assistant") return
+  // Live reasoning ordinals count only reasoning parts, not the mixed content array.
+  let reasoningOrdinal = 0
  for (const part of message.content) {
    if (part.type === "text") {
      await connection.sessionUpdate({
@@ -472,7 +474,7 @@ async function replayMessage(
        sessionId: sessionID,
        update: {
          sessionUpdate: "agent_thought_chunk",
-          messageId: message.id,
+          messageId: `${message.id}:reasoning:${reasoningOrdinal++}`,
          content: { type: "text", text: part.text },
        },
      })
@@ -41,7 +41,12 @@ import type {
 } from "@agentclientprotocol/sdk"
 import { OPENCODE_VERSION } from "../version"
 import { SessionMessage } from "@opencode/schema/session-message"
-import { buildConfigOptions, parseModelSelection, type ConfigOptionProvider } from "./config-option"
+import {
+  buildConfigOptions,
+  DEFAULT_VARIANT_VALUE,
+  parseModelSelection,
+  type ConfigOptionProvider,
+} from "./config-option"
 import { promptContentToParts } from "./content"
 import {
  ChildSessionUpdateMethod,
@@ -275,7 +280,7 @@ export function make(input: { readonly client: OpenCodeClient; readonly connecti
      if (typeof params.value !== "string") throw new ACPError.InvalidConfigOptionError({ configId: params.configId })
      switch (params.configId) {
        case "model": {
-          const selected = requireModel(state.catalog, params.value)
+          const selected = requireModel(state.catalog, params.value, state.model)
          state.model = selected
          await input.client.session.switchModel({ sessionID: state.id, model: selected })
          break
@@ -284,7 +289,10 @@ export function make(input: { readonly client: OpenCodeClient; readonly connecti
          const model = state.catalog.models.find(
            (item) => item.providerID === state.model.providerID && item.id === state.model.id,
          )
-          if (!model?.variants.some((variant) => variant.id === params.value))
+          if (
+            !model ||
+            (params.value !== DEFAULT_VARIANT_VALUE && !model.variants.some((variant) => variant.id === params.value))
+          )
            throw new ACPError.InvalidEffortError({ effort: params.value })
          state.model = { ...state.model, variant: params.value }
          await input.client.session.switchModel({ sessionID: state.id, model: state.model })
@@ -453,7 +461,7 @@ function providers(models: readonly ModelInfo[]): ConfigOptionProvider[] {
    }))
 }

-function requireModel(catalog: Catalog, modelID: string): ModelRef {
+function requireModel(catalog: Catalog, modelID: string, current: ModelRef): ModelRef {
  const selected = parseModelSelection(modelID, catalog.providers)
  const model = catalog.models.find(
    (item) => item.providerID === selected.model.providerID && item.id === selected.model.modelID,
@@ -461,7 +469,14 @@ function requireModel(catalog: Catalog, modelID: string): ModelRef {
  if (!model) throw new ACPError.InvalidModelError({ providerId: selected.model.providerID, modelId: modelID })
  if (selected.variant && !model.variants.some((variant) => variant.id === selected.variant))
    throw new ACPError.InvalidEffortError({ effort: selected.variant })
-  return { providerID: model.providerID, id: model.id, variant: selected.variant }
+  const variant =
+    selected.variant ??
+    (current.providerID === model.providerID &&
+    current.id === model.id &&
+    (current.variant === DEFAULT_VARIANT_VALUE || model.variants.some((variant) => variant.id === current.variant))
+      ? current.variant
+      : undefined)
+  return { providerID: model.providerID, id: model.id, variant }
 }

 async function selectMode(client: OpenCodeClient, state: Attached, modeID: string) {
@@ -36,7 +36,7 @@ const PermissionParams = {
 }

 const Root = Spec.make(typeof OPENCODE_CLI_NAME === "string" ? OPENCODE_CLI_NAME : "opencode", {
-  description: "OpenCode 2.0 preview command line interface",
+  description: "OpenCode command line interface",
  params: {
    ...ServerParams,
    ...PermissionParams,
@@ -47,7 +47,7 @@ const handler = Effect.fn("cli.session.list")(function* (
          null,
          2,
        )
-      : formatTable(page.data)) + EOL
+      : formatList(page.data)) + EOL
  const write = Effect.tryPromise(
    () =>
      new Promise<void>((resolve, reject) => {
@@ -96,18 +96,14 @@ export default Runtime.handler(Commands.commands.session.commands.list, (input)
  ),
 )

-function formatTable(sessions: ReadonlyArray<SessionInfo>) {
-  const rows = sessions.map((session) => ({
-    id: session.id,
-    title: (session.title ?? "Untitled session").replace(/[\r\n\t]/g, " "),
-    updated: new Date(session.time.updated).toLocaleString(),
-  }))
-  const idWidth = Math.max(20, ...rows.map((row) => row.id.length))
-  const titleWidth = Math.max(25, ...rows.map((row) => row.title.length))
-  const header = `${"Session ID".padEnd(idWidth)}  ${"Title".padEnd(titleWidth)}  Updated`
-  return [
-    header,
-    "─".repeat(header.length),
-    ...rows.map((row) => `${row.id.padEnd(idWidth)}  ${row.title.padEnd(titleWidth)}  ${row.updated}`),
-  ].join(EOL)
+function formatList(sessions: ReadonlyArray<SessionInfo>) {
+  return sessions
+    .map((session) =>
+      [
+        session.id,
+        (session.title ?? "Untitled session").replace(/[\r\n\t]/g, " "),
+        new Date(session.time.updated).toLocaleString(),
+      ].join("\t"),
+    )
+    .join(EOL)
 }
@@ -105,7 +105,7 @@ const make = Effect.gen(function* () {
      global.home,
      ".opencode",
      "bin",
-      process.platform === "win32" ? "opencode2.exe" : "opencode2",
+      process.platform === "win32" ? "opencode.exe" : "opencode",
    )
    if (path.resolve(process.execPath) === path.resolve(binary)) return "curl"
    if (!installedPackage) return
@@ -186,7 +186,10 @@ const make = Effect.gen(function* () {
        "npm",
        "install",
        "--global",
-        ...(installedPackage && packageName !== installedPackage ? ["--force"] : []),
+        ...((OPENCODE_ARTIFACT === "cli" && !installedPackage?.endsWith("/cli-node")) ||
+        (installedPackage && packageName !== installedPackage)
+          ? ["--force"]
+          : []),
        target,
      ],
      pnpm: ["pnpm", "add", "--global", `--allow-build=${packageName}`, target],
@@ -50,11 +50,11 @@ describe("acp config option subprocess", () => {
    const effort = requireSelectOption((await newSession(acp, fixture.home)).configOptions, "effort")

    expect(effort.category).toBe("thought_level")
-    expect(effort.currentValue).toBe("low")
-    expect(flattenSelectOptions(effort).map((option) => option.value)).toEqual(["low", "high"])
+    expect(effort.currentValue).toBe("default")
+    expect(flattenSelectOptions(effort).map((option) => option.value)).toEqual(["low", "high", "default"])
  }, 60_000)

-  test("effort switch updates currentValue", async () => {
+  test("effort survives model synchronization and can be reset to default", async () => {
    await using fixture = await createAcpFixture()
    const acp = fixture.spawn()
    await initialize(acp)
@@ -70,5 +70,23 @@ describe("acp config option subprocess", () => {
    )

    expect(selectConfigOption(updated.configOptions, "effort")?.currentValue).toBe(nextEffort)
+
+    const synchronized = expectOk(
+      await acp.request<SetSessionConfigOptionResponse>("session/set_config_option", {
+        sessionId: session.sessionId,
+        configId: "model",
+        value: requireSelectOption(session.configOptions, "model").currentValue,
+      }),
+    )
+    expect(selectConfigOption(synchronized.configOptions, "effort")?.currentValue).toBe(nextEffort)
+
+    const reset = expectOk(
+      await acp.request<SetSessionConfigOptionResponse>("session/set_config_option", {
+        sessionId: session.sessionId,
+        configId: "effort",
+        value: "default",
+      }),
+    )
+    expect(selectConfigOption(reset.configOptions, "effort")?.currentValue).toBe("default")
  }, 60_000)
 })
@@ -91,7 +91,7 @@ describe("acp event behavior", () => {
    }
  })

-  test("preserves text and reasoning order before returning the terminal response", async () => {
+  test("preserves reasoning boundaries and update order during streaming and replay", async () => {
    const firstUpdate = Promise.withResolvers<void>()
    const releaseUpdate = Promise.withResolvers<void>()
    const allUpdates = Promise.withResolvers<void>()
@@ -108,6 +108,14 @@ describe("acp event behavior", () => {
            delta: "think-1",
          }),
        )
+        send(
+          ephemeralEvent("session.reasoning.delta", {
+            sessionID: "ses_order",
+            assistantMessageID: "msg_order",
+            ordinal: 0,
+            delta: " continued",
+          }),
+        )
        send(
          ephemeralEvent("session.text.delta", {
            sessionID: "ses_order",
@@ -120,7 +128,7 @@ describe("acp event behavior", () => {
          ephemeralEvent("session.reasoning.delta", {
            sessionID: "ses_order",
            assistantMessageID: "msg_order",
-            ordinal: 2,
+            ordinal: 1,
            delta: "think-2",
          }),
        )
@@ -144,7 +152,7 @@ describe("acp event behavior", () => {
          firstUpdate.resolve()
          await releaseUpdate.promise
        }
-        if (updates.length === 3) allUpdates.resolve()
+        if (updates.length === 4) allUpdates.resolve()
      },
      requestPermission: async () => ({ outcome: { outcome: "cancelled" } }),
    } satisfies Connection
@@ -171,18 +179,48 @@ describe("acp event behavior", () => {
          ) {
            return [
              item.update.sessionUpdate,
+              item.update.messageId,
              item.update.content.type === "text" ? item.update.content.text : undefined,
            ]
          }
          return [item.update.sessionUpdate, undefined]
        }),
      ).toEqual([
-        ["agent_thought_chunk", "think-1"],
-        ["agent_message_chunk", "answer"],
-        ["agent_thought_chunk", "think-2"],
+        ["agent_thought_chunk", "msg_order:reasoning:0", "think-1"],
+        ["agent_thought_chunk", "msg_order:reasoning:0", " continued"],
+        ["agent_message_chunk", "msg_order", "answer"],
+        ["agent_thought_chunk", "msg_order:reasoning:1", "think-2"],
      ])
      expect(fixture.requests.at(-1)?.path).toBe("/api/session/ses_order/message/msg_order")
      expect(response).toMatchObject({ stopReason: "end_turn", usage: { totalTokens: 2 } })
+
+      const replayed: SessionUpdateParams[] = []
+      await replayMessages(recordingConnection(replayed), "ses_order", "/workspace", [
+        {
+          id: "msg_order",
+          type: "assistant",
+          agent: "build",
+          model: { providerID: "test", id: "test-model" },
+          time: { created: 1 },
+          content: [
+            { type: "reasoning", text: "think-1 continued" },
+            { type: "text", text: "answer" },
+            { type: "reasoning", text: "think-2" },
+          ],
+        },
+      ])
+      expect(replayed).toEqual([
+        {
+          sessionId: "ses_order",
+          update: {
+            sessionUpdate: "agent_thought_chunk",
+            messageId: "msg_order:reasoning:0",
+            content: { type: "text", text: "think-1 continued" },
+          },
+        },
+        updates[2],
+        updates[3],
+      ])
    } finally {
      releaseUpdate.resolve()
      releaseSubmit.resolve()
@@ -222,7 +222,7 @@ describe("acp service directory behavior", () => {
    await fixture.service.setSessionMode({ sessionId: session.sessionId, modeId: "build" })

    expect(currentValue(selectedModel, "model")).toBe("test/second-model")
-    expect(currentValue(selectedModel, "effort")).toBe("low")
+    expect(currentValue(selectedModel, "effort")).toBe("default")
    expect(currentValue(selectedEffort, "effort")).toBe("medium")
    expect(currentValue(selectedMode, "mode")).toBe("plan")
    expect(
@@ -32,7 +32,7 @@ describe("acp service lifecycle", () => {
        model: { providerID: "test", id: "second-model" },
      },
    })
-    expect(currentValue(created, "effort")).toBe("none")
+    expect(currentValue(created, "effort")).toBe("default")
  })

  test("loads and forks with paginated replay while resume does not replay", async () => {
@@ -423,24 +423,24 @@ test("updates effective duplicate canonical keybinds", async () => {
  const file = path.join(directory.path, "cli.json")
  await Bun.write(
    file,
-    `{"keybinds":{"session.delete":"first","session.delete":"last","permission.mode":"off","permission.mode":"on"}}`,
+    `{"keybinds":{"session.delete":"first","session.delete":"last","opencode.settings":"off","opencode.settings":"on"}}`,
  )

  const config = await run(
    directory.path,
    Effect.gen(function* () {
      const service = yield* Config.Service
-      expect((yield* service.get()).keybinds).toEqual({ "session.delete": "last", "permission.mode": "on" })
+      expect((yield* service.get()).keybinds).toEqual({ "session.delete": "last", "opencode.settings": "on" })
      return yield* service.update((draft) => {
-        draft.keybinds = { ...draft.keybinds, "session.delete": "changed", "permission.mode": "changed" }
+        draft.keybinds = { ...draft.keybinds, "session.delete": "changed", "opencode.settings": "changed" }
      })
    }),
  )

-  expect(config.keybinds).toEqual({ "session.delete": "changed", "permission.mode": "changed" })
+  expect(config.keybinds).toEqual({ "session.delete": "changed", "opencode.settings": "changed" })
  expect(parse(await Bun.file(file).text()).keybinds).toEqual({
    "session.delete": "changed",
-    "permission.mode": "changed",
+    "opencode.settings": "changed",
  })
 })

@@ -100,7 +100,7 @@ function fixture(
 }

 const installs = [
-  { method: "npm", command: ["npm", "install", "--global", "@opencode/cli@2.3.4-beta.1"] },
+  { method: "npm", command: ["npm", "install", "--global", "--force", "@opencode/cli@2.3.4-beta.1"] },
  {
    method: "pnpm",
    command: ["pnpm", "add", "--global", "--allow-build=@opencode/cli", "@opencode/cli@2.3.4-beta.1"],
@@ -1,7 +1,7 @@
 {
  "$schema": "https://json.schemastore.org/package.json",
  "name": "@opencode/client",
-  "version": "1.17.13",
+  "version": "2.0.0",
  "type": "module",
  "license": "MIT",
  "repository": {
@@ -11,13 +11,13 @@ import type { RelativePath } from "@opencode/schema/schema"
 import type { Brand } from "effect"
 import type { Model } from "@opencode/schema/model"
 import type { DateTime } from "effect"
+import type { Permission } from "@opencode/schema/permission"
 import type { SessionMessage } from "@opencode/schema/session-message"
 import type { SessionInbox } from "@opencode/schema/session-inbox"
 import type { PromptInput } from "@opencode/schema/prompt-input"
 import type { AgentAttachment } from "@opencode/schema/prompt"
 import type { Skill } from "@opencode/schema/skill"
 import type { Event } from "@opencode/schema/event"
-import type { FileDiff } from "@opencode/schema/file-diff"
 import type { InstructionEntry } from "@opencode/schema/instruction-entry"
 import type { Schema } from "effect"
 import type { EventLog } from "@opencode/schema/event-log"
@@ -27,7 +27,6 @@ import type { Integration } from "@opencode/schema/integration"
 import type { Form } from "@opencode/schema/form"
 import type { Mcp } from "@opencode/schema/mcp"
 import type { Credential } from "@opencode/schema/credential"
-import type { Permission } from "@opencode/schema/permission"
 import type { PermissionSaved } from "@opencode/schema/permission-saved"
 import type { FileSystem } from "@opencode/schema/filesystem"
 import type { Command } from "@opencode/schema/command"
@@ -37,6 +36,7 @@ import type { PtyTicket } from "@opencode/schema/pty-ticket"
 import type { Reference } from "@opencode/schema/reference"
 import type { Worktree } from "@opencode/schema/worktree"
 import type { Vcs } from "@opencode/schema/vcs"
+import type { FileDiff } from "@opencode/schema/file-diff"
 import type { WebSearch } from "@opencode/schema/websearch"
 import type { Config } from "@opencode/schema/config"

@@ -209,6 +209,7 @@ export type SessionCreateInput = {
  readonly model?: Model.Ref | undefined
  readonly location?: Location.Ref | undefined
  readonly metadata?: Session.Metadata | undefined
+  readonly permissions?: Permission.Ruleset | undefined
 }
 export type SessionCreateOutput = Session.Info
 export type SessionCreateOperation<E = never> = (input?: SessionCreateInput) => Effect.Effect<SessionCreateOutput, E>
@@ -360,15 +361,6 @@ export type SessionContextInput = { readonly sessionID: Session.ID }
 export type SessionContextOutput = ReadonlyArray<SessionMessage.Info>
 export type SessionContextOperation<E = never> = (input: SessionContextInput) => Effect.Effect<SessionContextOutput, E>

-export type SessionDiffInput = {
-  readonly sessionID: Session.ID
-  readonly messageID?: SessionMessage.ID | undefined
-  readonly to?: SessionMessage.ID | undefined
-  readonly context?: number | undefined
-}
-export type SessionDiffOutput = ReadonlyArray<FileDiff.Info>
-export type SessionDiffOperation<E = never> = (input: SessionDiffInput) => Effect.Effect<SessionDiffOutput, E>
-
 export type SessionInboxListInput = { readonly sessionID: Session.ID }
 export type SessionInboxListOutput = ReadonlyArray<SessionInbox.Info>
 export type SessionInboxListOperation<E = never> = (
@@ -446,6 +438,7 @@ export type SessionLogOutput =
            readonly agent?: Agent.ID | undefined
            readonly model?: Model.Ref | undefined
            readonly metadata?: Session.Metadata | undefined
+            readonly permissions?: Permission.Ruleset | undefined
            readonly version: string
          }
        }
@@ -498,6 +491,15 @@ export type SessionLogOutput =
          readonly location?: Location.Ref | undefined
          readonly data: { readonly sessionID: Session.ID; readonly title: string }
        }
+      | {
+          readonly id: Event.ID
+          readonly created: number
+          readonly metadata?: { readonly [x: string]: unknown } | undefined
+          readonly type: "session.permissions.updated"
+          readonly durable: { readonly aggregateID: string; readonly seq: Event.Seq; readonly version: Event.Version }
+          readonly location?: Location.Ref | undefined
+          readonly data: { readonly sessionID: Session.ID; readonly permissions: Permission.Ruleset }
+        }
      | {
          readonly id: Event.ID
          readonly created: number
@@ -1148,7 +1150,6 @@ export interface SessionApi<E = never> {
    readonly commit: SessionRevertCommitOperation<E>
  }
  readonly context: SessionContextOperation<E>
-  readonly diff: SessionDiffOperation<E>
  readonly inbox: {
    readonly list: SessionInboxListOperation<E>
    readonly cancel: SessionInboxCancelOperation<E>
@@ -1595,6 +1596,12 @@ export type PermissionReplyOperation<E = never> = (
  input: PermissionReplyInput,
 ) => Effect.Effect<PermissionReplyOutput, E>

+export type PermissionRulesInput = { readonly sessionID: Session.ID; readonly permissions: Permission.Ruleset }
+export type PermissionRulesOutput = void
+export type PermissionRulesOperation<E = never> = (
+  input: PermissionRulesInput,
+) => Effect.Effect<PermissionRulesOutput, E>
+
 export interface PermissionApi<E = never> {
  readonly request: { readonly list: PermissionRequestListOperation<E> }
  readonly saved: { readonly list: PermissionSavedListOperation<E>; readonly remove: PermissionSavedRemoveOperation<E> }
@@ -1602,6 +1609,7 @@ export interface PermissionApi<E = never> {
  readonly list: PermissionListOperation<E>
  readonly get: PermissionGetOperation<E>
  readonly reply: PermissionReplyOperation<E>
+  readonly rules: PermissionRulesOperation<E>
 }

 export type FileListInput = {
--- a/Show More
+++ b/Show More
Author	SHA1	Message	Date
Aiden Cline	a1e99be863	fix(tui): guard streaming part appends with a part index Every session.text.delta calls appendPart, which ran setRows(produce(...)) plus a hasPart full-tree scan even when the part already existed. On a root-array Solid store the no-op update still unwraps the whole rows tree, so per-delta cost scaled with loaded history (~580ms per 50-delta burst at 4000 messages). Maintain a lazily-rebuilt Set of part keys: rebuilt from the plain reduce() output on each reconcile, extended on append, and checked O(1) before touching the store. Reasoning completion still flows through to flip the group flag. Measured with a headless createTestRenderer probe (mixed markdown + tools, 50 deltas + settle): 578ms -> 34ms at 4000 messages, 133ms -> 34ms at 1000 messages.	2026-09-11 16:26:20 -05:00
Dax	312651f68e	feat(tui): configure session permission handling (#48545 )	2026-09-11 19:50:39 +00:00
Dax	654f411a5b	feat(release): isolate v2 stable publishing (#48546 )	2026-09-11 15:50:35 -04:00
Aiden Cline	68b862da49	refactor(codemode): make functions program objects with name and length (#48541 )	2026-09-11 14:50:07 -05:00
Dax	8f2b13e864	fix(tui): display slash command aliases (#48544 )	2026-09-11 15:34:33 -04:00
Dax	bb564f96a3	feat(cli): rename command to opencode	2026-09-11 15:03:45 -04:00
Aiden Cline	a555f24ded	refactor: flatten provider settings across core and the AI package (#48429 )	2026-09-11 13:37:24 -05:00
Aiden Cline	10ae321bf3	refactor(codemode): give programs their own objects and arrays (#48527 )	2026-09-11 13:36:35 -05:00
Aiden Cline	6e8ff0c2b7	refactor(core): remove dead AISDK hooks (#48523 )	2026-09-11 12:41:20 -05:00
Aiden Cline	923bda07a2	feat(ai): always send interleaved thinking beta for Anthropic (#48521 )	2026-09-11 12:29:18 -05:00
Aiden Cline	4c34ee5eab	test(codemode): run test262 files verbatim from a local checkout (#48446 )	2026-09-11 10:40:02 -05:00
Shoubhit Dash	2df00955cb	fix(cli): preserve ACP reasoning boundaries and effort selections (#48493 )	2026-09-11 18:46:54 +05:30
opencode-agent[bot]andnexxeln	e0deffa083	docs(plugin): fix Effect system part example (#48492 ) Co-authored-by: nexxeln <95541290+nexxeln@users.noreply.github.com>	2026-09-11 18:12:11 +05:30
Wu Shuwen	4e97b78d98	docs: fix v2 plugin system example (#48452 )	2026-09-11 18:12:01 +05:30
Brendan Allan	6bb4b35399	fix(app): preserve loopback server host (#44296 )	2026-09-11 06:41:38 +00:00
Brendan Allan	0ce1383030	fix(app): preserve file search results while loading (#43832 )	2026-09-11 06:13:32 +00:00
Brendan AllanandBrendonovich	010cd6131e	fix(app): merge adjacent file-changing tools (#44977 ) Co-authored-by: Brendonovich <Brendonovich@users.noreply.github.com>	2026-09-11 05:58:02 +00:00
Brendan Allan	c2348f8f69	fix(app): coordinate notifications across windows (#44613 )	2026-09-11 05:28:13 +00:00
Aiden Cline	71317ec7c9	fix(core): report provider initialization failures (#48433 )	2026-09-11 00:26:55 -05:00
opencode-agent[bot]andBrendonovich	4261fe749d	fix(desktop): repair SSH server updates (#48442 ) Co-authored-by: Brendonovich <14191578+Brendonovich@users.noreply.github.com>	2026-09-11 13:24:38 +08:00
Aiden Cline	9dd7149e75	feat(core): append comment guidance to Anthropic system prompt (#48444 )	2026-09-11 00:05:05 -05:00
Brendan Allan	0c0a431c9f	fix(ui): make comment cancel action ghost (#46742 )	2026-09-11 13:03:49 +08:00
Brendan Allan	3368e049d2	fix(app): refine home row hover actions (#47270 )	2026-09-11 04:39:26 +00:00
Aiden Cline	f0b5da1c11	feat(ai): type provider configuration errors (#48440 )	2026-09-10 23:29:55 -05:00
Aiden Cline	15c525dcfb	fix(merman): support & node groups in flowchart statements (#48425 )	2026-09-10 22:24:59 -05:00
James Long	4d12e01824	refactor(tui): project production subgroups through tree engine (#48399 )	2026-09-10 23:15:58 -04:00
James Long	70afbac80c	docs(www): add core CLI theme concepts (#48426 )	2026-09-10 23:13:53 -04:00
James Long	1c723c56fa	feat(tui): add recursive grouping tree (#48394 )	2026-09-10 21:40:23 -04:00
Aiden Cline	181428a2f3	fix(codemode): use Bun's wording for the missing atob/btoa argument (#48381 )	2026-09-10 20:08:36 -05:00
Dax Raad	9b1891fb7e	docs(cli): expand provider connections	2026-09-10 20:43:17 -04:00
Dax Raad	d4bf78b348	docs: add websearch guides	2026-09-10 20:30:58 -04:00
Dax Raad	d4ceffe787	docs: refresh v2 guides	2026-09-10 20:12:08 -04:00
Aiden Cline	872e38055e	fix(ai): normalize flat Responses stream errors (#48376 )	2026-09-10 18:23:32 -05:00
James Long	0c1dfa9186	refactor(tui): extract shared session rendering primitives (#48393 )	2026-09-10 18:07:14 -04:00
Dax Raad	8f4d706647	feat(cli): add command docs and simplify session list	2026-09-10 16:27:12 -04:00
Aiden Cline	929374cdfd	feat(core): parse JSON text results from MCP tools without an output schema (#48357 )	2026-09-10 15:20:13 -05:00
opencode-agent[bot]	cfa5ba700e	fix(stats): canonicalize DeepSeek Flash usage (#48373 )	2026-09-10 13:58:51 -05:00
Shoubhit Dash	45a2ed9a97	feat(core): per-session permission rules (#48351 )	2026-09-10 22:30:53 +05:30
opencode-agent[bot]andnexxeln	f6333546f8	fix(tui): disambiguate plugin actions (#48354 ) Co-authored-by: nexxeln <95541290+nexxeln@users.noreply.github.com>	2026-09-10 22:10:01 +05:30
opencode-agent[bot]andnexxeln	9e153ce7b3	fix(core): raise instruction entry limit (#48350 ) Co-authored-by: nexxeln <nexxeln@users.noreply.github.com>	2026-09-10 20:59:38 +05:30
Simon Klee	eb357f17cf	fix(tui): strip NUL characters before clipboard writes (#48337 )	2026-09-10 15:47:03 +02:00
Shoubhit Dash	573d76933f	fix(ai): make xAI Responses websockets work and enable them (#48318 )	2026-09-10 17:27:43 +05:30
usrnk1andBrendonovich	bb8194395a	feat(desktop): respect follow-up behavior for slash commands (#48169 ) Co-authored-by: Brendonovich <14191578+Brendonovich@users.noreply.github.com>	2026-09-10 19:24:23 +08:00