fix(core): trigger compaction at 85% of the input window

packages/ai/src/protocols/open-responses.ts

@@ -454,6 +454,11 @@ export const lowerTool = Effect.fn("OpenResponses.lowerTool")(function* (protoco
   }
 })
 
+export const lowerTools = (tools: ReadonlyArray<ToolDefinition>, adapter: ProviderAdapter) =>
+  Effect.forEach(tools, (tool) =>
+    tool.native !== undefined && adapter.nativeTool ? adapter.nativeTool(tool.native) : lowerTool(adapter.name, tool),
+  )
+
 export const lowerToolChoice = (protocolName: string, toolChoice: NonNullable<LLMRequest["toolChoice"]>) =>
   ProviderShared.matchToolChoice(protocolName, toolChoice, {
     auto: () => "auto" as const,
@@ -815,14 +820,7 @@ export const fromRequestWithAdapter = Effect.fn("OpenResponses.fromRequestWithAd
   return {
     ...(yield* lowerConversation(projected.request, adapter)),
     ...lowerGeneration(request),
-    tools:
-      projected.tools.length === 0
-        ? undefined
-        : yield* Effect.forEach(projected.tools, (tool) =>
-            tool.native !== undefined && adapter.nativeTool
-              ? adapter.nativeTool(tool.native)
-              : lowerTool(adapter.name, tool),
-          ),
+    tools: projected.tools.length === 0 ? undefined : yield* lowerTools(projected.tools, adapter),
     tool_choice:
       allowedToolChoice(request) ??
       (request.toolChoice ? yield* lowerToolChoice(adapter.name, request.toolChoice) : undefined),

packages/ai/src/protocols/openai-responses.ts

@@ -5,7 +5,7 @@ import { Auth } from "../route/auth.js"
 import { Endpoint } from "../route/endpoint.js"
 import { Protocol } from "../route/protocol.js"
 import { HttpTransport } from "../route/transport/index.js"
-import { LLMRequest, mergeJsonRecords, type ToolDefinition, type ToolEntry } from "../schema/index.js"
+import { LLMRequest, type ToolDefinition, type ToolEntry } from "../schema/index.js"
 import { resolveEffortUpdates } from "../effort-updates.js"
 import { OpenResponses } from "./open-responses.js"
 import { OpenResponsesOptions } from "./utils/open-responses-options.js"
@@ -135,11 +135,6 @@ export const CompactionTrigger = Schema.Struct({ type: Schema.Literal("compactio
 const CheckpointBody = Schema.Struct({
   ...OpenAIResponsesBody.fields,
   input: Schema.Array(Schema.Union([OpenAIResponsesInputItem, CompactionTrigger])),
-  store: Schema.Literal(false),
-  prompt_cache_retention: optionalNull(Schema.String),
-  prompt_cache_options: optionalNull(
-    Schema.Struct({ mode: Schema.optional(Schema.String), ttl: Schema.optional(Schema.String) }),
-  ),
 })
 
 const adapter = {
@@ -191,6 +186,8 @@ const lowerToolEntry = Effect.fn("OpenAIResponses.lowerToolEntry")(function* (to
   }
 })
 
+const lowerTools = (request: LLMRequest) => Effect.forEach(request.tools, lowerToolEntry)
+
 const lowerToolChoice = (toolChoice: NonNullable<LLMRequest["toolChoice"]>, tools: ReadonlyArray<ToolEntry>) =>
   ProviderShared.matchToolChoice(NAME, toolChoice, {
     auto: () => "auto" as const,
@@ -214,7 +211,7 @@ const fromRequest = Effect.fn("OpenAIResponses.fromRequest")(function* (request:
     ...(yield* OpenResponses.lowerConversation(updates.request, adapter)),
     ...OpenResponses.lowerGeneration(request, { ...options, reasoningEffort: updates.effort }),
     context_management: management?.map((edit) => ({ type: edit.type, compact_threshold: edit.compactThreshold })),
-    tools: request.tools.length === 0 ? undefined : yield* Effect.forEach(request.tools, lowerToolEntry),
+    tools: request.tools.length === 0 ? undefined : yield* lowerTools(request),
     tool_choice:
       request.tools.length === 0
         ? undefined
@@ -226,7 +223,6 @@ const fromRequest = Effect.fn("OpenAIResponses.fromRequest")(function* (request:
 const checkpointBody = {
   schema: CheckpointBody,
   from: Effect.fn("OpenAIResponses.checkpointBody")(function* (request: LLMRequest) {
-    const native = yield* fromRequest(LLMRequest.update(request, { toolChoice: undefined }))
     const overlay = request.http?.body
     // Complete history is required for stateless replay and SSE recovery. Raw input overrides bypass that contract.
     if (
@@ -237,18 +233,13 @@ const checkpointBody = {
       return yield* ProviderShared.invalidRequest(
         "Trigger compaction requires complete canonical history, not an input or continuation override",
       )
-    return yield* ProviderShared.validateWith(Schema.decodeUnknownEffect(CheckpointBody))({
-      ...mergeJsonRecords(native, overlay),
-      input: [...native.input, { type: "compaction_trigger" }],
-      stream: true,
-      store: false,
-      parallel_tool_calls: true,
-      tool_choice: undefined,
-      context_management: undefined,
-      text: undefined,
-      max_output_tokens: undefined,
-      max_tool_calls: undefined,
-    })
+    if (overlay?.stream !== undefined && overlay.stream !== true)
+      return yield* ProviderShared.invalidRequest("Trigger compaction requires a streamed response")
+    const native = yield* fromRequest(request)
+    return {
+      ...native,
+      input: [...native.input, { type: "compaction_trigger" as const }],
+    }
   }),
 }
 
@@ -330,7 +321,10 @@ export const transport = channelTransport({
 })
 
 export const route = Route.make({
-  compact: { endpoint: ResponsesCompaction.make(adapter), trigger: ResponsesCheckpoint.make(checkpointBody) },
+  compact: {
+    endpoint: ResponsesCompaction.make(adapter, lowerTools),
+    trigger: ResponsesCheckpoint.make(checkpointBody),
+  },
   id: ADAPTER,
   provider: "openai",
   providerMetadataKey: "openai",

packages/ai/src/protocols/utils/responses-checkpoint.ts

@@ -1,7 +1,7 @@
 import { Effect, Schema, Stream } from "effect"
 import { Route, type RouteBody, type TriggerCompactOperation } from "../../route/client.js"
 import { Protocol } from "../../route/protocol.js"
-import { CompactionCheckpointResponse, HttpOptions, LLMEvent, LLMRequest } from "../../schema/index.js"
+import { CompactionCheckpointResponse, LLMEvent, LLMRequest } from "../../schema/index.js"
 import { OpenResponses } from "../open-responses.js"
 import { ProviderShared } from "../shared.js"
 
@@ -109,12 +109,8 @@ export const make = <Body>(body: RouteBody<Body>): TriggerCompactOperation =>
       transport: source.transport,
     })
     const native = yield* body.from(request)
-    // The body builder already applied and validated overlays. Do not let transport reapply them.
-    const preparedRequest = LLMRequest.update(request, {
-      http: request.http === undefined ? undefined : new HttpOptions({ ...request.http, body: undefined }),
-    })
-    const prepared = yield* route.prepareTransport(native, preparedRequest, options)
-    yield* route.streamPrepared(prepared, preparedRequest, { http: executor }, options).pipe(Stream.runDrain)
+    const prepared = yield* route.prepareTransport(native, request, options)
+    yield* route.streamPrepared(prepared, request, { http: executor }, options).pipe(Stream.runDrain)
     if (!result) return yield* ProviderShared.eventError(source.id, "Compaction response ended without a checkpoint")
     return result
   })

packages/ai/src/protocols/utils/responses-compaction.ts

@@ -19,12 +19,18 @@ import { OpenResponses } from "../open-responses.js"
 import { JsonObject, optionalNull, ProviderShared } from "../shared.js"
 import { Media } from "../../media.js"
 
+// /compact has a smaller wire contract than /responses; keep the request controls it accepts.
 const Body = Schema.Struct({
   model: Schema.String,
   input: Schema.Array(Schema.Unknown),
   instructions: optionalNull(Schema.String),
   previous_response_id: optionalNull(Schema.String),
   service_tier: optionalNull(Schema.String),
+  reasoning: Schema.optional(JsonObject),
+  text: Schema.optional(JsonObject),
+  include: OpenResponses.coreFields.include,
+  parallel_tool_calls: OpenResponses.coreFields.parallel_tool_calls,
+  tools: Schema.optional(Schema.Array(JsonObject)),
   prompt_cache_key: optionalNull(Schema.String),
   prompt_cache_retention: optionalNull(Schema.String),
   prompt_cache_options: optionalNull(
@@ -74,17 +80,27 @@ const Response = Schema.Struct({
   usage: Schema.optional(Schema.StructWithRest(OpenResponses.OpenResponsesUsage, [JsonObject])),
 })
 
-export const make = (adapter: OpenResponses.ProviderAdapter): CompactOperation =>
+export const make = (
+  adapter: OpenResponses.ProviderAdapter,
+  lowerTools: (request: LLMRequest) => Effect.Effect<ReadonlyArray<Record<string, unknown>>, AIError>,
+): CompactOperation =>
   Effect.fn("ResponsesCompaction.execute")(function* (request, executor, options) {
     const route = request.model.route
     // The standalone compaction endpoint rejects histories containing configuration updates.
     const native = yield* OpenResponses.lowerConversation(stripEffortUpdates(request), adapter)
+    const generation = OpenResponses.lowerGeneration(request)
+    const tools = request.tools.length === 0 ? undefined : yield* lowerTools(request)
     const body = yield* ProviderShared.validateWith(Schema.decodeUnknownEffect(Body))(
       mergeJsonRecords(
         {
           ...native,
-          service_tier: request.providerOptions?.serviceTier,
-          prompt_cache_key: ProviderShared.promptCacheKey(request),
+          service_tier: generation.service_tier,
+          reasoning: generation.reasoning,
+          text: generation.text,
+          include: generation.include,
+          parallel_tool_calls: generation.parallel_tool_calls,
+          tools,
+          prompt_cache_key: generation.prompt_cache_key,
         },
         request.http?.body,
       ),

packages/ai/src/protocols/xai-responses.ts

@@ -50,7 +50,8 @@ const fromRequest = Effect.fn("XAIResponses.fromRequest")(function* (request: LL
       operation: "in-band-compaction",
       provider: request.model.provider,
       route: request.model.route.id,
-      message: "xAI requires explicit compaction through LLMClient.compact; automatic context management is not supported",
+      message:
+        "xAI requires explicit compaction through LLMClient.compact; automatic context management is not supported",
     })
   return yield* decodeBody(yield* OpenResponses.fromRequestWithAdapter(request, adapter))
 })
@@ -93,6 +94,8 @@ export const protocol = Protocol.make({
   },
 })
 
-export const compact = ResponsesCompaction.make(adapter)
+export const compact = ResponsesCompaction.make(adapter, (request) =>
+  OpenResponses.lowerTools(ProviderShared.flattenTools(request.tools), adapter),
+)
 
 export * as XAIResponses from "./xai-responses.js"

packages/ai/src/providers/openai-options.ts

@@ -32,10 +32,7 @@ const openAIProviderOptions = (options: OpenAIOptionsInput | undefined): Provide
   return result
 }
 
-export const gpt5DefaultOptions = (
-  modelID: string,
-  options: { readonly textVerbosity?: boolean } = {},
-): ProviderOptions | undefined => {
+export const gpt5DefaultOptions = (modelID: string): ProviderOptions | undefined => {
   const id = modelID.toLowerCase()
   if (!id.includes("gpt-5") || id.includes("gpt-5-chat") || id.includes("gpt-5-pro")) return undefined
   return openAIProviderOptions({
@@ -47,27 +44,19 @@ export const gpt5DefaultOptions = (
     // this, callers using the default model facade get reasoning summaries
     // they cannot replay statelessly.
     include: ["reasoning.encrypted_content"],
-    textVerbosity:
-      options.textVerbosity === true && id.includes("gpt-5.") && !id.includes("codex") && !id.includes("-chat")
-        ? "low"
-        : undefined,
   })
 }
 
-export const openAIDefaultOptions = (
-  modelID: string,
-  options: { readonly textVerbosity?: boolean } = {},
-): ProviderOptions | undefined =>
-  mergeProviderOptions(openAIProviderOptions({ store: false }), gpt5DefaultOptions(modelID, options))
+export const openAIDefaultOptions = (modelID: string): ProviderOptions | undefined =>
+  mergeProviderOptions(openAIProviderOptions({ store: false }), gpt5DefaultOptions(modelID))
 
 export const withOpenAIOptions = <Options extends { readonly providerOptions?: OpenAIProviderOptionsInput }>(
   modelID: string,
   options: Options,
-  defaults: { readonly textVerbosity?: boolean } = {},
 ): Omit<Options, "providerOptions"> & { readonly providerOptions?: ProviderOptions } => {
   return {
     ...options,
-    providerOptions: mergeProviderOptions(openAIDefaultOptions(modelID, defaults), options.providerOptions),
+    providerOptions: mergeProviderOptions(openAIDefaultOptions(modelID), options.providerOptions),
   }
 }
 

packages/ai/src/providers/openai.ts

@@ -100,7 +100,7 @@ export const configure = (input: Config = {}) => {
   const modelDefaults = defaults(input)
   const responses = (id: string | ModelID) =>
     responsesRoute
-      .with(withOpenAIOptions(id, modelDefaults, { textVerbosity: true }))
+      .with(withOpenAIOptions(id, modelDefaults))
       .model<OpenAIProviderOptionsInput>({ id })
   const chat = (id: string | ModelID) =>
     chatRoute.with(withOpenAIOptions(id, modelDefaults)).model<OpenAIProviderOptionsInput>({

packages/ai/test/fixtures/recordings/openai-responses-websocket/continues-a-tool-call-over-one-socket.json

@@ -30,7 +30,7 @@
         {
           "direction": "client",
           "kind": "text",
-          "body": "{\"type\":\"response.create\",\"model\":\"gpt-5.5\",\"input\":[{\"type\":\"message\",\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"What is the weather in Paris?\"}]}],\"instructions\":\"Call get_weather once, then reply exactly: Paris is sunny.\",\"tools\":[{\"type\":\"function\",\"name\":\"get_weather\",\"description\":\"Get current weather for a city.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"city\":{\"type\":\"string\"}},\"required\":[\"city\"],\"additionalProperties\":false},\"strict\":false}],\"store\":false,\"include\":[\"reasoning.encrypted_content\"],\"reasoning\":{\"effort\":\"medium\",\"summary\":\"auto\"},\"text\":{\"verbosity\":\"low\"},\"max_output_tokens\":50}"
+          "body": "{\"type\":\"response.create\",\"model\":\"gpt-5.5\",\"input\":[{\"type\":\"message\",\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"What is the weather in Paris?\"}]}],\"instructions\":\"Call get_weather once, then reply exactly: Paris is sunny.\",\"tools\":[{\"type\":\"function\",\"name\":\"get_weather\",\"description\":\"Get current weather for a city.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"city\":{\"type\":\"string\"}},\"required\":[\"city\"],\"additionalProperties\":false},\"strict\":false}],\"store\":false,\"include\":[\"reasoning.encrypted_content\"],\"reasoning\":{\"effort\":\"medium\",\"summary\":\"auto\"},\"max_output_tokens\":50}"
         },
         {
           "direction": "server",
@@ -525,7 +525,7 @@
         {
           "direction": "client",
           "kind": "text",
-          "body": "{\"type\":\"response.create\",\"model\":\"gpt-5.5\",\"input\":[{\"type\":\"function_call_output\",\"call_id\":\"call_hejtTYDa1IfLyNIzb3fq9gJs\",\"output\":\"{\\\"temperature\\\":22,\\\"condition\\\":\\\"sunny\\\"}\"}],\"instructions\":\"Call get_weather once, then reply exactly: Paris is sunny.\",\"tools\":[{\"type\":\"function\",\"name\":\"get_weather\",\"description\":\"Get current weather for a city.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"city\":{\"type\":\"string\"}},\"required\":[\"city\"],\"additionalProperties\":false},\"strict\":false}],\"store\":false,\"include\":[\"reasoning.encrypted_content\"],\"reasoning\":{\"effort\":\"medium\",\"summary\":\"auto\"},\"text\":{\"verbosity\":\"low\"},\"max_output_tokens\":50,\"previous_response_id\":\"resp_0d9a44b6df400533016aa8c8e36de887d1be260913d131b2ca\"}"
+          "body": "{\"type\":\"response.create\",\"model\":\"gpt-5.5\",\"input\":[{\"type\":\"function_call_output\",\"call_id\":\"call_hejtTYDa1IfLyNIzb3fq9gJs\",\"output\":\"{\\\"temperature\\\":22,\\\"condition\\\":\\\"sunny\\\"}\"}],\"instructions\":\"Call get_weather once, then reply exactly: Paris is sunny.\",\"tools\":[{\"type\":\"function\",\"name\":\"get_weather\",\"description\":\"Get current weather for a city.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"city\":{\"type\":\"string\"}},\"required\":[\"city\"],\"additionalProperties\":false},\"strict\":false}],\"store\":false,\"include\":[\"reasoning.encrypted_content\"],\"reasoning\":{\"effort\":\"medium\",\"summary\":\"auto\"},\"max_output_tokens\":50,\"previous_response_id\":\"resp_0d9a44b6df400533016aa8c8e36de887d1be260913d131b2ca\"}"
         },
         {
           "direction": "server",

packages/ai/test/fixtures/recordings/openai-responses-websocket/reconstructs-full-context-after-reconnect.json

@@ -30,7 +30,7 @@
         {
           "direction": "client",
           "kind": "text",
-          "body": "{\"type\":\"response.create\",\"model\":\"gpt-5.5\",\"input\":[{\"type\":\"message\",\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"Reply exactly: Alpha.\"}]}],\"instructions\":\"Follow the user's exact reply instruction.\",\"store\":false,\"include\":[\"reasoning.encrypted_content\"],\"reasoning\":{\"effort\":\"medium\",\"summary\":\"auto\"},\"text\":{\"verbosity\":\"low\"},\"max_output_tokens\":30}"
+          "body": "{\"type\":\"response.create\",\"model\":\"gpt-5.5\",\"input\":[{\"type\":\"message\",\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"Reply exactly: Alpha.\"}]}],\"instructions\":\"Follow the user's exact reply instruction.\",\"store\":false,\"include\":[\"reasoning.encrypted_content\"],\"reasoning\":{\"effort\":\"medium\",\"summary\":\"auto\"},\"max_output_tokens\":30}"
         },
         {
           "direction": "server",
@@ -109,7 +109,7 @@
         {
           "direction": "client",
           "kind": "text",
-          "body": "{\"type\":\"response.create\",\"model\":\"gpt-5.5\",\"input\":[{\"type\":\"message\",\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"Reply exactly: Alpha.\"}]},{\"type\":\"reasoning\",\"id\":\"rs_0a6277dd90b94da1016aa8c946e33487d1b725d8e9dc874d82\",\"summary\":[],\"encrypted_content\":\"gAAAAABqqMlHtG6yltzgW_UjUcYxvl2hoMkk7cSEH5SJMe9CR5gKIaCwCh4peUo8XZrd-EU-tPCXthv7JzHxYXGVG1fIgTJ7BjBoOh-jgd_oGlyCHj2jVPj7nVoB763tZHdyw_ovL3V7GpJ6VeLGIsRGqNTnBz47ZuKikTKrbTDupn6fT2wOM_69zwdg4-UVnoF2J9E_jIS0E5XRtRwOidqANl61HCwo7LV-Ut1aqCbXb-59vkrVGWPxD_8n4smf7Qjywc0FCH9zwuEDX4cPdqj3MmjvYPrn0jand9LKkAy9rblaKFJSQfVuritbHrdQrnF7hLxu7QCpehOzNXOkpNEqTmXnAZjfMc53hq-ahH6_KoyYlZpompwyGPngFeI2fKRqP8rlpDilD1BHBsh-bl7kXzI8HQ_jameXwPZ1La6gjtFkThXp53BD6BOx11SB9Nypqolu2at5rR32UYcGrYeGtTu-HmYGp0oHVFkHumTNuHKmGXV14dI9swgryfygLhX8EAJrWrjm3e8rvAkHKpAZ0IkmlCcp15UCFDKNeS560fQVRaKXWnQ7m0Ih3C1xG0ifJ89j27c8GHo1kAhEJk-lSB1-FZr9Ls_w7N772kmZ2a7LLswu3kNW78kPas_CtcOnBOHwE1DhcVh5YpxwNftOHZnK1v8NPNF8EWqio4ArZy1thMCzH5zXBNcFzgd8tePMFukBblSP8QGQoVYRqTne2sFoOZXjslXnDvEe-Ycj8X38zWgRiwAk7guOloFC7Se36KCDP357773Vah86gWCt55mSEyhVW_GF1oTuHvJ18GZsXcyN21scF1PSr8YaCM_jR7ZkU1GYXbQK2Y4oTAV9XDptQA5YzEREnn7muC_6v5ZTAglZF1lhn9Q0NwmylZEAXJdSGHaqXt1Hv-vQlprA_9m22vrreBOTLPnVK946J8absKrwfe-jK_1n_9YQR43uwH8XwFBFND0c4lICCQGbxwM8pX4ACWR0c19aORCYm-M5FrJsxmG29_aDVNhcvkoQ3mlP7ITQeqzkrjfytSwLb2BYpXYZKjEHNfV9j3JoxJobUkK5hrxXBhTvZzbVBnE0LSXQNwR-JAcOliP_jXBEeQ-28B-aGW8TI1vEP2i260QKgzOzPC-pOoFp0-vCvxojNO0kE16ECVTLfDNLNGSMZs9vekdLJP2akBp6PsUXsQUTbWO_wr1E2oNU7ctfMRxoh-yP0ZW2_xz_NjE72O5LF_6zuDVV7Q==\"},{\"type\":\"message\",\"id\":\"msg_0a6277dd90b94da1016aa8c947253887d184c150fcbcbcd8da\",\"role\":\"assistant\",\"content\":[{\"type\":\"output_text\",\"text\":\"Alpha.\"}],\"phase\":\"final_answer\",\"status\":\"completed\"},{\"type\":\"message\",\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"Reply exactly: Beta.\"}]}],\"instructions\":\"Follow the user's exact reply instruction.\",\"store\":false,\"include\":[\"reasoning.encrypted_content\"],\"reasoning\":{\"effort\":\"medium\",\"summary\":\"auto\"},\"text\":{\"verbosity\":\"low\"},\"max_output_tokens\":30}"
+          "body": "{\"type\":\"response.create\",\"model\":\"gpt-5.5\",\"input\":[{\"type\":\"message\",\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"Reply exactly: Alpha.\"}]},{\"type\":\"reasoning\",\"id\":\"rs_0a6277dd90b94da1016aa8c946e33487d1b725d8e9dc874d82\",\"summary\":[],\"encrypted_content\":\"gAAAAABqqMlHtG6yltzgW_UjUcYxvl2hoMkk7cSEH5SJMe9CR5gKIaCwCh4peUo8XZrd-EU-tPCXthv7JzHxYXGVG1fIgTJ7BjBoOh-jgd_oGlyCHj2jVPj7nVoB763tZHdyw_ovL3V7GpJ6VeLGIsRGqNTnBz47ZuKikTKrbTDupn6fT2wOM_69zwdg4-UVnoF2J9E_jIS0E5XRtRwOidqANl61HCwo7LV-Ut1aqCbXb-59vkrVGWPxD_8n4smf7Qjywc0FCH9zwuEDX4cPdqj3MmjvYPrn0jand9LKkAy9rblaKFJSQfVuritbHrdQrnF7hLxu7QCpehOzNXOkpNEqTmXnAZjfMc53hq-ahH6_KoyYlZpompwyGPngFeI2fKRqP8rlpDilD1BHBsh-bl7kXzI8HQ_jameXwPZ1La6gjtFkThXp53BD6BOx11SB9Nypqolu2at5rR32UYcGrYeGtTu-HmYGp0oHVFkHumTNuHKmGXV14dI9swgryfygLhX8EAJrWrjm3e8rvAkHKpAZ0IkmlCcp15UCFDKNeS560fQVRaKXWnQ7m0Ih3C1xG0ifJ89j27c8GHo1kAhEJk-lSB1-FZr9Ls_w7N772kmZ2a7LLswu3kNW78kPas_CtcOnBOHwE1DhcVh5YpxwNftOHZnK1v8NPNF8EWqio4ArZy1thMCzH5zXBNcFzgd8tePMFukBblSP8QGQoVYRqTne2sFoOZXjslXnDvEe-Ycj8X38zWgRiwAk7guOloFC7Se36KCDP357773Vah86gWCt55mSEyhVW_GF1oTuHvJ18GZsXcyN21scF1PSr8YaCM_jR7ZkU1GYXbQK2Y4oTAV9XDptQA5YzEREnn7muC_6v5ZTAglZF1lhn9Q0NwmylZEAXJdSGHaqXt1Hv-vQlprA_9m22vrreBOTLPnVK946J8absKrwfe-jK_1n_9YQR43uwH8XwFBFND0c4lICCQGbxwM8pX4ACWR0c19aORCYm-M5FrJsxmG29_aDVNhcvkoQ3mlP7ITQeqzkrjfytSwLb2BYpXYZKjEHNfV9j3JoxJobUkK5hrxXBhTvZzbVBnE0LSXQNwR-JAcOliP_jXBEeQ-28B-aGW8TI1vEP2i260QKgzOzPC-pOoFp0-vCvxojNO0kE16ECVTLfDNLNGSMZs9vekdLJP2akBp6PsUXsQUTbWO_wr1E2oNU7ctfMRxoh-yP0ZW2_xz_NjE72O5LF_6zuDVV7Q==\"},{\"type\":\"message\",\"id\":\"msg_0a6277dd90b94da1016aa8c947253887d184c150fcbcbcd8da\",\"role\":\"assistant\",\"content\":[{\"type\":\"output_text\",\"text\":\"Alpha.\"}],\"phase\":\"final_answer\",\"status\":\"completed\"},{\"type\":\"message\",\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"Reply exactly: Beta.\"}]}],\"instructions\":\"Follow the user's exact reply instruction.\",\"store\":false,\"include\":[\"reasoning.encrypted_content\"],\"reasoning\":{\"effort\":\"medium\",\"summary\":\"auto\"},\"max_output_tokens\":30}"
         },
         {
           "direction": "server",

packages/ai/test/fixtures/recordings/openai-responses-websocket/recovers-from-explicit-continuation-rejection.json

@@ -30,7 +30,7 @@
         {
           "direction": "client",
           "kind": "text",
-          "body": "{\"type\":\"response.create\",\"model\":\"gpt-5.5\",\"input\":[{\"type\":\"message\",\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"Reply exactly: Ready.\"}]}],\"instructions\":\"Follow the user's exact reply instruction.\",\"store\":false,\"include\":[\"reasoning.encrypted_content\"],\"reasoning\":{\"effort\":\"medium\",\"summary\":\"auto\"},\"text\":{\"verbosity\":\"low\"},\"max_output_tokens\":30}"
+          "body": "{\"type\":\"response.create\",\"model\":\"gpt-5.5\",\"input\":[{\"type\":\"message\",\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"Reply exactly: Ready.\"}]}],\"instructions\":\"Follow the user's exact reply instruction.\",\"store\":false,\"include\":[\"reasoning.encrypted_content\"],\"reasoning\":{\"effort\":\"medium\",\"summary\":\"auto\"},\"max_output_tokens\":30}"
         },
         {
           "direction": "server",
@@ -109,7 +109,7 @@
         {
           "direction": "client",
           "kind": "text",
-          "body": "{\"type\":\"response.create\",\"model\":\"gpt-5.5\",\"input\":[{\"type\":\"message\",\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"Reply exactly: Recovered.\"}]}],\"instructions\":\"Follow the user's exact reply instruction.\",\"store\":false,\"include\":[\"reasoning.encrypted_content\"],\"reasoning\":{\"effort\":\"medium\",\"summary\":\"auto\"},\"text\":{\"verbosity\":\"low\"},\"max_output_tokens\":30,\"previous_response_id\":\"resp_01cc0cda24c36acf016aa8ca3c1d3c87d1853283f43675e411\"}"
+          "body": "{\"type\":\"response.create\",\"model\":\"gpt-5.5\",\"input\":[{\"type\":\"message\",\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"Reply exactly: Recovered.\"}]}],\"instructions\":\"Follow the user's exact reply instruction.\",\"store\":false,\"include\":[\"reasoning.encrypted_content\"],\"reasoning\":{\"effort\":\"medium\",\"summary\":\"auto\"},\"max_output_tokens\":30,\"previous_response_id\":\"resp_01cc0cda24c36acf016aa8ca3c1d3c87d1853283f43675e411\"}"
         },
         {
           "direction": "server",
@@ -119,7 +119,7 @@
         {
           "direction": "client",
           "kind": "text",
-          "body": "{\"type\":\"response.create\",\"model\":\"gpt-5.5\",\"input\":[{\"type\":\"message\",\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"Reply exactly: Ready.\"}]},{\"type\":\"reasoning\",\"id\":\"rs_01cc0cda24c36acf016aa8ca3ced8c87d1a14c5c6a2ced8544\",\"summary\":[],\"encrypted_content\":\"gAAAAABqqMo9M9B15LsV1CLsXpNJpI08GiwCAK97hTheQ2s1lAkKgARMs1HIIXVeAr-tbUq88yN51fiQE8HvGwdF9ZcNI6bZle8D2PS7m8O7UE7C1Z_HX_fz2f0pRHyRfWpVAT2gYQWIDeZu6UdgxbKPBW0lXWzlk_relHG5x6nXkYzZoEeVasqavWoyMMSX7cexe-IYJh6e_3DgRpOchueS8Z-70P0w5R83Ea7UXZQNhMA0yDEYu_td2PmE2Pd2PUTOB3mxF2pb1z7-2t6S0UryhHx0az7Gh2eT60GGUqz9CIZzNE_FX--tszeuO0eI92Cen5tirOUHBTyyDqE0eG26DRl_p_U-xDZwaQONbtYbvkvrj-G7FA2oZXxjJPHuQZsNgBslXS-H0KT1lx3Y8XJ9QMVjFLFaucFG64wCmXPfCH8dYtX_YqfYQR4lwNfiSbyJEX2oDvTVVD_aCJ9NRo7c0aCTtmKBvr6fvvAy3MAFxAp_Sm2nMx4P5GYO4qAmJDByywKw-VK1vHlv3NRmVsAgbArIFgm-axoCs2PLpvZjDqeQGPavaq8zKWTyZYqBsEzKZUtGOZfYjD4mud0Z08I4i2H4K-L00ccVauode3548ZipOIuslbhJxonQXsF6TFdW2Hj8E5JjoEr5IbmwHyI0PBcDWW5AmkjHLwr9v08mFppRoD-2wzPAd5igROAuUbvJhiQd2A-uOaohwMjdpFxrjyUqgGTlI5g7tmI1ceeQWms0bKm8Pd0wIqVM3Nq6YvV7XfEyeogfRMVUQezr_lES42ZMVAoKBSFzMysDwCFkhVNVclcTUpcUUbVp21FChG7Ag-xuq8Cl6OGLA8nWX1C0aCf2HNa-n3dkYr1DtUziurh1MD-UIs5jdGiq3ptrc0VaVZwNdD4jVfAoHB_Ws7GiISXuclfpqsG3DTJEfzlbukI1vxXrt3FArsHiQvQjW5UM7gGel32M6p8AlXRxnez9PgIuU1WrtBUJetk7m39AZwp_aqbqC-AJ-MF70xP1VJZwFN-GeNL3VZsRHePFG4h7Pj---CCZRGlmzuE1-b-sIE7Bn_gbue_qHFMZJhAY55MO25vfZbSoLWHZCGmLMjJVHOYCPoy6l7zyxNmoIcC58QILNWal31KLCDmCsASmZC-xQRjyFwt-kvLbvk38Dc02IKcP3ujhf6WRRr1A0hh1K-gXmv_XI_MUFDcOVIjjhB-rXgjWSCoKfaKI9GCIsb9mnNaBq65BWg==\"},{\"type\":\"message\",\"id\":\"msg_01cc0cda24c36acf016aa8ca3d4a0087d1950121aed4802434\",\"role\":\"assistant\",\"content\":[{\"type\":\"output_text\",\"text\":\"Ready.\"}],\"phase\":\"final_answer\",\"status\":\"completed\"},{\"type\":\"message\",\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"Reply exactly: Recovered.\"}]}],\"instructions\":\"Follow the user's exact reply instruction.\",\"store\":false,\"include\":[\"reasoning.encrypted_content\"],\"reasoning\":{\"effort\":\"medium\",\"summary\":\"auto\"},\"text\":{\"verbosity\":\"low\"},\"max_output_tokens\":30}"
+          "body": "{\"type\":\"response.create\",\"model\":\"gpt-5.5\",\"input\":[{\"type\":\"message\",\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"Reply exactly: Ready.\"}]},{\"type\":\"reasoning\",\"id\":\"rs_01cc0cda24c36acf016aa8ca3ced8c87d1a14c5c6a2ced8544\",\"summary\":[],\"encrypted_content\":\"gAAAAABqqMo9M9B15LsV1CLsXpNJpI08GiwCAK97hTheQ2s1lAkKgARMs1HIIXVeAr-tbUq88yN51fiQE8HvGwdF9ZcNI6bZle8D2PS7m8O7UE7C1Z_HX_fz2f0pRHyRfWpVAT2gYQWIDeZu6UdgxbKPBW0lXWzlk_relHG5x6nXkYzZoEeVasqavWoyMMSX7cexe-IYJh6e_3DgRpOchueS8Z-70P0w5R83Ea7UXZQNhMA0yDEYu_td2PmE2Pd2PUTOB3mxF2pb1z7-2t6S0UryhHx0az7Gh2eT60GGUqz9CIZzNE_FX--tszeuO0eI92Cen5tirOUHBTyyDqE0eG26DRl_p_U-xDZwaQONbtYbvkvrj-G7FA2oZXxjJPHuQZsNgBslXS-H0KT1lx3Y8XJ9QMVjFLFaucFG64wCmXPfCH8dYtX_YqfYQR4lwNfiSbyJEX2oDvTVVD_aCJ9NRo7c0aCTtmKBvr6fvvAy3MAFxAp_Sm2nMx4P5GYO4qAmJDByywKw-VK1vHlv3NRmVsAgbArIFgm-axoCs2PLpvZjDqeQGPavaq8zKWTyZYqBsEzKZUtGOZfYjD4mud0Z08I4i2H4K-L00ccVauode3548ZipOIuslbhJxonQXsF6TFdW2Hj8E5JjoEr5IbmwHyI0PBcDWW5AmkjHLwr9v08mFppRoD-2wzPAd5igROAuUbvJhiQd2A-uOaohwMjdpFxrjyUqgGTlI5g7tmI1ceeQWms0bKm8Pd0wIqVM3Nq6YvV7XfEyeogfRMVUQezr_lES42ZMVAoKBSFzMysDwCFkhVNVclcTUpcUUbVp21FChG7Ag-xuq8Cl6OGLA8nWX1C0aCf2HNa-n3dkYr1DtUziurh1MD-UIs5jdGiq3ptrc0VaVZwNdD4jVfAoHB_Ws7GiISXuclfpqsG3DTJEfzlbukI1vxXrt3FArsHiQvQjW5UM7gGel32M6p8AlXRxnez9PgIuU1WrtBUJetk7m39AZwp_aqbqC-AJ-MF70xP1VJZwFN-GeNL3VZsRHePFG4h7Pj---CCZRGlmzuE1-b-sIE7Bn_gbue_qHFMZJhAY55MO25vfZbSoLWHZCGmLMjJVHOYCPoy6l7zyxNmoIcC58QILNWal31KLCDmCsASmZC-xQRjyFwt-kvLbvk38Dc02IKcP3ujhf6WRRr1A0hh1K-gXmv_XI_MUFDcOVIjjhB-rXgjWSCoKfaKI9GCIsb9mnNaBq65BWg==\"},{\"type\":\"message\",\"id\":\"msg_01cc0cda24c36acf016aa8ca3d4a0087d1950121aed4802434\",\"role\":\"assistant\",\"content\":[{\"type\":\"output_text\",\"text\":\"Ready.\"}],\"phase\":\"final_answer\",\"status\":\"completed\"},{\"type\":\"message\",\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"Reply exactly: Recovered.\"}]}],\"instructions\":\"Follow the user's exact reply instruction.\",\"store\":false,\"include\":[\"reasoning.encrypted_content\"],\"reasoning\":{\"effort\":\"medium\",\"summary\":\"auto\"},\"max_output_tokens\":30}"
         },
         {
           "direction": "server",

packages/ai/test/fixtures/recordings/openai-responses/openai-responses-gpt-5-5-reasoning-continuation.json

@@ -26,7 +26,7 @@
         "headers": {
           "content-type": "application/json"
         },
-        "body": "{\"model\":\"gpt-5.5\",\"input\":[{\"type\":\"message\",\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"Think briefly, then reply exactly with: Hello!\"}]}],\"instructions\":\"Show concise reasoning when the provider supports visible reasoning summaries.\",\"store\":false,\"include\":[\"reasoning.encrypted_content\"],\"reasoning\":{\"effort\":\"low\",\"summary\":\"auto\"},\"text\":{\"verbosity\":\"low\"},\"max_output_tokens\":120,\"stream\":true}"
+        "body": "{\"model\":\"gpt-5.5\",\"input\":[{\"type\":\"message\",\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"Think briefly, then reply exactly with: Hello!\"}]}],\"instructions\":\"Show concise reasoning when the provider supports visible reasoning summaries.\",\"store\":false,\"include\":[\"reasoning.encrypted_content\"],\"reasoning\":{\"effort\":\"low\",\"summary\":\"auto\"},\"max_output_tokens\":120,\"stream\":true}"
       },
       "response": {
         "status": 200,
@@ -44,7 +44,7 @@
         "headers": {
           "content-type": "application/json"
         },
-        "body": "{\"model\":\"gpt-5.5\",\"input\":[{\"type\":\"message\",\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"Think briefly, then reply exactly with: Hello!\"}]},{\"type\":\"reasoning\",\"id\":\"rs_052e7ec551f55289016aa8c8d63eac87d19d6b921611f123e7\",\"summary\":[],\"encrypted_content\":\"gAAAAABqqMjWZ2Eei8_Gf-5FeEFAYp-gSzFL4D4lQBKL_fyyTYXv5iJ-2jql1mOq0wZpqHL8O9MWxebQGW56Ahd-p21qrDD52CyUBqKKlF87eC1d-cTgjXQlFMsPxvwyQeuU6A2l8tanTtJ48sKtzZtHrDuBXZ35u-lONnovjFGMX3Q83xoqG_um_w5rT420TA_SyU4fGt7oiQvOPS1q4PNo97O824oRnI7n_BC1jPYCaJhl2I1rPJg4afuOpjG-u7JcXRD4JPwZdqMa5o2d0uDKHuUYwP25qiPKKDqTTqFka5cDJjZNPF3ZkVHR-cagjZGvMnizXXxgUpPJ9j83gqY4QJLKCkzcaBj9H7mAL-v9yl4I5kn_9_DhpMILs2SZkC8AvIYNgmel3sDV_BG4XZ2JXciZz86ukQ6DwXgQqS4HOB91g-sGHOWMU1ohsZlEvvJBjGkJ_rAdXVMqAbvi2zvE3_NI4sTAGUrIugGJePrQYTe8gqL8f9NsYac6pzHNQL1e_jQNUvp49bu7EsPzCP3KPYVvZCFohDdwe7sMd6wrztrCJwM4CLdAQK61A7sYzU0HyglLtPidmS5QkSmV6U_xgih7JbKnY0oAeCyYw4ADYqdNTi0axmBErh-lbh-XKNG_TnoMa-2IS3X041N8OsDfSdQp3QsAm53fF8seQuLFa27Iaq2etMj3yGeqWVjA-Mae3K34mt2YPGjQ-HbIOMVmYXBLzNr-s2fT35Sp6SDEsFyvzXb0Vij58s1wW5zkuKgaJiroGgkY86NImuaa3_-wpMK3_9O_wwAbRwV4uBCVzT_rY6rDQHR9-VkM0MbGK8drbdtjXwy3KtAzkux4N4g2nadYU0IIIEUNj_JChUFSHC7VRg7L7LpZMgAFGwHmaUyzQt31LyiVix9WFtcKfBgzehoRV6vstln-oBRd-vFjUW-7WLm7R_lFNHQZ0CKUvKCSpQxdIevczpYT0_lQiDU7Rfp1UBBnicndpq4YQwgRppdZX-QHG5IZxxHNWYdIBtn3eO3ooDXmI-rYryyVcFG6VY5xqssf8GgXqgNy84X_SgYfhGmNiPPQTl5wIBa49f1sxAI-7ri1xLr6FJhQMasOoxm_VuZYNkq6NrAXprtKa70cBRpm1wWBNcPdMul4GNMDQatQZRDENcNLe45iMX-HB1YYSvyc-y5rfSJkP4EgKXC-OqcrzxAlRfB3Bm4jOQL_PSsNBbh26CWAdAYSftM88fbnuqZX2w==\"},{\"type\":\"message\",\"role\":\"assistant\",\"content\":[{\"type\":\"output_text\",\"text\":\"Hello!\"}],\"status\":\"completed\"},{\"type\":\"message\",\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"Now reply exactly with: Done.\"}]}],\"store\":false,\"include\":[\"reasoning.encrypted_content\"],\"reasoning\":{\"effort\":\"low\",\"summary\":\"auto\"},\"text\":{\"verbosity\":\"low\"},\"max_output_tokens\":40,\"stream\":true}"
+        "body": "{\"model\":\"gpt-5.5\",\"input\":[{\"type\":\"message\",\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"Think briefly, then reply exactly with: Hello!\"}]},{\"type\":\"reasoning\",\"id\":\"rs_052e7ec551f55289016aa8c8d63eac87d19d6b921611f123e7\",\"summary\":[],\"encrypted_content\":\"gAAAAABqqMjWZ2Eei8_Gf-5FeEFAYp-gSzFL4D4lQBKL_fyyTYXv5iJ-2jql1mOq0wZpqHL8O9MWxebQGW56Ahd-p21qrDD52CyUBqKKlF87eC1d-cTgjXQlFMsPxvwyQeuU6A2l8tanTtJ48sKtzZtHrDuBXZ35u-lONnovjFGMX3Q83xoqG_um_w5rT420TA_SyU4fGt7oiQvOPS1q4PNo97O824oRnI7n_BC1jPYCaJhl2I1rPJg4afuOpjG-u7JcXRD4JPwZdqMa5o2d0uDKHuUYwP25qiPKKDqTTqFka5cDJjZNPF3ZkVHR-cagjZGvMnizXXxgUpPJ9j83gqY4QJLKCkzcaBj9H7mAL-v9yl4I5kn_9_DhpMILs2SZkC8AvIYNgmel3sDV_BG4XZ2JXciZz86ukQ6DwXgQqS4HOB91g-sGHOWMU1ohsZlEvvJBjGkJ_rAdXVMqAbvi2zvE3_NI4sTAGUrIugGJePrQYTe8gqL8f9NsYac6pzHNQL1e_jQNUvp49bu7EsPzCP3KPYVvZCFohDdwe7sMd6wrztrCJwM4CLdAQK61A7sYzU0HyglLtPidmS5QkSmV6U_xgih7JbKnY0oAeCyYw4ADYqdNTi0axmBErh-lbh-XKNG_TnoMa-2IS3X041N8OsDfSdQp3QsAm53fF8seQuLFa27Iaq2etMj3yGeqWVjA-Mae3K34mt2YPGjQ-HbIOMVmYXBLzNr-s2fT35Sp6SDEsFyvzXb0Vij58s1wW5zkuKgaJiroGgkY86NImuaa3_-wpMK3_9O_wwAbRwV4uBCVzT_rY6rDQHR9-VkM0MbGK8drbdtjXwy3KtAzkux4N4g2nadYU0IIIEUNj_JChUFSHC7VRg7L7LpZMgAFGwHmaUyzQt31LyiVix9WFtcKfBgzehoRV6vstln-oBRd-vFjUW-7WLm7R_lFNHQZ0CKUvKCSpQxdIevczpYT0_lQiDU7Rfp1UBBnicndpq4YQwgRppdZX-QHG5IZxxHNWYdIBtn3eO3ooDXmI-rYryyVcFG6VY5xqssf8GgXqgNy84X_SgYfhGmNiPPQTl5wIBa49f1sxAI-7ri1xLr6FJhQMasOoxm_VuZYNkq6NrAXprtKa70cBRpm1wWBNcPdMul4GNMDQatQZRDENcNLe45iMX-HB1YYSvyc-y5rfSJkP4EgKXC-OqcrzxAlRfB3Bm4jOQL_PSsNBbh26CWAdAYSftM88fbnuqZX2w==\"},{\"type\":\"message\",\"role\":\"assistant\",\"content\":[{\"type\":\"output_text\",\"text\":\"Hello!\"}],\"status\":\"completed\"},{\"type\":\"message\",\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"Now reply exactly with: Done.\"}]}],\"store\":false,\"include\":[\"reasoning.encrypted_content\"],\"reasoning\":{\"effort\":\"low\",\"summary\":\"auto\"},\"max_output_tokens\":40,\"stream\":true}"
       },
       "response": {
         "status": 200,

packages/ai/test/fixtures/recordings/openai-responses/openai-responses-gpt-5-5-reasoning.json

@@ -24,7 +24,7 @@
         "headers": {
           "content-type": "application/json"
         },
-        "body": "{\"model\":\"gpt-5.5\",\"input\":[{\"type\":\"message\",\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"Think briefly, then reply exactly with: Hello!\"}]}],\"instructions\":\"Show concise reasoning when the provider supports visible reasoning summaries.\",\"store\":false,\"include\":[\"reasoning.encrypted_content\"],\"reasoning\":{\"effort\":\"low\",\"summary\":\"auto\"},\"text\":{\"verbosity\":\"low\"},\"max_output_tokens\":120,\"stream\":true}"
+        "body": "{\"model\":\"gpt-5.5\",\"input\":[{\"type\":\"message\",\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"Think briefly, then reply exactly with: Hello!\"}]}],\"instructions\":\"Show concise reasoning when the provider supports visible reasoning summaries.\",\"store\":false,\"include\":[\"reasoning.encrypted_content\"],\"reasoning\":{\"effort\":\"low\",\"summary\":\"auto\"},\"max_output_tokens\":120,\"stream\":true}"
       },
       "response": {
         "status": 200,

packages/ai/test/fixtures/recordings/openai-responses/openai-responses-gpt-5-5-tool-loop.json

@@ -25,7 +25,7 @@
         "headers": {
           "content-type": "application/json"
         },
-        "body": "{\"model\":\"gpt-5.5\",\"input\":[{\"type\":\"message\",\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"What is the weather in Paris?\"}]}],\"instructions\":\"Use the get_weather tool exactly once. After the tool result, reply exactly: Paris is sunny.\",\"tools\":[{\"type\":\"function\",\"name\":\"get_weather\",\"description\":\"Get current weather for a city.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"city\":{\"type\":\"string\"}},\"required\":[\"city\"],\"additionalProperties\":false},\"strict\":false}],\"store\":false,\"include\":[\"reasoning.encrypted_content\"],\"reasoning\":{\"effort\":\"medium\",\"summary\":\"auto\"},\"text\":{\"verbosity\":\"low\"},\"max_output_tokens\":80,\"stream\":true}"
+        "body": "{\"model\":\"gpt-5.5\",\"input\":[{\"type\":\"message\",\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"What is the weather in Paris?\"}]}],\"instructions\":\"Use the get_weather tool exactly once. After the tool result, reply exactly: Paris is sunny.\",\"tools\":[{\"type\":\"function\",\"name\":\"get_weather\",\"description\":\"Get current weather for a city.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"city\":{\"type\":\"string\"}},\"required\":[\"city\"],\"additionalProperties\":false},\"strict\":false}],\"store\":false,\"include\":[\"reasoning.encrypted_content\"],\"reasoning\":{\"effort\":\"medium\",\"summary\":\"auto\"},\"max_output_tokens\":80,\"stream\":true}"
       },
       "response": {
         "status": 200,
@@ -43,7 +43,7 @@
         "headers": {
           "content-type": "application/json"
         },
-        "body": "{\"model\":\"gpt-5.5\",\"input\":[{\"type\":\"message\",\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"What is the weather in Paris?\"}]},{\"type\":\"function_call\",\"id\":\"fc_09525c04931d1487016aa8c8d8e12887d193bd327a0f313bd7\",\"call_id\":\"call_p57PJbKe0bX44nj908fpHdRn\",\"name\":\"get_weather\",\"arguments\":\"{\\\"city\\\":\\\"Paris\\\"}\"},{\"type\":\"function_call_output\",\"call_id\":\"call_p57PJbKe0bX44nj908fpHdRn\",\"output\":\"{\\\"temperature\\\":22,\\\"condition\\\":\\\"sunny\\\"}\"}],\"instructions\":\"Use the get_weather tool exactly once. After the tool result, reply exactly: Paris is sunny.\",\"tools\":[{\"type\":\"function\",\"name\":\"get_weather\",\"description\":\"Get current weather for a city.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"city\":{\"type\":\"string\"}},\"required\":[\"city\"],\"additionalProperties\":false},\"strict\":false}],\"store\":false,\"include\":[\"reasoning.encrypted_content\"],\"reasoning\":{\"effort\":\"medium\",\"summary\":\"auto\"},\"text\":{\"verbosity\":\"low\"},\"max_output_tokens\":80,\"stream\":true}"
+        "body": "{\"model\":\"gpt-5.5\",\"input\":[{\"type\":\"message\",\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"What is the weather in Paris?\"}]},{\"type\":\"function_call\",\"id\":\"fc_09525c04931d1487016aa8c8d8e12887d193bd327a0f313bd7\",\"call_id\":\"call_p57PJbKe0bX44nj908fpHdRn\",\"name\":\"get_weather\",\"arguments\":\"{\\\"city\\\":\\\"Paris\\\"}\"},{\"type\":\"function_call_output\",\"call_id\":\"call_p57PJbKe0bX44nj908fpHdRn\",\"output\":\"{\\\"temperature\\\":22,\\\"condition\\\":\\\"sunny\\\"}\"}],\"instructions\":\"Use the get_weather tool exactly once. After the tool result, reply exactly: Paris is sunny.\",\"tools\":[{\"type\":\"function\",\"name\":\"get_weather\",\"description\":\"Get current weather for a city.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"city\":{\"type\":\"string\"}},\"required\":[\"city\"],\"additionalProperties\":false},\"strict\":false}],\"store\":false,\"include\":[\"reasoning.encrypted_content\"],\"reasoning\":{\"effort\":\"medium\",\"summary\":\"auto\"},\"max_output_tokens\":80,\"stream\":true}"
       },
       "response": {
         "status": 200,

packages/ai/test/provider/checkpoint.test.ts

@@ -22,21 +22,22 @@ testEffect(
       expect(body).toMatchObject({
         model: "fixture",
         stream: true,
-        store: false,
+        store: true,
         instructions: "Keep the context",
-        parallel_tool_calls: true,
+        parallel_tool_calls: false,
         prompt_cache_key: "session-key",
         service_tier: "priority",
         reasoning: { effort: "high", summary: "auto" },
+        context_management: [{ type: "compaction" }],
+        max_tool_calls: 1,
+        tool_choice: "required",
+        text: { verbosity: "high", format: { type: "json_object" } },
         prompt_cache_retention: "24h",
         prompt_cache_options: { mode: "session", ttl: "1h" },
         input: [{ role: "user", content: [{ type: "input_text", text: "hello" }] }, { type: "compaction_trigger" }],
       })
       expect(body.tools).toHaveLength(1)
       expect(body.tools[0].name).toBe("lookup")
-      expect(body.tool_choice).toBeUndefined()
-      expect(body.context_management).toBeUndefined()
-      expect(body.text).toBeUndefined()
       expect(body.max_output_tokens).toBeUndefined()
       expect(body.previous_response_id).toBeUndefined()
       return respond(
@@ -57,7 +58,7 @@ testEffect(
       )
     }),
   ),
-).effect("trigger uses normal request preparation, configured deployment, and supplied subscription headers", () =>
+).effect("trigger keeps request controls, configured deployment, and supplied subscription headers", () =>
   Effect.gen(function* () {
     const calls: string[] = []
     const input = LLM.request({
@@ -67,12 +68,14 @@ testEffect(
       promptCacheKey: "session-key",
       tools: [{ name: "lookup", description: "Lookup", inputSchema: { type: "object", properties: {} } }],
       toolChoice: { type: "tool", name: "lookup" },
-      generation: { maxTokens: 1 },
       providerOptions: {
         store: true,
         reasoningEffort: "high",
         reasoningSummary: "auto",
         contextManagement: [{ type: "compaction" }],
+        parallelToolCalls: false,
+        maxToolCalls: 1,
+        textVerbosity: "low",
       },
       http: {
         headers: { "chatgpt-account-id": "fixture-account", "x-codex-beta-features": "remote_compaction_v2" },
@@ -82,8 +85,7 @@ testEffect(
           prompt_cache_retention: "24h",
           prompt_cache_options: { mode: "session", ttl: "1h" },
           store: true,
-          stream: false,
-          text: { format: { type: "json_object" } },
+          text: { verbosity: "high", format: { type: "json_object" } },
           tool_choice: "required",
         },
       },
@@ -114,6 +116,75 @@ testEffect(
   }),
 )
 
+testEffect(
+  dynamicResponse(({ text, respond }) =>
+    Effect.sync(() => {
+      expect(JSON.parse(text).text).toEqual({ verbosity: "low", format: { type: "json_object" } })
+      return respond(sseEvents({ type: "response.completed", response: { id: "resp_1", output: [checkpoint] } }), {
+        headers: { "content-type": "text/event-stream" },
+      })
+    }),
+  ),
+).effect("keeps explicit verbosity on a trigger checkpoint for prompt cache reuse", () =>
+  LLMClient.compact(
+    LLM.request({
+      model: OpenAI.configure({ apiKey: "fixture" }).responses("gpt-5.5"),
+      prompt: "Hello.",
+      providerOptions: { textVerbosity: "low" },
+      http: { body: { text: { format: { type: "json_object" } } } },
+    }),
+    trigger,
+  ),
+)
+
+testEffect(
+  dynamicResponse(({ text, respond }) =>
+    Effect.sync(() => {
+      const body = JSON.parse(text)
+      expect(body.text).toEqual({ verbosity: "high", format: { type: "json_object" } })
+      expect(body.max_output_tokens).toBe(20_000)
+      return respond(sseEvents({ type: "response.completed", response: { id: "resp_1", output: [checkpoint] } }), {
+        headers: { "content-type": "text/event-stream" },
+      })
+    }),
+  ),
+).effect("keeps the effective body-overlay verbosity and text formatting", () =>
+  LLMClient.compact(
+    LLM.request({
+      model: OpenAI.configure({ apiKey: "fixture" }).responses("gpt-5.5"),
+      prompt: "Hello.",
+      generation: { maxTokens: 20_000 },
+      providerOptions: { textVerbosity: "low" },
+      http: { body: { text: { verbosity: "high", format: { type: "json_object" } } } },
+    }),
+    trigger,
+  ),
+)
+
+testEffect(
+  dynamicResponse(({ text, respond }) =>
+    Effect.sync(() => {
+      expect(JSON.parse(text).max_output_tokens).toBe(128)
+      return respond(JSON.stringify({ error: { message: "max_output_tokens must be at least 20000" } }), {
+        status: 400,
+        headers: { "content-type": "application/json" },
+      })
+    }),
+  ),
+).effect("passes configured output limits through and leaves rejection to the provider", () =>
+  Effect.gen(function* () {
+    const error = yield* LLMClient.compact(
+      LLM.request({
+        model: OpenAI.configure({ apiKey: "fixture" }).responses("gpt-5.5"),
+        prompt: "Hello.",
+        generation: { maxTokens: 128 },
+      }),
+      trigger,
+    ).pipe(Effect.flip)
+    expect(error.message).toContain("at least 20000")
+  }),
+)
+
 const idless = { type: "compaction", encrypted_content: "opaque" }
 testEffect(
   fixedResponse(
@@ -184,7 +255,7 @@ testEffect(fixedResponse(sseEvents({ type: "response.output_item.done", item: ch
       expect(error.reason._tag).toBe("InvalidProviderOutput")
     }),
 )
-for (const body of [{ input: [] }, { previous_response_id: "stale" }]) {
+for (const body of [{ input: [] }, { previous_response_id: "stale" }, { stream: false }]) {
   testEffect(dynamicResponse(() => Effect.die("Must reject before sending"))).effect(
     `rejects caller-supplied ${Object.keys(body)[0]} before sending trigger`,
     () =>

packages/ai/test/provider/explicit-compaction.test.ts

@@ -110,11 +110,16 @@ for (const model of [
     dynamicResponse(({ request, text, respond }) =>
       Effect.sync(() => {
         expect(new URL(request.url).pathname).toEndWith("/responses/compact")
-        expect(JSON.parse(text)).toEqual({ model: "fixture", input: [item], instructions: "Keep the context" })
+        expect(JSON.parse(text)).toEqual({
+          model: "fixture",
+          input: [item],
+          instructions: "Keep the context",
+          include: ["reasoning.encrypted_content"],
+        })
         return respond(JSON.stringify({ object: "response.compaction", output: [checkpoint] }))
       }),
     ),
-  ).effect(`${model.provider} compacts provider-specific history without lowering generation settings`, () =>
+  ).effect(`${model.provider} validates tools but ignores unrelated unsupported generation settings`, () =>
     Effect.gen(function* () {
       const request = LLM.request({
         model,
@@ -151,6 +156,11 @@ for (const model of [
       ] as const) {
         const error = yield* LLMClient.generate(candidate).pipe(Effect.flip)
         expect(error.reason._tag).toBe(tag)
+        if (candidate.tools.length > 0) {
+          const compactError = yield* LLMClient.compact(candidate).pipe(Effect.flip)
+          expect(compactError.reason._tag).toBe("InvalidRequest")
+          continue
+        }
         const response = yield* LLMClient.compact(candidate)
         expect(response.replacement[0]?.content[0]?.type).toBe("compaction")
       }
@@ -255,6 +265,13 @@ for (const overlay of [undefined, { service_tier: "priority", prompt_cache_key:
           model: "fixture",
           input: [{ type: "message", role: "user", content: [{ type: "input_text", text: "hello" }] }],
           service_tier: overlay?.service_tier ?? "flex",
+          reasoning: { effort: "low" },
+          text: { verbosity: "low", format: { type: "json_object" } },
+          include: ["reasoning.encrypted_content"],
+          parallel_tool_calls: false,
+          tools: [
+            { type: "function", name: "lookup", description: "Lookup", parameters: { type: "object" }, strict: false },
+          ],
           prompt_cache_key: overlay?.prompt_cache_key ?? "affinity",
           prompt_cache_retention: "24h",
           prompt_cache_options: { mode: "explicit", ttl: "30m" },
@@ -268,12 +285,20 @@ for (const overlay of [undefined, { service_tier: "priority", prompt_cache_key:
         model: OpenAI.configure({ apiKey: "test" }).responses("fixture"),
         prompt: "hello",
         promptCacheKey: "affinity",
-        providerOptions: { serviceTier: "flex" },
+        providerOptions: {
+          serviceTier: "flex",
+          reasoningEffort: "low",
+          textVerbosity: "low",
+          include: ["reasoning.encrypted_content"],
+          parallelToolCalls: false,
+        },
         generation: { maxTokens: 100 },
+        tools: [{ name: "lookup", description: "Lookup", inputSchema: {} }],
         http: {
           body: {
             stream: true,
             store: false,
+            text: { format: { type: "json_object" } },
             prompt_cache_retention: "24h",
             prompt_cache_options: { mode: "explicit", ttl: "30m" },
             ...overlay,
@@ -396,6 +421,8 @@ for (const model of [
             model: model.id,
             input: [{ type: "message", role: "user", content: [{ type: "input_text", text: "original" }] }],
             instructions: "system",
+            include: ["reasoning.encrypted_content"],
+            ...(model.id === "gpt-5.3-codex" ? { reasoning: { effort: "medium", summary: "auto" } } : {}),
           })
           return respond(
             JSON.stringify({
@@ -407,7 +434,10 @@ for (const model of [
           )
         }
         expect(new URL(request.url).pathname.endsWith("/responses")).toBe(true)
-        expect(body.input).toEqual([...output, { type: "message", role: "user", content: [{ type: "input_text", text: "continue" }] }])
+        expect(body.input).toEqual([
+          ...output,
+          { type: "message", role: "user", content: [{ type: "input_text", text: "continue" }] },
+        ])
         return respond(sseEvents({ type: "response.completed", response: { id: "resp_1", output: [] } }), {
           headers: { "content-type": "text/event-stream" },
         })

packages/ai/test/provider/openai-responses.test.ts

@@ -1945,7 +1945,7 @@ describe("OpenAI Responses route", () => {
       expect(prepared.body.prompt_cache_key).toBe("session_123")
       expect(prepared.body.include).toEqual(["reasoning.encrypted_content"])
       expect(prepared.body.reasoning).toEqual({ effort: "high", summary: "auto" })
-      expect(prepared.body.text).toEqual({ verbosity: "low" })
+      expect(prepared.body.text).toBeUndefined()
       expect(prepared.body.metadata).toEqual({ environment: "test", tenant: "acme" })
       expect(prepared.body.safety_identifier).toBe("user_123")
       expect(prepared.body.stream_options).toEqual({ include_obfuscation: false })

packages/core/src/plugin/internal.ts

@@ -101,6 +101,7 @@ import { VcsHgPlugin } from "./vcs/hg.js"
 import { ToolInputRepairPlugin } from "./tool-input-repair.js"
 import { OptimizePlugin } from "./optimize.js"
 import { VcsGitPlugin } from "./vcs/git.js"
+import { VerbosityPlugin } from "./verbosity.js"
 import { WarmingPlugin } from "./warming.js"
 import { WellKnownPlugin } from "../wellknown/plugin.js"
 
@@ -230,6 +231,7 @@ const pre = [
   PatchTool.Plugin,
   // Render model prompts after the patch plugin selects the available editing tools.
   ...OptimizePlugin.Plugins,
+  VerbosityPlugin.Plugin,
   IdentityPlugin.Plugin,
   EditTool.Plugin,
   GlobTool.Plugin,

packages/core/src/plugin/verbosity.ts

@@ -0,0 +1,54 @@
+export * as VerbosityPlugin from "./verbosity.js"
+
+import { define } from "@opencode/plugin/effect/plugin"
+import type { SessionRequest } from "@opencode/plugin/effect/session"
+import { Effect } from "effect"
+import { Model } from "../model.js"
+import { Provider } from "../provider.js"
+import type { PluginInternal } from "./internal.js"
+
+const direct = new Set([
+  "@opencode/ai/providers/openai",
+  "@opencode/ai/providers/openai/responses",
+  "@opencode/ai/providers/azure",
+  "@opencode/ai/providers/azure/responses",
+])
+const gateways = new Set(["@opencode/ai/providers/cloudflare-ai-gateway", Provider.aisdk("@ai-sdk/gateway")])
+
+export const Plugin = define({
+  id: "opencode.prompt.verbosity",
+  effect: Effect.fn("VerbosityPlugin")(function* (ctx) {
+    const models = yield* Model.Service
+    const hook = (event: SessionRequest) =>
+      Effect.gen(function* () {
+        if (event.options.textVerbosity !== undefined) return
+        const model = yield* models.get(event.model.providerID, event.model.id)
+        if (!model) return
+        const id = openAIModelID(model)
+        if (!id || !supportsVerbosity(id)) return
+        if (model.settings?.textVerbosity !== undefined) return
+        const variant = model.variants.find((item) => item.id === event.model.variant)
+        if (variant?.settings?.textVerbosity !== undefined) return
+        event.options.textVerbosity = "low"
+      })
+    yield* ctx.session.hook("context", hook)
+    yield* ctx.session.hook("compaction", hook)
+    yield* ctx.session.hook("generate", hook)
+    yield* ctx.session.hook("title", hook)
+  }),
+} satisfies PluginInternal.InternalPlugin)
+
+function supportsVerbosity(id: string) {
+  if (id.includes("gpt-6")) return true
+  if (id.includes("-chat") || id.includes("-image")) return false
+  // New GPT-5 minor versions remain unset until their support is known.
+  return /(?:^|[/.])gpt-5\.[1-6](?:[.:-]|$)/.test(id) || /(?:^|[/.])gpt-5(?:-(?:mini|nano)(?:[.:-]|$)|$)/.test(id)
+}
+
+function openAIModelID(model: Model.Info) {
+  const id = model.modelID.toLowerCase()
+  if (direct.has(model.package ?? "")) return id
+  if (model.package === "@opencode/ai/providers/amazon-bedrock/mantle/responses" && id.startsWith("openai."))
+    return id.slice("openai.".length)
+  if (gateways.has(model.package ?? "") && id.startsWith("openai/")) return id.slice("openai/".length)
+}

packages/core/src/session/compaction.ts

@@ -37,9 +37,8 @@ import { toLLMMessages } from "./runner/to-llm-message.js"
 import type { AgentNotFoundError } from "./error.js"
 import type { Instructions } from "../instructions/index.js"
 
-const DEFAULT_BUFFER = 20_000
+const AUTO_THRESHOLD = 0.85
 const DEFAULT_KEEP_TOKENS = 15_000
-const OUTPUT_TOKEN_MAX = 32_000
 const TOOL_OUTPUT_MAX_CHARS = 2_000
 const IMAGE_TOKEN_ESTIMATE = 1_500
 const PDF_TOKEN_ESTIMATE = 2_000
@@ -89,7 +88,7 @@ const LEGACY_HEADING = "## Additional Context"
 
 export type Settings = {
   auto: boolean
-  buffer: number
+  buffer?: number
   tokens: number
 }
 
@@ -401,7 +400,7 @@ export const layer = Layer.effect(
 
     const state = State.create<Settings & { readonly native: NativeStrategy[] }, Editor>({
       name: "session-compaction",
-      initial: () => ({ auto: true, buffer: DEFAULT_BUFFER, tokens: DEFAULT_KEEP_TOKENS, native: [] }),
+      initial: () => ({ auto: true, tokens: DEFAULT_KEEP_TOKENS, native: [] }),
       editor: (editor) => ({
         configure: (settings) => {
           if (settings.auto !== undefined) editor.auto = settings.auto
@@ -754,11 +753,9 @@ export const layer = Layer.effect(
       const limit = input.resolved.limit
       const context = limit.context
       if (context <= 0) return false
-      const output = Math.min(limit.output, OUTPUT_TOKEN_MAX)
-      const promptCeiling = Math.min(
-        limit.input === undefined ? Number.POSITIVE_INFINITY : limit.input - config.buffer,
-        context - Math.max(output, config.buffer),
-      )
+      const usable = Math.min(context, limit.input ?? context)
+      const promptCeiling =
+        config.buffer === undefined ? Math.floor(usable * AUTO_THRESHOLD) : usable - config.buffer
       return estimateTokens(input) >= promptCeiling
     }
     const compactManual = Effect.fn("SessionCompaction.compactManual")(function* (input: ManualInput) {

packages/core/test/aisdk.test.ts

@@ -375,6 +375,18 @@ it.effect("routes AI Gateway model options by upstream prefix", () =>
       bedrock: { reasoningConfig: { type: "enabled" } },
     })
 
+    const openai = yield* aisdk.model({
+      ...model("@ai-sdk/gateway", { gateway: { order: ["openai"] } }),
+      modelID: Model.ID.make("openai/gpt-5.5"),
+    })
+    const openaiPrepared = yield* compileRequest(
+      LLM.request({ model: openai, prompt: "Hello", providerOptions: { textVerbosity: "low" } }),
+    )
+    expect(openaiPrepared.body.providerOptions).toEqual({
+      gateway: { order: ["openai"] },
+      openai: { textVerbosity: "low" },
+    })
+
     const fallback = yield* aisdk.model({
       ...model("@ai-sdk/gateway", { reasoningEffort: "high" }),
       modelID: Model.ID.make("deepseek/deepseek-v4"),

packages/core/test/config/compaction.test.ts

@@ -172,5 +172,5 @@ const input = (tokens: number) => {
     },
   }
 }
-const bufferedInput = input(85_000)
+const bufferedInput = input(82_000)
 const nearInput = input(95_000)

packages/core/test/plugin/verbosity.test.ts

@@ -0,0 +1,163 @@
+import { expect } from "bun:test"
+import { Effect } from "effect"
+import { Agent } from "@opencode/core/agent"
+import { Model } from "@opencode/core/model"
+import { Plugin } from "@opencode/core/plugin"
+import { PluginHooks } from "@opencode/core/plugin/hooks"
+import { PluginHost } from "@opencode/core/plugin/host"
+import { VerbosityPlugin } from "@opencode/core/plugin/verbosity"
+import { Provider } from "@opencode/core/provider"
+import { Session } from "@opencode/core/session"
+import type { SessionHooks } from "@opencode/plugin/effect/session"
+import { testEffect } from "../lib/effect"
+import { PluginTestLayer } from "./fixture"
+
+const it = testEffect(PluginTestLayer)
+const ref = (providerID: string, id: string, variant?: string) =>
+  Model.Ref.make({
+    providerID: Provider.ID.make(providerID),
+    id: Model.ID.make(id),
+    ...(variant ? { variant: Model.VariantID.make(variant) } : {}),
+  })
+const request = (model: Model.Ref, options: SessionHooks["context"]["options"] = {}): SessionHooks["context"] => ({
+  sessionID: Session.ID.make("ses_verbosity"),
+  agent: Agent.ID.make("build"),
+  model,
+  system: [],
+  messages: [],
+  tools: {},
+  options,
+})
+
+it.effect("sets known OpenAI Responses defaults without overriding configured or unknown models", () =>
+  Effect.gen(function* () {
+    const providers = yield* Provider.Service
+    const plugins = yield* Plugin.Service
+    const hooks = yield* PluginHooks.Service
+    yield* providers.transform((editor) => {
+      editor.add({
+        info: { ...Provider.Info.empty(Provider.ID.make("openai")), package: "@opencode/ai/providers/openai" },
+        models: [
+          { ...Model.Info.default(Provider.ID.make("openai"), Model.ID.make("gpt-5.5")) },
+          { ...Model.Info.default(Provider.ID.make("openai"), Model.ID.make("gpt-5.6-luna-fast")) },
+          { ...Model.Info.default(Provider.ID.make("openai"), Model.ID.make("gpt-5.2-codex")) },
+          { ...Model.Info.default(Provider.ID.make("openai"), Model.ID.make("gpt-5-mini-fast")) },
+          { ...Model.Info.default(Provider.ID.make("openai"), Model.ID.make("gpt-6-astra-pro")) },
+          { ...Model.Info.default(Provider.ID.make("openai"), Model.ID.make("gpt-4o")) },
+          { ...Model.Info.default(Provider.ID.make("openai"), Model.ID.make("gpt-7")) },
+          { ...Model.Info.default(Provider.ID.make("openai"), Model.ID.make("gpt-5.7")) },
+          { ...Model.Info.default(Provider.ID.make("openai"), Model.ID.make("gpt-5.5-chat")) },
+          { ...Model.Info.default(Provider.ID.make("openai"), Model.ID.make("gpt-5.4-image-2")) },
+          {
+            ...Model.Info.default(Provider.ID.make("openai"), Model.ID.make("gpt-6-astra")),
+            variants: [{ id: Model.VariantID.make("quiet"), settings: { textVerbosity: null } }],
+          },
+          {
+            ...Model.Info.default(Provider.ID.make("openai"), Model.ID.make("chat")),
+            modelID: Model.ID.make("gpt-5.5"),
+            package: "@opencode/ai/providers/openai/chat",
+          },
+        ],
+      })
+      editor.add({
+        info: {
+          ...Provider.Info.empty(Provider.ID.make("opencode")),
+          package: "@opencode/ai/providers/openai-compatible",
+        },
+        models: [
+          {
+            ...Model.Info.default(Provider.ID.make("opencode"), Model.ID.make("astra-alias")),
+            modelID: Model.ID.make("gpt-6-astra"),
+            package: "@opencode/ai/providers/openai/responses",
+          },
+          {
+            ...Model.Info.default(Provider.ID.make("opencode"), Model.ID.make("no-default")),
+            modelID: Model.ID.make("gpt-6-astra"),
+            package: "@opencode/ai/providers/openai/responses",
+            settings: { textVerbosity: null },
+          },
+        ],
+      })
+      editor.add({
+        info: { ...Provider.Info.empty(Provider.ID.make("openrouter")), package: "@opencode/ai/providers/openrouter" },
+        models: [Model.Info.default(Provider.ID.make("openrouter"), Model.ID.make("gpt-5.5"))],
+      })
+      editor.add({
+        info: {
+          ...Provider.Info.empty(Provider.ID.make("configured")),
+          package: "@opencode/ai/providers/openai",
+          settings: { textVerbosity: "medium" },
+        },
+        models: [Model.Info.default(Provider.ID.make("configured"), Model.ID.make("gpt-5.5"))],
+      })
+      for (const [providerID, packageName, modelID] of [
+        ["azure", "@opencode/ai/providers/azure/responses", "gpt-5.5"],
+        ["bedrock-mantle", "@opencode/ai/providers/amazon-bedrock/mantle/responses", "openai.gpt-6-sol"],
+        ["cloudflare", "@opencode/ai/providers/cloudflare-ai-gateway", "openai/gpt-5.6-sol"],
+        ["vercel", Provider.aisdk("@ai-sdk/gateway"), "openai/gpt-6-astra-fast"],
+        ["azure-chat", "@opencode/ai/providers/azure/chat", "gpt-5.5"],
+        ["bedrock-converse", "@opencode/ai/providers/amazon-bedrock", "global.openai.gpt-6-sol"],
+        ["cloudflare-chat", "@opencode/ai/providers/cloudflare-ai-gateway", "workers-ai/gpt-5.5"],
+        ["vercel-other", Provider.aisdk("@ai-sdk/gateway"), "anthropic/gpt-5.5"],
+      ] as const) {
+        editor.add({
+          info: { ...Provider.Info.empty(Provider.ID.make(providerID)), package: packageName },
+          models: [
+            {
+              ...Model.Info.default(Provider.ID.make(providerID), Model.ID.make("selected")),
+              modelID: Model.ID.make(modelID),
+            },
+          ],
+        })
+      }
+    })
+    yield* VerbosityPlugin.Plugin.effect(yield* PluginHost.make(plugins))
+
+    for (const kind of ["context", "compaction", "generate", "title"] as const) {
+      const event = request(ref("openai", "gpt-5.5"))
+      yield* hooks.trigger("session", kind, event)
+      expect(event.options.textVerbosity).toBe("low")
+    }
+
+    for (const id of ["gpt-5.6-luna-fast", "gpt-5.2-codex", "gpt-5-mini-fast", "gpt-6-astra-pro"]) {
+      const event = request(ref("openai", id))
+      yield* hooks.trigger("session", "context", event)
+      expect(event.options.textVerbosity).toBe("low")
+    }
+
+    for (const model of [
+      ref("openai", "gpt-4o"),
+      ref("openai", "gpt-7"),
+      ref("openai", "gpt-5.7"),
+      ref("openai", "gpt-5.5-chat"),
+      ref("openai", "gpt-5.4-image-2"),
+      ref("openai", "chat"),
+      ref("openrouter", "gpt-5.5"),
+      ref("configured", "gpt-5.5"),
+      ref("azure-chat", "selected"),
+      ref("bedrock-converse", "selected"),
+      ref("cloudflare-chat", "selected"),
+      ref("vercel-other", "selected"),
+      ref("openai", "gpt-6-astra", "quiet"),
+      ref("opencode", "no-default"),
+    ]) {
+      const event = request(model)
+      yield* hooks.trigger("session", "context", event)
+      expect(event.options.textVerbosity).toBeUndefined()
+    }
+
+    const alias = request(ref("opencode", "astra-alias"))
+    yield* hooks.trigger("session", "context", alias)
+    expect(alias.options.textVerbosity).toBe("low")
+
+    for (const providerID of ["azure", "bedrock-mantle", "cloudflare", "vercel"]) {
+      const event = request(ref(providerID, "selected"))
+      yield* hooks.trigger("session", "context", event)
+      expect(event.options.textVerbosity).toBe("low")
+    }
+
+    const overridden = request(ref("openai", "gpt-5.5"), { textVerbosity: "high" })
+    yield* hooks.trigger("session", "context", overridden)
+    expect(overridden.options.textVerbosity).toBe("high")
+  }),
+)

packages/core/test/session-compaction.test.ts

@@ -153,7 +153,7 @@ test("compaction prompts prohibit task execution", () => {
     expect(SessionCompaction.buildPrompt(update)).toContain("Do not continue the task or call tools")
 })
 
-it.effect("auto compaction estimates current content against the buffered prompt ceiling", () =>
+it.effect("auto compaction uses 85% by default and a configured buffer instead", () =>
   Effect.gen(function* () {
     const compaction = yield* SessionCompaction.Service
     const session = Session.Info.make({
@@ -205,23 +205,27 @@ it.effect("auto compaction estimates current content against the buffered prompt
     }
 
     const inputLimited = { context: 400_000, input: 272_000, output: 128_000 }
-    expect(compaction.required(input(251_999, inputLimited))).toBe(false)
-    expect(compaction.required(input(252_000, inputLimited))).toBe(true)
+    expect(compaction.required(input(231_199, inputLimited))).toBe(false)
+    expect(compaction.required(input(231_200, inputLimited))).toBe(true)
     const native = (tokens: number, limit: { context: number; input?: number; output: number } = inputLimited) => {
       const selected = input(tokens, limit)
       return { ...selected, resolved: { ...selected.resolved, compaction: { type: "native" as const } } }
     }
-    expect(compaction.required(native(251_999))).toBe(false)
-    expect(compaction.required(native(252_000))).toBe(true)
+    expect(compaction.required(native(231_199))).toBe(false)
+    expect(compaction.required(native(231_200))).toBe(true)
     expect(compaction.required(native(1_000_000, { context: 0, input: undefined, output: 0 }))).toBe(false)
 
     const contextLimited = { context: 100_000, output: 10_000 }
-    expect(compaction.required(input(79_999, contextLimited))).toBe(false)
-    expect(compaction.required(input(80_000, contextLimited))).toBe(true)
+    expect(compaction.required(input(84_999, contextLimited))).toBe(false)
+    expect(compaction.required(input(85_000, contextLimited))).toBe(true)
 
     const outputLimited = { context: 100_000, output: 30_000 }
-    expect(compaction.required(input(69_999, outputLimited))).toBe(false)
-    expect(compaction.required(input(70_000, outputLimited))).toBe(true)
+    expect(compaction.required(input(84_999, outputLimited))).toBe(false)
+    expect(compaction.required(input(85_000, outputLimited))).toBe(true)
+
+    const smallWindow = { context: 32_000, output: 32_000 }
+    expect(compaction.required(input(27_199, smallWindow))).toBe(false)
+    expect(compaction.required(input(27_200, smallWindow))).toBe(true)
 
     const assistant = input(79_000, contextLimited).messages[0]
     const tool = SessionMessage.AssistantTool.make({
@@ -233,7 +237,9 @@ it.effect("auto compaction estimates current content against the buffered prompt
     })
     const grown = { ...input(79_000, contextLimited), messages: [{ ...assistant, content: [tool] }] }
     expect(SessionCompaction.estimateTokens(grown)).toBe(80_000)
-    expect(compaction.required(grown)).toBe(true)
+    expect(compaction.required(grown)).toBe(false)
+    const near = input(84_000, contextLimited)
+    expect(compaction.required({ ...near, messages: [{ ...near.messages[0], content: [tool] }] })).toBe(true)
 
     const interrupted = { ...assistant, id: SessionMessage.ID.create(), tokens: undefined }
     expect(SessionCompaction.estimateTokens({ ...grown, messages: [...grown.messages, interrupted] })).toBe(80_001)
@@ -285,6 +291,13 @@ it.effect("auto compaction estimates current content against the buffered prompt
       time: { created: 0, completed: 0 },
     })
     expect(compaction.required({ ...grown, messages: [checkpoint] })).toBe(false)
+
+    yield* compaction.transform((editor) => editor.configure({ buffer: 10_000 }))
+    expect(compaction.required(input(89_999, contextLimited))).toBe(false)
+    expect(compaction.required(input(90_000, contextLimited))).toBe(true)
+    yield* compaction.transform((editor) => editor.configure({ buffer: 0 }))
+    expect(compaction.required(input(99_999, contextLimited))).toBe(false)
+    expect(compaction.required(input(100_000, contextLimited))).toBe(true)
   }),
 )
 

packages/core/test/session-native-compaction.test.ts

@@ -362,6 +362,7 @@ it.live("manual and automatic endpoint compaction keep the provider replacement
     expect(JSON.stringify(replacement)).not.toContain("Original user")
     expect(fixture.state.calls).toBe(2)
     expect(fixture.headers[0]?.get("x-http-hook")).toBe("compaction")
+    expect(fixture.bodies[0]).toMatchObject({ tools: [expect.objectContaining({ name: "read" })] })
     expect(fixture.bodies[0]).not.toHaveProperty("context_management")
   }),
 )

packages/core/test/session-runner.test.ts

@@ -2883,7 +2883,7 @@ describe("SessionRunnerLLM", () => {
         agent.steps = 2
       }),
     )
-    yield* s.llm.push(TestLLM.textWithUsage("Earlier answer", "before-native", 10_000))
+    yield* s.llm.push(TestLLM.textWithUsage("Earlier answer", "before-native", 36_000))
     yield* s.runPrompt("First real request")
     const checkpoint = (encrypted: string) =>
       CompactionCheckpointResponse.make({
@@ -2903,7 +2903,7 @@ describe("SessionRunnerLLM", () => {
     const installed = (yield* s.messages).filter((message) => message.type === "compaction")
     expect(installed).toMatchObject([{ status: "completed", reason: "auto", providerContext: { version: 1 } }])
     // New input without a post-checkpoint usage anchor must not retrigger compaction.
-    yield* s.llm.push(TestLLM.textWithUsage("Measured", "measured", 10_000))
+    yield* s.llm.push(TestLLM.textWithUsage("Measured", "measured", 36_000))
     yield* s.runPrompt("Third real request")
     expect(s.requests).toHaveLength(5)
     yield* s.llm.push(checkpoint("second"), TestLLM.textWithUsage("Continued", "continued", 10_000))
@@ -2929,7 +2929,7 @@ describe("SessionRunnerLLM", () => {
     s.currentModel = LanguageModel.make({ id: "native", provider: "openai", route: OpenAIResponses.route })
     modelLimits.set("native", { context: 42_000, output: 32_000 })
     s.compaction = { type: "native" }
-    yield* s.llm.push(TestLLM.textWithUsage("Earlier answer", "before-native", 10_000))
+    yield* s.llm.push(TestLLM.textWithUsage("Earlier answer", "before-native", 36_000))
     yield* s.runPrompt("Original durable request")
     yield* s.llm.push(
       CompactionCheckpointResponse.make({

services/www/src/docs/content/compaction.mdx

@@ -78,17 +78,20 @@ Reusing that ID for another record returns a conflict.
 ## Automatic
 
 Before each model call, OpenCode estimates the final size of the system prompt,
-messages, and advertised tools. It starts compaction at this ceiling:
+messages, and advertised tools. By default it starts compaction at 85% of the
+smaller context or input limit. A configured `buffer` replaces that percentage
+with an absolute number of tokens to reserve:
 
 ```text
-estimated tokens >= min(input limit - buffer, context limit - max(output reserve, buffer))
+usable window = min(context limit, input limit if present)
+compact when estimated tokens >= (buffer configured ? usable window - buffer : floor(85% of usable window))
 ```
 
-For example, with a 128,000-token input limit and the default 20,000-token
-buffer, the input-limit side of the ceiling is 108,000 tokens.
+For example, with a 200,000-token context and a 128,000-token input limit,
+the default threshold is 108,800 tokens:
 
 ```text
-128,000 - 20,000 = 108,000
+128,000 × 85% = 108,800
 ```
 
 The estimate follows these rules:
@@ -97,8 +100,7 @@ The estimate follows these rules:
   newer content are then added.
 - Without provider usage, OpenCode estimates text, media, instructions, and
   tools locally.
-- The reserved model output is capped at 32,000 tokens.
-- A model without an input limit is constrained by its context limit instead.
+- Without a separate input limit, the context limit sets the threshold.
 - A successful checkpoint rebuilds the same pending model step. It does not
   promote the input again or spend another agent step.
 
@@ -132,11 +134,11 @@ Add `compaction` to any [OpenCode configuration file](/config):
 | --- | ---: | --- |
 | `auto` | `true` | Enables preflight checks and one provider-overflow recovery attempt. It does not control manual compaction. |
 | `keep.tokens` | `15000` | Approximate recent serialized context retained beside a local summary, or real user input retained for a provider checkpoint. |
-| `buffer` | `20000` | Safety margin below an explicit input limit. Without one, it is the minimum context reserve; a larger model output allowance wins. |
+| `buffer` | unset | Replaces the 85% threshold with `usable window - buffer`. Set it to reserve a fixed number of tokens. |
 
 `keep.tokens` and `buffer` accept non-negative integers. Larger
 `keep.tokens` preserves more recent detail but leaves less room for new work;
-larger `buffer` starts automatic compaction earlier.
+larger `buffer` starts automatic compaction earlier when configured.
 
 ## Providers
 
@@ -161,7 +163,7 @@ A model setting overrides the provider setting.
 
 | Topic | Provider behavior |
 | --- | --- |
-| Threshold | OpenCode uses the model's usable input ceiling configured by its context, input, and output limits plus the global compaction buffer. |
+| Threshold | OpenCode starts at 85% of the smaller context or input limit, unless `buffer` replaces it with a fixed reserve. |
 | Scheduling | Checkpoints run at normal safe step boundaries. `compaction.auto: false` disables all new automatic work. |
 | Usage | After a native checkpoint is installed, automatic checks wait for fresh model usage because encrypted checkpoint bytes cannot provide a meaningful token count. |
 | OpenAI | Responses routes use a streamed compaction trigger when supported; endpoint-only routes use the standalone compaction endpoint. Deployment and model support vary. |