@bastani/pi-ai 0.9.18 → 0.9.19-alpha.10

package/CHANGELOG.md

@@ -4,6 +4,38 @@ This package is a Bastani fork of `@earendil-works/pi-ai`. Upstream history at t
 
 ## [Unreleased]
 
+## [0.9.19-alpha.4] - 2026-09-10
+
+### Added
+
+- Enabled native deferred tool loading for Fireworks Messages models. Use `ToolSearch` or `tool_search` as the loader name for prompt-prefix deferral ([#9323](https://github.com/earendil-works/pi/issues/9323)).
+
+### Fixed
+
+- Fixed OpenAI Codex requests to send the model's Off reasoning effort instead of omitting it, while respecting unsupported Off mappings ([#9191](https://github.com/earendil-works/pi/issues/9191)).
+- Fixed Fireworks unsigned thinking replay and reasoning effort selection using catalog metadata, with verified DeepSeek V4 and Qwen3.8 fallbacks and removal of redundant GLM 5.2 and Kimi K3 effort aliases ([#9323](https://github.com/earendil-works/pi/issues/9323)).
+- Fixed OpenRouter requests to send `x-session-id` from `sessionId` for Chat Completions and Anthropic Messages models when prompt caching is enabled ([#9102](https://github.com/earendil-works/pi/issues/9102)).
+- Fixed the DeepSeek catalog to advertise `deepseek-flash` for DeepSeek V4.1 Flash instead of retired Flash aliases, and refreshed DeepSeek pricing metadata ([#9423](https://github.com/earendil-works/pi/issues/9423)).
+- Removed GPT-5.4 and GPT-5.4 mini from the OpenAI Codex catalog after they became unavailable to ChatGPT accounts ([#9394](https://github.com/earendil-works/pi/issues/9394)).
+- Fixed Mistral-hosted GLM-5.2 reasoning requests to use `reasoning_effort` instead of the ignored `prompt_mode` ([#9375](https://github.com/earendil-works/pi/issues/9375)).
+
+## [0.9.19-alpha.3] - 2026-09-09
+
+### Fixed
+
+- Capped shared assistant retry backoff with `RetryPolicy.maxAgentDelayMs` (60 seconds by default), including summary calls ([#8826](https://github.com/earendil-works/pi/issues/8826)).
+
+## [0.9.19-alpha.2] - 2026-09-08
+
+### Fixed
+
+- Fixed Anthropic Messages tool requests advertising root object-union parameter schemas without fields by projecting their branch fields into Anthropic-compatible object schemas ([#2190](https://github.com/bastani-inc/atomic/pull/2190) by [@elefthei](https://github.com/elefthei)).
+- Fixed Anthropic object-union tool projection dropping own property names such as `__proto__` and incorrectly projecting explicitly nonobject branches with `properties` metadata; authored schemas and runtime validation remain unchanged ([#2190](https://github.com/bastani-inc/atomic/pull/2190), [#2189](https://github.com/bastani-inc/atomic/issues/2189)).
+- Updated Anthropic, Bedrock, Google and HTTP proxy SDKs to upstream Pi's current versions, and mapped Google's consecutive tool-call limit to an error stop reason.
+- Routed all GitHub Copilot GPT model IDs through the Responses API ([#9209](https://github.com/earendil-works/pi/issues/9209)).
+- Fixed quadratic CPU usage when draining buffered `EventStream` events ([#9055](https://github.com/earendil-works/pi/issues/9055)).
+- Fixed reasoning-capable Mistral Medium aliases using unsupported `prompt_mode` instead of `reasoning_effort` ([#8700](https://github.com/earendil-works/pi/issues/8700)).
+
 ## [0.9.18] - 2026-09-05
 
 Cumulative release of the `0.9.18-alpha.5` through `0.9.18-alpha.7` prereleases. Per-change details remain in the unchanged prerelease sections below.

package/README.md

@@ -575,6 +575,12 @@ context.messages.push({
 });
 ```
 
+### Fireworks deferred tools
+
+Fireworks models using the Anthropic Messages API support native deferred tool loading. Keep your discovery tool in `context.tools`, add discovered tool definitions there, and set `addedToolNames` on the discovery tool's result to the names you added. The adapter sends deferred schemas and `tool_reference` blocks while preserving ordinary result text and unsigned thinking during replay.
+
+Name the discovery tool `ToolSearch` or `tool_search` for Fireworks prompt-prefix deferral. Other discovery names serialize correctly but do not get that prefix optimization. Set `model.compat.supportsToolReferences` to `false` to send ordinary tool schemas instead. This feature does not apply to Fireworks Chat Completions models.
+
 ### Streaming Tool Calls with Partial JSON
 
 During streaming, tool call arguments are progressively parsed as they arrive. This enables real-time UI updates before the complete arguments are available:
@@ -1183,7 +1189,7 @@ interface OpenAICompletionsCompat {
   supportsUsageInStreaming?: boolean; // Whether provider supports `stream_options: { include_usage: true }` (default: true)
   supportsStrictMode?: boolean;      // Whether provider supports `strict` in tool definitions (default: true)
   supportsOpenAIGrammarTools?: boolean; // Whether to emit OpenAI custom Lark/regex grammar tools; false falls back to normal function tools (default: false; the generated catalog enables it for capable models)
-  sendSessionAffinityHeaders?: boolean; // Send session-affinity data from `sessionId` (default: false)
+  sendSessionAffinityHeaders?: boolean; // Send session-affinity data from `sessionId` (default: true for OpenRouter, false otherwise)
   sessionAffinityFormat?: 'openai' | 'openai-nosession' | 'openrouter'; // Format for session affinity: 'openai' uses `prompt_cache_key`, `session_id`, `x-client-request-id`, and `x-session-affinity`; 'openai-nosession' uses `prompt_cache_key`, `x-client-request-id`, and `x-session-affinity`; 'openrouter' uses `x-session-id` (default: auto-detected)
   maxTokensField?: 'max_completion_tokens' | 'max_tokens';  // Which field name to use (default: max_completion_tokens)
   requiresToolResultName?: boolean;  // Whether tool results require the `name` field (default: false)
@@ -1211,6 +1217,8 @@ interface OpenAIResponsesCompat {
 }
 ```
 
+OpenRouter requests send `x-session-id` from `sessionId` when prompt caching is enabled. Chat Completions and Anthropic Messages both auto-detect OpenRouter endpoints unless `sendSessionAffinityHeaders` is explicitly false. On Anthropic-compatible models, `sessionAffinityFormat: "openrouter"` selects `x-session-id`; when unset on non-OpenRouter endpoints, the existing `x-session-affinity` format is used. Explicit request headers take precedence over generated headers.
+
 If `compat` is not set, the library falls back to URL-based detection. If `compat` is partially set, unspecified fields use the detected defaults. This is useful for:
 
 - **LiteLLM proxies**: May not support `store` field

package/dist/api/anthropic-messages.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"anthropic-messages.d.ts","sourceRoot":"","sources":["../../src/api/anthropic-messages.ts"],"names":[],"mappings":"AAAA,OAAO,SAAS,MAAM,mBAAmB,CAAC;AAU1C,OAAO,KAAK,EAWX,mBAAmB,EAEnB,cAAc,EACd,aAAa,EAOb,MAAM,aAAa,CAAC;AA0JrB,MAAM,MAAM,eAAe,GAAG,KAAK,GAAG,QAAQ,GAAG,MAAM,GAAG,OAAO,GAAG,KAAK,CAAC;AAE1E,MAAM,MAAM,wBAAwB,GAAG,YAAY,GAAG,SAAS,CAAC;AA8LhE,MAAM,WAAW,gBAAiB,SAAQ,aAAa;IACtD;;;;;;OAMG;IACH,eAAe,CAAC,EAAE,OAAO,CAAC;IAC1B;;;;OAIG;IACH,oBAAoB,CAAC,EAAE,MAAM,CAAC;IAC9B;;;;;;;;;;;OAWG;IACH,MAAM,CAAC,EAAE,eAAe,CAAC;IACzB;;;;;;;;;;;OAWG;IACH,eAAe,CAAC,EAAE,wBAAwB,CAAC;IAC3C;;;;;OAKG;IACH,mBAAmB,CAAC,EAAE,OAAO,CAAC;IAC9B;;;;OAIG;IACH,UAAU,CAAC,EAAE,MAAM,GAAG,KAAK,GAAG,MAAM,GAAG;QAAE,IAAI,EAAE,MAAM,CAAC;QAAC,IAAI,EAAE,MAAM,CAAA;KAAE,CAAC;IACtE;;;;OAIG;IACH,MAAM,CAAC,EAAE,SAAS,CAAC;CACnB;AA6OD,eAAO,MAAM,MAAM,EAAE,cAAc,CAAC,oBAAoB,EAAE,gBAAgB,CAsWzE,CAAC;AA4BF,eAAO,MAAM,YAAY,EAAE,cAAc,CAAC,oBAAoB,EAAE,mBAAmB,CA8ClF,CAAC","sourcesContent":["import Anthropic from \"@anthropic-ai/sdk\";\nimport type {\n\tCacheControlEphemeral,\n\tContentBlockParam,\n\tMessageCreateParamsStreaming,\n\tMessageParam,\n\tRawMessageStreamEvent,\n\tRefusalStopDetails,\n} from \"@anthropic-ai/sdk/resources/messages.js\";\nimport { calculateCost } from \"../models.ts\";\nimport type {\n\tAnthropicMessagesCompat,\n\tApi,\n\tAssistantMessage,\n\tCacheRetention,\n\tContext,\n\tImageContent,\n\tMessage,\n\tModel,\n\tProviderEnv,\n\tProviderHeaders,\n\tSimpleStreamOptions,\n\tStopReason,\n\tStreamFunction,\n\tStreamOptions,\n\tTextContent,\n\tThinkingContent,\n\tTool,\n\tToolCall,\n\tToolResultMessage,\n\tUsage,\n} from \"../types.ts\";\nimport { splitDeferredTools } from \"../utils/deferred-tools.ts\";\nimport { appendAssistantMessageDiagnostic } from \"../utils/diagnostics.ts\";\nimport { assertSupportedDocumentMimeType } from \"../utils/document-input.ts\";\nimport { AssistantMessageEventStream } from \"../utils/event-stream.ts\";\nimport { headersToRecord } from \"../utils/headers.ts\";\nimport { parseJsonWithRepair, parseStreamingJson } from \"../utils/json-parse.ts\";\nimport { getPiUserAgent } from \"../utils/pi-user-agent.ts\";\nimport { getProviderEnvValue } from \"../utils/provider-env.ts\";\nimport { retryProviderRequest } from \"../utils/provider-retry.ts\";\nimport { sanitizeSurrogates } from \"../utils/sanitize-unicode.ts\";\nimport { createStreamDeadline, withStreamDeadline } from \"../utils/stream-deadline.ts\";\n\nimport { getJsonSchemaToolParameters, resolveJsonSchemaStrictSampling } from \"./constrained-sampling.ts\";\nimport {\n\tbuildCopilotDynamicHeaders,\n\thasCopilotVisionInput,\n\tpreserveCopilotIntegrationHeader,\n} from \"./github-copilot-headers.ts\";\nimport { adjustMaxTokensForThinking, buildBaseOptions, clampMaxTokensToContext } from \"./simple-options.ts\";\nimport { transformMessages } from \"./transform-messages.ts\";\n\n/**\n * Resolve cache retention preference.\n * Defaults to \"short\" and uses PI_CACHE_RETENTION for backward compatibility.\n */\nfunction resolveCacheRetention(cacheRetention?: CacheRetention, env?: ProviderEnv): CacheRetention {\n\tif (cacheRetention) {\n\t\treturn cacheRetention;\n\t}\n\tif (getProviderEnvValue(\"PI_CACHE_RETENTION\", env) === \"long\") {\n\t\treturn \"long\";\n\t}\n\treturn \"short\";\n}\n\nfunction getCacheControl(\n\tmodel: Model<\"anthropic-messages\">,\n\tcacheRetention?: CacheRetention,\n\tenv?: ProviderEnv,\n): { retention: CacheRetention; cacheControl?: CacheControlEphemeral } {\n\tconst retention = resolveCacheRetention(cacheRetention, env);\n\tif (retention === \"none\") {\n\t\treturn { retention };\n\t}\n\tconst ttl = retention === \"long\" && getAnthropicCompat(model).supportsLongCacheRetention ? \"1h\" : undefined;\n\treturn {\n\t\tretention,\n\t\tcacheControl: { type: \"ephemeral\", ...(ttl && { ttl }) },\n\t};\n}\n\n// Stealth mode: Mimic Claude Code's tool naming exactly\n//\n// Anthropic gates newer models on the `claude-cli/<version>` user agent alone and rejects an\n// older one with `claude_code_version_too_old`. `claude-fable-5-1` requires >= 2.1.251, bisected\n// against the live API: 2.1.250 -> 400, 2.1.251 -> 200. This is pinned to that exact published\n// minimum rather than the newest release, and it is a strict superset of the previous 2.1.75 --\n// every model this provider ships answers 200 at 2.1.251. Raise it only when a model rejects\n// this value; a caller can override it for one client through the `headers` option, whose\n// lowercase `user-agent` key is merged last.\nconst claudeCodeVersion = \"2.1.251\";\n\n// Claude Code 2.x tool names (canonical casing)\n// Source: https://cchistory.mariozechner.at/data/prompts-2.1.11.md\n// To update: https://github.com/badlogic/cchistory\nconst claudeCodeTools = [\n\t\"Read\",\n\t\"Write\",\n\t\"Edit\",\n\t\"Bash\",\n\t\"Grep\",\n\t\"Glob\",\n\t\"AskUserQuestion\",\n\t\"EnterPlanMode\",\n\t\"ExitPlanMode\",\n\t\"KillShell\",\n\t\"NotebookEdit\",\n\t\"Skill\",\n\t\"Task\",\n\t\"TaskOutput\",\n\t\"TodoWrite\",\n\t\"WebFetch\",\n\t\"WebSearch\",\n];\n\nconst ccToolLookup = new Map(claudeCodeTools.map((t) => [t.toLowerCase(), t]));\n\n// Convert tool name to CC canonical casing if it matches (case-insensitive)\nconst toClaudeCodeName = (name: string) => ccToolLookup.get(name.toLowerCase()) ?? name;\nconst fromClaudeCodeName = (name: string, tools?: Tool[]) => {\n\tif (tools && tools.length > 0) {\n\t\tconst lowerName = name.toLowerCase();\n\t\tconst matchedTool = tools.find((tool) => tool.name.toLowerCase() === lowerName);\n\t\tif (matchedTool) return matchedTool.name;\n\t}\n\treturn name;\n};\n\n/**\n * Convert tool-result content blocks to Anthropic API format.\n *\n * Tool results carry text and images only. `DocumentContent` appears exclusively in user\n * messages, which `convertMessages` serializes on its own path.\n */\nfunction convertContentBlocks(content: (TextContent | ImageContent)[]):\n\t| string\n\t| Array<\n\t\t\t| { type: \"text\"; text: string }\n\t\t\t| {\n\t\t\t\t\ttype: \"image\";\n\t\t\t\t\tsource: {\n\t\t\t\t\t\ttype: \"base64\";\n\t\t\t\t\t\tmedia_type: \"image/jpeg\" | \"image/png\" | \"image/gif\" | \"image/webp\";\n\t\t\t\t\t\tdata: string;\n\t\t\t\t\t};\n\t\t\t  }\n\t  > {\n\t// If only text blocks, return as concatenated string for simplicity\n\tconst hasImages = content.some((c) => c.type === \"image\");\n\tif (!hasImages) {\n\t\treturn sanitizeSurrogates(content.map((c) => (c as TextContent).text).join(\"\\n\"));\n\t}\n\n\t// If we have images, convert to content block array\n\tconst blocks = content.map((block) => {\n\t\tif (block.type === \"text\") {\n\t\t\treturn {\n\t\t\t\ttype: \"text\" as const,\n\t\t\t\ttext: sanitizeSurrogates(block.text),\n\t\t\t};\n\t\t}\n\t\treturn {\n\t\t\ttype: \"image\" as const,\n\t\t\tsource: {\n\t\t\t\ttype: \"base64\" as const,\n\t\t\t\tmedia_type: block.mimeType as \"image/jpeg\" | \"image/png\" | \"image/gif\" | \"image/webp\",\n\t\t\t\tdata: block.data,\n\t\t\t},\n\t\t};\n\t});\n\n\t// If only images (no text), add placeholder text block\n\tconst hasText = blocks.some((b) => b.type === \"text\");\n\tif (!hasText) {\n\t\tblocks.unshift({\n\t\t\ttype: \"text\" as const,\n\t\t\ttext: \"(see attached image)\",\n\t\t});\n\t}\n\n\treturn blocks;\n}\n\nexport type AnthropicEffort = \"low\" | \"medium\" | \"high\" | \"xhigh\" | \"max\";\n\nexport type AnthropicThinkingDisplay = \"summarized\" | \"omitted\";\n\ntype MessageCreateParamsStreamingWithFallbacks = MessageCreateParamsStreaming & {\n\tfallbacks?: readonly { model: string }[];\n};\n\nconst FINE_GRAINED_TOOL_STREAMING_BETA = \"fine-grained-tool-streaming-2025-05-14\";\nconst INTERLEAVED_THINKING_BETA = \"interleaved-thinking-2025-05-14\";\nconst SERVER_SIDE_FALLBACK_BETA = \"server-side-fallback-2026-07-01\";\nconst THINKING_BINDING_CONTROLS_BETA = \"thinking-binding-controls-2026-08-01\";\nconst MID_CONVERSATION_OUTPUT_CONFIG_BETA = \"mid-conversation-output-config-2026-07-01\";\n\nfunction shouldUseServerSideFallbackBeta(model: Model<\"anthropic-messages\">): boolean {\n\treturn (model.compat?.allowedFallbackModels?.length ?? 0) > 0;\n}\n\n/**\n * Claude Fable 5.1 binds each thinking block to the conversation prefix that produced it and,\n * for Anthropic accounts created on or after 2026-08-31, rejects a replay behind a changed\n * `system` prompt, `tools` array, or earlier message with a 400 `invalid_request_error`.\n * Atomic rebuilds those inputs between turns (dynamic system prompt, tool availability changes,\n * model switches), so it opts into `prefix_mismatch_behavior: \"drop_block\"`: the API discards\n * the affected thinking blocks and answers the turn instead of failing the session.\n *\n * This covers live prefix mismatches *between* compaction boundaries. It is not what handles\n * compaction itself: Atomic's client-side `preserve_recent` tail compaction serializes the\n * protected tail into a single boundary message, so no signed thinking block survives a\n * boundary to be replayed behind it. That is Anthropic's documented keep-tail remedy applied\n * structurally rather than a case `drop_block` has to catch.\n * https://platform.claude.com/docs/en/build-with-claude/preserved-thinking\n */\nfunction shouldUseThinkingBindingControlsBeta(model: Model<\"anthropic-messages\">): boolean {\n\treturn model.compat?.enforcesPreservedThinkingBinding === true;\n}\n\n/**\n * Per-turn effort and preserved-thinking binding are independent capabilities.\n * `supportsMidConvoEffort` adds provider effort markers and requires binding controls;\n * Atomic's older `enforcesPreservedThinkingBinding` remains the broader opt-in for\n * transports that need drop recovery without supporting effort-only system messages.\n */\nfunction supportsMidConvoEffort(model: Model<\"anthropic-messages\">): boolean {\n\treturn model.compat?.supportsMidConvoEffort === true;\n}\n\n/** One entry of the `input_transformations` array the block-binding controls beta adds. */\ninterface AnthropicInputTransformation {\n\ttype?: string;\n\tpath?: string;\n\treason?: string;\n}\n\nfunction getInputTransformations(message: unknown): AnthropicInputTransformation[] | undefined {\n\tconst transformations = (message as { input_transformations?: AnthropicInputTransformation[] })\n\t\t?.input_transformations;\n\treturn Array.isArray(transformations) ? transformations : undefined;\n}\n\n/** Record thinking blocks the API dropped from this request without exposing block contents. */\nfunction recordInputTransformations(output: AssistantMessage, transformations: AnthropicInputTransformation[]): void {\n\tif (transformations.length === 0) return;\n\tappendAssistantMessageDiagnostic(output, {\n\t\ttype: \"anthropic_input_transformations\",\n\t\ttimestamp: Date.now(),\n\t\tdetails: {\n\t\t\tdroppedBlockCount: transformations.length,\n\t\t\treasons: [...new Set(transformations.map((entry) => entry.reason).filter((r): r is string => !!r))],\n\t\t\tpaths: transformations.map((entry) => entry.path).filter((p): p is string => !!p),\n\t\t},\n\t});\n}\n\n/** One entry of the per-attempt `usage.iterations` array a server-side fallback response carries. */\ninterface AnthropicUsageIteration {\n\ttype?: string;\n\tmodel?: string;\n\tinput_tokens?: number;\n\toutput_tokens?: number;\n\tcache_read_input_tokens?: number;\n\t/** The aggregate cache-write count. `cache_creation` splits the same tokens by TTL. */\n\tcache_creation_input_tokens?: number;\n\t/** \"Breakdown of cached tokens by TTL\", per `BetaMessageIterationUsage`. */\n\tcache_creation?: { ephemeral_1h_input_tokens?: number; ephemeral_5m_input_tokens?: number } | null;\n}\n\n/**\n * Bill the attempts that ran *before* the one which produced the returned message.\n *\n * \"Every attempt that produced output, including one that declined partway through its response,\n * is billed separately at the rates of the model that ran it. The `usage.iterations` array is the\n * per-attempt record of what you're billed. The top-level `usage` counts describe only the attempt\n * that produced the returned message. Tokens from different models are never summed into one\n * field.\" https://platform.claude.com/docs/en/build-with-claude/refusals-and-fallback\n *\n * Two rules follow, and both matter for not double-counting:\n *\n * - Only `type: \"message\"` entries are added. The `fallback_message` entry *is* the serving\n *   attempt, and its tokens are already in the top-level usage that `calculateCost` just priced.\n * - Only entries with output are added: \"An attempt that declined before producing any output is\n *   not billed: its tokens are reported on its `usage.iterations` entry but not charged.\"\n *\n * Token counts are deliberately left alone — the docs forbid summing across models and `Usage` has\n * no per-attempt shape — so this contributes to `usage.cost` only.\n */\nfunction addEarlierAttemptCosts(\n\toutput: AssistantMessage,\n\tmodel: Model<\"anthropic-messages\">,\n\tservingModel: Model<\"anthropic-messages\">,\n\tevent: unknown,\n): void {\n\tconst iterations = (event as { usage?: { iterations?: AnthropicUsageIteration[] } })?.usage?.iterations;\n\tif (!Array.isArray(iterations) || iterations.length === 0) return;\n\n\tfor (const iteration of iterations) {\n\t\tif (iteration.type !== \"message\") continue;\n\t\tif (!iteration.output_tokens) continue;\n\n\t\t// Price at the rates of the model that ran the attempt, which is not the serving model.\n\t\tconst attemptCost =\n\t\t\titeration.model === model.id\n\t\t\t\t? model.cost\n\t\t\t\t: iteration.model === servingModel.id\n\t\t\t\t\t? servingModel.cost\n\t\t\t\t\t: model.compat?.allowedFallbackModels?.find(\n\t\t\t\t\t\t\t(fallback) => fallback.provider === model.provider && fallback.model === iteration.model,\n\t\t\t\t\t\t)?.cost;\n\t\tif (!attemptCost) continue;\n\n\t\tconst attemptUsage: Usage = {\n\t\t\tinput: iteration.input_tokens ?? 0,\n\t\t\toutput: iteration.output_tokens,\n\t\t\tcacheRead: iteration.cache_read_input_tokens ?? 0,\n\t\t\t// `cacheWrite` stays the aggregate: `calculateCost` derives the 5-minute share by\n\t\t\t// subtracting `cacheWrite1h` from it, so setting this to the 5-minute count instead\n\t\t\t// would under-charge and adding the two together would double-charge. Without the 1h\n\t\t\t// split, an hour-long write bills at the 5-minute rate rather than 2x base input.\n\t\t\tcacheWrite: iteration.cache_creation_input_tokens ?? 0,\n\t\t\tcacheWrite1h: iteration.cache_creation?.ephemeral_1h_input_tokens ?? 0,\n\t\t\ttotalTokens: 0,\n\t\t\tcost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0, total: 0 },\n\t\t};\n\t\tcalculateCost({ ...model, id: iteration.model ?? model.id, cost: attemptCost }, attemptUsage);\n\n\t\toutput.usage.cost.input += attemptUsage.cost.input;\n\t\toutput.usage.cost.output += attemptUsage.cost.output;\n\t\toutput.usage.cost.cacheRead += attemptUsage.cost.cacheRead;\n\t\toutput.usage.cost.cacheWrite += attemptUsage.cost.cacheWrite;\n\t\toutput.usage.cost.total += attemptUsage.cost.total;\n\t}\n}\n\nfunction getAnthropicCompat(\n\tmodel: Model<\"anthropic-messages\">,\n): Required<\n\tOmit<\n\t\tAnthropicMessagesCompat,\n\t\t| \"forceAdaptiveThinking\"\n\t\t| \"allowedFallbackModels\"\n\t\t| \"enforcesPreservedThinkingBinding\"\n\t\t| \"delegatesThinkingModelBinding\"\n\t\t| \"supportsForcedToolChoice\"\n\t\t| \"supportsMidConvoEffort\"\n\t>\n> {\n\treturn {\n\t\tsupportsEagerToolInputStreaming: model.compat?.supportsEagerToolInputStreaming ?? true,\n\t\tsupportsLongCacheRetention: model.compat?.supportsLongCacheRetention ?? true,\n\t\tsendSessionAffinityHeaders: model.compat?.sendSessionAffinityHeaders ?? false,\n\t\tsupportsCacheControlOnTools: model.compat?.supportsCacheControlOnTools ?? true,\n\t\tsupportsTemperature: model.compat?.supportsTemperature ?? true,\n\t\tallowEmptySignature: model.compat?.allowEmptySignature ?? false,\n\t\tsupportsStrictTools: model.compat?.supportsStrictTools ?? false,\n\t\tsupportsToolReferences: model.compat?.supportsToolReferences ?? defaultSupportsToolReferences(model),\n\t};\n}\n\n/**\n * Default for `supportsToolReferences`: first-party Anthropic models except\n * Haiku (rejects client-side tool_reference blocks) and models that predate\n * tool search (Claude 3.x, Opus/Sonnet 4.0, Opus 4.1).\n */\nfunction defaultSupportsToolReferences(model: Model<\"anthropic-messages\">): boolean {\n\tif (model.provider !== \"anthropic\" || model.id.includes(\"haiku\")) return false;\n\tconst version = model.id.match(/^claude-(?:opus|sonnet|fable)-(\\d+)(?:-(\\d+))?(?:-|$)/);\n\tif (!version) return false;\n\tconst major = Number(version[1]);\n\tconst minor = version[2] && version[2].length < 8 ? Number(version[2]) : 0;\n\treturn major > 4 || (major === 4 && minor >= 5);\n}\n\nexport interface AnthropicOptions extends StreamOptions {\n\t/**\n\t * Enable extended thinking.\n\t * For adaptive thinking models: the model decides when/how much to think.\n\t * For older models: uses budget-based thinking with thinkingBudgetTokens.\n\t * Default: undefined (thinking is omitted unless `streamSimple()` maps\n\t * a simple reasoning level to this option, or callers set it explicitly).\n\t */\n\tthinkingEnabled?: boolean;\n\t/**\n\t * Token budget for extended thinking (older models only).\n\t * Ignored for adaptive thinking models.\n\t * Default: 1024 when `thinkingEnabled` is true and no budget is provided.\n\t */\n\tthinkingBudgetTokens?: number;\n\t/**\n\t * Effort level for adaptive thinking models.\n\t * Controls how much thinking Claude allocates:\n\t * - \"max\": Always thinks with no constraints (Opus 4.6 only)\n\t * - \"xhigh\": Highest reasoning level (Opus 4.7+, Fable 5)\n\t * - \"high\": Always thinks, deep reasoning\n\t * - \"medium\": Moderate thinking, may skip for simple queries\n\t * - \"low\": Minimal thinking, skips for simple tasks\n\t * Ignored for older models.\n\t * Default: omitted unless `streamSimple()` maps a simple reasoning\n\t * level to this option.\n\t */\n\teffort?: AnthropicEffort;\n\t/**\n\t * Controls how thinking content is returned in API responses.\n\t * - \"summarized\": Thinking blocks contain summarized thinking text.\n\t * - \"omitted\": Thinking blocks return an empty thinking field; the encrypted\n\t *   signature still travels back for multi-turn continuity. Use for faster\n\t *   time-to-first-text-token when your UI does not surface thinking.\n\t *\n\t * Note: Anthropic's API default for Claude Opus 4.7 and Claude Mythos Preview\n\t * is \"omitted\". We default to \"summarized\" here to keep behavior consistent\n\t * with older Claude 4 models. Set this explicitly to \"omitted\" to opt in.\n\t * Default: \"summarized\" when thinking is enabled.\n\t */\n\tthinkingDisplay?: AnthropicThinkingDisplay;\n\t/**\n\t * Whether to request the interleaved thinking beta header for non-adaptive\n\t * thinking models. Adaptive thinking models have interleaved thinking built in,\n\t * so the header is skipped for them regardless of this setting.\n\t * Default: true.\n\t */\n\tinterleavedThinking?: boolean;\n\t/**\n\t * Anthropic tool choice behavior. String values map to Anthropic's built-in\n\t * choices; `{ type: \"tool\", name }` forces a specific tool.\n\t * Default: omitted (Anthropic default behavior, currently equivalent to auto).\n\t */\n\ttoolChoice?: \"auto\" | \"any\" | \"none\" | { type: \"tool\"; name: string };\n\t/**\n\t * Pre-built Anthropic client instance. When provided, skips internal client\n\t * construction entirely. Use this to inject alternative SDK clients such as\n\t * `AnthropicVertex` that shares the same messaging API.\n\t */\n\tclient?: Anthropic;\n}\n\nfunction mergeHeaders(...headerSources: (ProviderHeaders | undefined)[]): ProviderHeaders {\n\tconst merged: ProviderHeaders = {};\n\tfor (const headers of headerSources) {\n\t\tif (headers) {\n\t\t\tObject.assign(merged, headers);\n\t\t}\n\t}\n\treturn merged;\n}\n\nfunction mergeClientHeaders(...headerSources: (ProviderHeaders | undefined)[]): ProviderHeaders {\n\treturn mergeHeaders({ \"User-Agent\": getPiUserAgent() }, ...headerSources);\n}\n\nfunction hasHeader(headers: ProviderHeaders | undefined, name: string): boolean {\n\tif (!headers) return false;\n\tconst expected = name.toLowerCase();\n\tfor (const [key, value] of Object.entries(headers)) {\n\t\tif (key.toLowerCase() === expected && value !== null && value.trim().length > 0) return true;\n\t}\n\treturn false;\n}\n\nfunction assertRequestAuth(provider: string, apiKey: string | undefined, headers: ProviderHeaders | undefined): void {\n\tif (apiKey) return;\n\tif (\n\t\thasHeader(headers, \"authorization\") ||\n\t\thasHeader(headers, \"x-api-key\") ||\n\t\thasHeader(headers, \"cf-aig-authorization\")\n\t) {\n\t\treturn;\n\t}\n\tthrow new Error(`No API key for provider: ${provider}`);\n}\n\ninterface ServerSentEvent {\n\tevent: string | null;\n\tdata: string;\n\traw: string[];\n}\n\ninterface SseDecoderState {\n\tevent: string | null;\n\tdata: string[];\n\traw: string[];\n}\n\nconst ANTHROPIC_MESSAGE_EVENTS: ReadonlySet<string> = new Set([\n\t\"message_start\",\n\t\"message_delta\",\n\t\"message_stop\",\n\t\"content_block_start\",\n\t\"content_block_delta\",\n\t\"content_block_stop\",\n]);\n\nfunction flushSseEvent(state: SseDecoderState): ServerSentEvent | null {\n\tif (!state.event && state.data.length === 0) {\n\t\treturn null;\n\t}\n\n\tconst event: ServerSentEvent = {\n\t\tevent: state.event,\n\t\tdata: state.data.join(\"\\n\"),\n\t\traw: [...state.raw],\n\t};\n\tstate.event = null;\n\tstate.data = [];\n\tstate.raw = [];\n\treturn event;\n}\n\nfunction decodeSseLine(line: string, state: SseDecoderState): ServerSentEvent | null {\n\tif (line === \"\") {\n\t\treturn flushSseEvent(state);\n\t}\n\n\tstate.raw.push(line);\n\tif (line.startsWith(\":\")) {\n\t\treturn null;\n\t}\n\n\tconst delimiterIndex = line.indexOf(\":\");\n\tconst fieldName = delimiterIndex === -1 ? line : line.slice(0, delimiterIndex);\n\tlet value = delimiterIndex === -1 ? \"\" : line.slice(delimiterIndex + 1);\n\tif (value.startsWith(\" \")) {\n\t\tvalue = value.slice(1);\n\t}\n\n\tif (fieldName === \"event\") {\n\t\tstate.event = value;\n\t} else if (fieldName === \"data\") {\n\t\tstate.data.push(value);\n\t}\n\n\treturn null;\n}\n\nfunction nextLineBreakIndex(text: string): number {\n\tconst carriageReturnIndex = text.indexOf(\"\\r\");\n\tconst newlineIndex = text.indexOf(\"\\n\");\n\tif (carriageReturnIndex === -1) {\n\t\treturn newlineIndex;\n\t}\n\tif (newlineIndex === -1) {\n\t\treturn carriageReturnIndex;\n\t}\n\treturn Math.min(carriageReturnIndex, newlineIndex);\n}\n\nfunction consumeLine(text: string): { line: string; rest: string } | null {\n\tconst lineBreakIndex = nextLineBreakIndex(text);\n\tif (lineBreakIndex === -1) {\n\t\treturn null;\n\t}\n\n\tlet nextIndex = lineBreakIndex + 1;\n\tif (text[lineBreakIndex] === \"\\r\" && text[nextIndex] === \"\\n\") {\n\t\tnextIndex += 1;\n\t}\n\n\treturn {\n\t\tline: text.slice(0, lineBreakIndex),\n\t\trest: text.slice(nextIndex),\n\t};\n}\n\nasync function* iterateSseMessages(\n\tbody: ReadableStream<Uint8Array>,\n\tsignal?: AbortSignal,\n): AsyncGenerator<ServerSentEvent> {\n\tconst reader = body.getReader();\n\tconst decoder = new TextDecoder();\n\tconst state: SseDecoderState = { event: null, data: [], raw: [] };\n\tlet buffer = \"\";\n\tconst onAbort = () => {\n\t\tvoid reader.cancel().catch(() => {});\n\t};\n\tsignal?.addEventListener(\"abort\", onAbort, { once: true });\n\n\ttry {\n\t\twhile (true) {\n\t\t\tif (signal?.aborted) {\n\t\t\t\tthrow new Error(\"Request was aborted\");\n\t\t\t}\n\n\t\t\tconst { value, done } = await reader.read();\n\t\t\tif (done) {\n\t\t\t\tbreak;\n\t\t\t}\n\n\t\t\tbuffer += decoder.decode(value, { stream: true });\n\t\t\tlet consumed = consumeLine(buffer);\n\t\t\twhile (consumed) {\n\t\t\t\tbuffer = consumed.rest;\n\t\t\t\tconst event = decodeSseLine(consumed.line, state);\n\t\t\t\tif (event) {\n\t\t\t\t\tyield event;\n\t\t\t\t}\n\t\t\t\tconsumed = consumeLine(buffer);\n\t\t\t}\n\t\t}\n\n\t\tbuffer += decoder.decode();\n\t\tlet consumed = consumeLine(buffer);\n\t\twhile (consumed) {\n\t\t\tbuffer = consumed.rest;\n\t\t\tconst event = decodeSseLine(consumed.line, state);\n\t\t\tif (event) {\n\t\t\t\tyield event;\n\t\t\t}\n\t\t\tconsumed = consumeLine(buffer);\n\t\t}\n\n\t\tif (buffer.length > 0) {\n\t\t\tconst event = decodeSseLine(buffer, state);\n\t\t\tif (event) {\n\t\t\t\tyield event;\n\t\t\t}\n\t\t}\n\n\t\tconst trailingEvent = flushSseEvent(state);\n\t\tif (trailingEvent) {\n\t\t\tyield trailingEvent;\n\t\t}\n\t} finally {\n\t\tsignal?.removeEventListener(\"abort\", onAbort);\n\t\ttry {\n\t\t\tawait reader.cancel();\n\t\t} catch {}\n\t\treader.releaseLock();\n\t}\n}\n\nasync function* iterateAnthropicEvents(\n\tresponse: Response,\n\tsignal?: AbortSignal,\n): AsyncGenerator<RawMessageStreamEvent> {\n\tif (!response.body) {\n\t\tthrow new Error(\"Attempted to iterate over an Anthropic response with no body\");\n\t}\n\n\tlet sawMessageStart = false;\n\tlet sawMessageEnd = false;\n\n\tfor await (const sse of iterateSseMessages(response.body, signal)) {\n\t\tif (sse.event === \"error\") {\n\t\t\tthrow new Error(sse.data);\n\t\t}\n\n\t\tif (!ANTHROPIC_MESSAGE_EVENTS.has(sse.event ?? \"\")) {\n\t\t\tcontinue;\n\t\t}\n\n\t\ttry {\n\t\t\tconst event = parseJsonWithRepair<RawMessageStreamEvent>(sse.data);\n\t\t\tif (event.type === \"message_start\") {\n\t\t\t\tsawMessageStart = true;\n\t\t\t} else if (event.type === \"message_stop\") {\n\t\t\t\tsawMessageEnd = true;\n\t\t\t}\n\t\t\tyield event;\n\t\t} catch (error) {\n\t\t\tconst message = error instanceof Error ? error.message : String(error);\n\t\t\tthrow new Error(\n\t\t\t\t`Could not parse Anthropic SSE event ${sse.event}: ${message}; data=${sse.data}; raw=${sse.raw.join(\"\\\\n\")}`,\n\t\t\t);\n\t\t}\n\t}\n\n\tif (sawMessageStart && !sawMessageEnd) {\n\t\tthrow new Error(\"Anthropic stream ended before message_stop\");\n\t}\n}\n\nexport const stream: StreamFunction<\"anthropic-messages\", AnthropicOptions> = (\n\tmodel: Model<\"anthropic-messages\">,\n\tcontext: Context,\n\toptions?: AnthropicOptions,\n): AssistantMessageEventStream => {\n\tconst stream = new AssistantMessageEventStream();\n\n\t(async () => {\n\t\tconst providerThinkingLevel = supportsMidConvoEffort(model) ? (options?.effort ?? \"high\") : undefined;\n\t\tconst output: AssistantMessage = {\n\t\t\trole: \"assistant\",\n\t\t\tcontent: [],\n\t\t\tapi: model.api as Api,\n\t\t\tprovider: model.provider,\n\t\t\tmodel: model.id,\n\t\t\t...(providerThinkingLevel === undefined ? {} : { providerThinkingLevel }),\n\t\t\tusage: {\n\t\t\t\tinput: 0,\n\t\t\t\toutput: 0,\n\t\t\t\tcacheRead: 0,\n\t\t\t\tcacheWrite: 0,\n\t\t\t\ttotalTokens: 0,\n\t\t\t\tcost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0, total: 0 },\n\t\t\t},\n\t\t\tstopReason: \"pending\",\n\t\t\ttimestamp: Date.now(),\n\t\t};\n\n\t\tconst streamDeadline = createStreamDeadline(options?.streamDeadlineMs, options?.signal);\n\n\t\tlet inputTransformations: AnthropicInputTransformation[] | undefined;\n\n\t\ttry {\n\t\t\tlet client: Anthropic;\n\t\t\tlet isOAuth: boolean;\n\t\t\tlet usageModel = model;\n\n\t\t\tif (options?.client) {\n\t\t\t\tclient = options.client;\n\t\t\t\tisOAuth = false;\n\t\t\t} else {\n\t\t\t\tconst apiKey = options?.apiKey;\n\t\t\t\tassertRequestAuth(model.provider, apiKey, options?.headers);\n\n\t\t\t\tlet copilotDynamicHeaders: Record<string, string> | undefined;\n\t\t\t\tif (model.provider === \"github-copilot\") {\n\t\t\t\t\tconst hasImages = hasCopilotVisionInput(context.messages);\n\t\t\t\t\tcopilotDynamicHeaders = preserveCopilotIntegrationHeader(\n\t\t\t\t\t\tmodel.headers,\n\t\t\t\t\t\tbuildCopilotDynamicHeaders({\n\t\t\t\t\t\t\tmessages: context.messages,\n\t\t\t\t\t\t\thasImages,\n\t\t\t\t\t\t\tapiKey,\n\t\t\t\t\t\t}),\n\t\t\t\t\t);\n\t\t\t\t}\n\n\t\t\t\tconst cacheRetention = resolveCacheRetention(options?.cacheRetention, options?.env);\n\t\t\t\tconst cacheSessionId = cacheRetention === \"none\" ? undefined : options?.sessionId;\n\n\t\t\t\tconst created = createClient(\n\t\t\t\t\tmodel,\n\t\t\t\t\tapiKey,\n\t\t\t\t\tmodel.reasoning && options?.thinkingEnabled === true && (options.interleavedThinking ?? true),\n\t\t\t\t\tshouldUseFineGrainedToolStreamingBeta(model, context),\n\t\t\t\t\tshouldUseServerSideFallbackBeta(model),\n\t\t\t\t\toptions?.headers,\n\t\t\t\t\toptions?.fetch,\n\t\t\t\t\tcopilotDynamicHeaders,\n\t\t\t\t\tcacheSessionId,\n\t\t\t\t);\n\t\t\t\tclient = created.client;\n\t\t\t\tisOAuth = created.isOAuthToken;\n\t\t\t}\n\t\t\tlet params = buildParams(model, context, isOAuth, options);\n\t\t\tconst nextParams = await options?.onPayload?.(params, model);\n\t\t\tif (nextParams !== undefined) {\n\t\t\t\tparams = nextParams as MessageCreateParamsStreaming;\n\t\t\t}\n\t\t\tconst requestOptions = {\n\t\t\t\t...(streamDeadline.signal ? { signal: streamDeadline.signal } : {}),\n\t\t\t\t...(options?.timeoutMs !== undefined ? { timeout: options.timeoutMs } : {}),\n\t\t\t\tmaxRetries: 0,\n\t\t\t};\n\t\t\tconst response = await retryProviderRequest(\n\t\t\t\t() => client.messages.create({ ...params, stream: true }, requestOptions).asResponse(),\n\t\t\t\t{\n\t\t\t\t\tmaxRetries: options?.maxRetries,\n\t\t\t\t\tmaxRetryDelayMs: options?.maxRetryDelayMs,\n\t\t\t\t\tsignal: streamDeadline.signal,\n\t\t\t\t},\n\t\t\t);\n\t\t\tawait options?.onResponse?.({ status: response.status, headers: headersToRecord(response.headers) }, model);\n\t\t\tstream.push({ type: \"start\", partial: output });\n\n\t\t\ttype Block = (ThinkingContent | TextContent | (ToolCall & { partialJson: string })) & { index: number };\n\t\t\tconst blocks = output.content as Block[];\n\n\t\t\tfor await (const event of withStreamDeadline(\n\t\t\t\titerateAnthropicEvents(response, streamDeadline.signal),\n\t\t\t\tstreamDeadline.deadlineMs,\n\t\t\t\tstreamDeadline.abort,\n\t\t\t)) {\n\t\t\t\tif (event.type === \"message_start\") {\n\t\t\t\t\toutput.responseId = event.message.id;\n\t\t\t\t\toutput.model = event.message.model;\n\t\t\t\t\tconst fallbackCost =\n\t\t\t\t\t\toutput.model === model.id\n\t\t\t\t\t\t\t? undefined\n\t\t\t\t\t\t\t: model.compat?.allowedFallbackModels?.find(\n\t\t\t\t\t\t\t\t\t(fallback) => fallback.provider === model.provider && fallback.model === output.model,\n\t\t\t\t\t\t\t\t)?.cost;\n\t\t\t\t\tusageModel = fallbackCost ? { ...model, id: output.model, cost: fallbackCost } : model;\n\t\t\t\t\t// Capture initial token usage from message_start event\n\t\t\t\t\t// This ensures we have input token counts even if the stream is aborted early\n\t\t\t\t\toutput.usage.input = event.message.usage.input_tokens || 0;\n\t\t\t\t\toutput.usage.output = event.message.usage.output_tokens || 0;\n\t\t\t\t\toutput.usage.cacheRead = event.message.usage.cache_read_input_tokens || 0;\n\t\t\t\t\toutput.usage.cacheWrite = event.message.usage.cache_creation_input_tokens || 0;\n\t\t\t\t\toutput.usage.cacheWrite1h = event.message.usage.cache_creation?.ephemeral_1h_input_tokens || 0;\n\t\t\t\t\t// Anthropic doesn't provide total_tokens, compute from components\n\t\t\t\t\toutput.usage.totalTokens =\n\t\t\t\t\t\toutput.usage.input + output.usage.output + output.usage.cacheRead + output.usage.cacheWrite;\n\t\t\t\t\tcalculateCost(usageModel, output.usage);\n\t\t\t\t\tinputTransformations = getInputTransformations(event.message) ?? inputTransformations;\n\t\t\t\t} else if (event.type === \"content_block_start\") {\n\t\t\t\t\tif (event.content_block.type === \"text\") {\n\t\t\t\t\t\tconst block: Block = {\n\t\t\t\t\t\t\ttype: \"text\",\n\t\t\t\t\t\t\ttext: event.content_block.text ?? \"\",\n\t\t\t\t\t\t\tindex: event.index,\n\t\t\t\t\t\t};\n\t\t\t\t\t\toutput.content.push(block);\n\t\t\t\t\t\tstream.push({ type: \"text_start\", contentIndex: output.content.length - 1, partial: output });\n\t\t\t\t\t} else if (event.content_block.type === \"thinking\") {\n\t\t\t\t\t\tconst block: Block = {\n\t\t\t\t\t\t\ttype: \"thinking\",\n\t\t\t\t\t\t\tthinking: event.content_block.thinking ?? \"\",\n\t\t\t\t\t\t\tthinkingSignature: event.content_block.signature ?? \"\",\n\t\t\t\t\t\t\tindex: event.index,\n\t\t\t\t\t\t};\n\t\t\t\t\t\toutput.content.push(block);\n\t\t\t\t\t\tstream.push({ type: \"thinking_start\", contentIndex: output.content.length - 1, partial: output });\n\t\t\t\t\t} else if (event.content_block.type === \"redacted_thinking\") {\n\t\t\t\t\t\tconst block: Block = {\n\t\t\t\t\t\t\ttype: \"thinking\",\n\t\t\t\t\t\t\tthinking: \"[Reasoning redacted]\",\n\t\t\t\t\t\t\tthinkingSignature: event.content_block.data,\n\t\t\t\t\t\t\tredacted: true,\n\t\t\t\t\t\t\tindex: event.index,\n\t\t\t\t\t\t};\n\t\t\t\t\t\toutput.content.push(block);\n\t\t\t\t\t\tstream.push({ type: \"thinking_start\", contentIndex: output.content.length - 1, partial: output });\n\t\t\t\t\t} else if (event.content_block.type === \"tool_use\") {\n\t\t\t\t\t\tconst block: Block = {\n\t\t\t\t\t\t\ttype: \"toolCall\",\n\t\t\t\t\t\t\tid: event.content_block.id,\n\t\t\t\t\t\t\tname: isOAuth\n\t\t\t\t\t\t\t\t? fromClaudeCodeName(event.content_block.name, context.tools)\n\t\t\t\t\t\t\t\t: event.content_block.name,\n\t\t\t\t\t\t\targuments: (event.content_block.input as Record<string, any>) ?? {},\n\t\t\t\t\t\t\tpartialJson: \"\",\n\t\t\t\t\t\t\tindex: event.index,\n\t\t\t\t\t\t};\n\t\t\t\t\t\toutput.content.push(block);\n\t\t\t\t\t\tstream.push({ type: \"toolcall_start\", contentIndex: output.content.length - 1, partial: output });\n\t\t\t\t\t} else if ((event.content_block as { type?: string }).type === \"fallback\") {\n\t\t\t\t\t\t// Server-side fallback boundary: \"the `fallback` block (an ordinary\n\t\t\t\t\t\t// `content_block_start` and `content_block_stop` pair with no deltas) marks the\n\t\t\t\t\t\t// boundary\", and clients must \"Keep it exactly where it appeared. The API uses\n\t\t\t\t\t\t// its position to validate the thinking blocks around it\".\n\t\t\t\t\t\t// https://platform.claude.com/docs/en/build-with-claude/refusals-and-fallback\n\t\t\t\t\t\t//\n\t\t\t\t\t\t// The SDK types lag this block, so read it through a narrow cast.\n\t\t\t\t\t\tconst fallbackBlock = event.content_block as unknown as {\n\t\t\t\t\t\t\tfrom?: { model?: string };\n\t\t\t\t\t\t\tto?: { model?: string };\n\t\t\t\t\t\t};\n\t\t\t\t\t\tconst toModel = fallbackBlock.to?.model;\n\t\t\t\t\t\toutput.content.push({\n\t\t\t\t\t\t\ttype: \"fallback\",\n\t\t\t\t\t\t\tfromModel: fallbackBlock.from?.model ?? output.model,\n\t\t\t\t\t\t\ttoModel: toModel ?? output.model,\n\t\t\t\t\t\t});\n\t\t\t\t\t\t// `message_start` named the requested model, so on a mid-output fallback the\n\t\t\t\t\t\t// serving model is only knowable here. Re-derive pricing the same way the\n\t\t\t\t\t\t// `message_start` branch does, so the returned message is costed at the rates\n\t\t\t\t\t\t// of the model that actually produced it.\n\t\t\t\t\t\tif (toModel && toModel !== output.model) {\n\t\t\t\t\t\t\toutput.model = toModel;\n\t\t\t\t\t\t\tconst fallbackCost = model.compat?.allowedFallbackModels?.find(\n\t\t\t\t\t\t\t\t(fallback) => fallback.provider === model.provider && fallback.model === toModel,\n\t\t\t\t\t\t\t)?.cost;\n\t\t\t\t\t\t\tusageModel = fallbackCost ? { ...model, id: toModel, cost: fallbackCost } : model;\n\t\t\t\t\t\t}\n\t\t\t\t\t}\n\t\t\t\t} else if (event.type === \"content_block_delta\") {\n\t\t\t\t\tif (event.delta.type === \"text_delta\") {\n\t\t\t\t\t\tconst index = blocks.findIndex((b) => b.index === event.index);\n\t\t\t\t\t\tconst block = blocks[index];\n\t\t\t\t\t\tif (block && block.type === \"text\") {\n\t\t\t\t\t\t\tblock.text += event.delta.text;\n\t\t\t\t\t\t\tstream.push({\n\t\t\t\t\t\t\t\ttype: \"text_delta\",\n\t\t\t\t\t\t\t\tcontentIndex: index,\n\t\t\t\t\t\t\t\tdelta: event.delta.text,\n\t\t\t\t\t\t\t\tpartial: output,\n\t\t\t\t\t\t\t});\n\t\t\t\t\t\t}\n\t\t\t\t\t} else if (event.delta.type === \"thinking_delta\") {\n\t\t\t\t\t\tconst index = blocks.findIndex((b) => b.index === event.index);\n\t\t\t\t\t\tconst block = blocks[index];\n\t\t\t\t\t\tif (block && block.type === \"thinking\") {\n\t\t\t\t\t\t\tblock.thinking += event.delta.thinking;\n\t\t\t\t\t\t\tstream.push({\n\t\t\t\t\t\t\t\ttype: \"thinking_delta\",\n\t\t\t\t\t\t\t\tcontentIndex: index,\n\t\t\t\t\t\t\t\tdelta: event.delta.thinking,\n\t\t\t\t\t\t\t\tpartial: output,\n\t\t\t\t\t\t\t});\n\t\t\t\t\t\t}\n\t\t\t\t\t} else if (event.delta.type === \"input_json_delta\") {\n\t\t\t\t\t\tconst index = blocks.findIndex((b) => b.index === event.index);\n\t\t\t\t\t\tconst block = blocks[index];\n\t\t\t\t\t\tif (block && block.type === \"toolCall\") {\n\t\t\t\t\t\t\tblock.partialJson += event.delta.partial_json;\n\t\t\t\t\t\t\tblock.arguments = parseStreamingJson(block.partialJson);\n\t\t\t\t\t\t\tstream.push({\n\t\t\t\t\t\t\t\ttype: \"toolcall_delta\",\n\t\t\t\t\t\t\t\tcontentIndex: index,\n\t\t\t\t\t\t\t\tdelta: event.delta.partial_json,\n\t\t\t\t\t\t\t\tpartial: output,\n\t\t\t\t\t\t\t});\n\t\t\t\t\t\t}\n\t\t\t\t\t} else if (event.delta.type === \"signature_delta\") {\n\t\t\t\t\t\tconst index = blocks.findIndex((b) => b.index === event.index);\n\t\t\t\t\t\tconst block = blocks[index];\n\t\t\t\t\t\tif (block && block.type === \"thinking\") {\n\t\t\t\t\t\t\tblock.thinkingSignature = block.thinkingSignature || \"\";\n\t\t\t\t\t\t\tblock.thinkingSignature += event.delta.signature;\n\t\t\t\t\t\t}\n\t\t\t\t\t}\n\t\t\t\t} else if (event.type === \"content_block_stop\") {\n\t\t\t\t\tconst index = blocks.findIndex((b) => b.index === event.index);\n\t\t\t\t\tconst block = blocks[index];\n\t\t\t\t\tif (block) {\n\t\t\t\t\t\tdelete (block as any).index;\n\t\t\t\t\t\tif (block.type === \"text\") {\n\t\t\t\t\t\t\tstream.push({\n\t\t\t\t\t\t\t\ttype: \"text_end\",\n\t\t\t\t\t\t\t\tcontentIndex: index,\n\t\t\t\t\t\t\t\tcontent: block.text,\n\t\t\t\t\t\t\t\tpartial: output,\n\t\t\t\t\t\t\t});\n\t\t\t\t\t\t} else if (block.type === \"thinking\") {\n\t\t\t\t\t\t\tstream.push({\n\t\t\t\t\t\t\t\ttype: \"thinking_end\",\n\t\t\t\t\t\t\t\tcontentIndex: index,\n\t\t\t\t\t\t\t\tcontent: block.thinking,\n\t\t\t\t\t\t\t\tpartial: output,\n\t\t\t\t\t\t\t});\n\t\t\t\t\t\t} else if (block.type === \"toolCall\") {\n\t\t\t\t\t\t\tblock.arguments = parseStreamingJson(block.partialJson);\n\t\t\t\t\t\t\t// Finalize in-place and strip the scratch buffer so replay only\n\t\t\t\t\t\t\t// carries parsed arguments.\n\t\t\t\t\t\t\tdelete (block as { partialJson?: string }).partialJson;\n\t\t\t\t\t\t\tstream.push({\n\t\t\t\t\t\t\t\ttype: \"toolcall_end\",\n\t\t\t\t\t\t\t\tcontentIndex: index,\n\t\t\t\t\t\t\t\ttoolCall: block,\n\t\t\t\t\t\t\t\tpartial: output,\n\t\t\t\t\t\t\t});\n\t\t\t\t\t\t}\n\t\t\t\t\t}\n\t\t\t\t} else if (event.type === \"message_delta\") {\n\t\t\t\t\tif (event.delta.stop_reason) {\n\t\t\t\t\t\toutput.rawStopReason = event.delta.stop_reason;\n\t\t\t\t\t\tconst stopReasonResult = mapStopReason(event.delta.stop_reason, event.delta.stop_details);\n\t\t\t\t\t\toutput.stopReason = stopReasonResult.stopReason;\n\t\t\t\t\t\tif (stopReasonResult.errorMessage) {\n\t\t\t\t\t\t\toutput.errorMessage = stopReasonResult.errorMessage;\n\t\t\t\t\t\t}\n\t\t\t\t\t}\n\t\t\t\t\t// Only update usage fields if present (not null).\n\t\t\t\t\t// Preserves input_tokens from message_start when proxies omit it in message_delta.\n\t\t\t\t\tif (event.usage) {\n\t\t\t\t\t\tif (event.usage.input_tokens != null) {\n\t\t\t\t\t\t\toutput.usage.input = event.usage.input_tokens;\n\t\t\t\t\t\t}\n\t\t\t\t\t\tif (event.usage.output_tokens != null) {\n\t\t\t\t\t\t\toutput.usage.output = event.usage.output_tokens;\n\t\t\t\t\t\t}\n\t\t\t\t\t\tif (event.usage.cache_read_input_tokens != null) {\n\t\t\t\t\t\t\toutput.usage.cacheRead = event.usage.cache_read_input_tokens;\n\t\t\t\t\t\t}\n\t\t\t\t\t\tif (event.usage.cache_creation_input_tokens != null) {\n\t\t\t\t\t\t\toutput.usage.cacheWrite = event.usage.cache_creation_input_tokens;\n\t\t\t\t\t\t}\n\t\t\t\t\t\t// Anthropic reports reasoning tokens in `output_tokens_details.thinking_tokens` on the\n\t\t\t\t\t\t// final message_delta usage (a subset of output_tokens). SDK 0.91.1 omits the field from\n\t\t\t\t\t\t// its Usage type, so read it through a narrow cast. Verified against the live API.\n\t\t\t\t\t\tconst thinkingTokens = (event.usage as { output_tokens_details?: { thinking_tokens?: number } })\n\t\t\t\t\t\t\t.output_tokens_details?.thinking_tokens;\n\t\t\t\t\t\tif (thinkingTokens != null) {\n\t\t\t\t\t\t\toutput.usage.reasoning = thinkingTokens;\n\t\t\t\t\t\t}\n\t\t\t\t\t}\n\t\t\t\t\t// Anthropic doesn't provide total_tokens, compute from components\n\t\t\t\t\toutput.usage.totalTokens =\n\t\t\t\t\t\toutput.usage.input + output.usage.output + output.usage.cacheRead + output.usage.cacheWrite;\n\t\t\t\t\tcalculateCost(usageModel, output.usage);\n\t\t\t\t\t// A non-empty final fallback report supersedes the request-start report with\n\t\t\t\t\t// the serving model's transformations. An empty report must not erase drops\n\t\t\t\t\t// already reported at message_start.\n\t\t\t\t\tconst servingTransformations = getInputTransformations(event);\n\t\t\t\t\tif (servingTransformations && servingTransformations.length > 0) {\n\t\t\t\t\t\tinputTransformations = servingTransformations;\n\t\t\t\t\t}\n\t\t\t\t\taddEarlierAttemptCosts(output, model, usageModel, event);\n\t\t\t\t}\n\t\t\t}\n\n\t\t\tif (inputTransformations) recordInputTransformations(output, inputTransformations);\n\t\t\tif (options?.signal?.aborted) {\n\t\t\t\tthrow new Error(\"Request was aborted\");\n\t\t\t}\n\n\t\t\tif (output.stopReason === \"pending\") {\n\t\t\t\tthrow new Error(\"Anthropic stream ended without a stop reason\");\n\t\t\t}\n\t\t\tif (output.stopReason === \"aborted\" || output.stopReason === \"error\") {\n\t\t\t\tthrow new Error(output.errorMessage || \"An unknown error occurred\");\n\t\t\t}\n\n\t\t\tstream.push({ type: \"done\", reason: output.stopReason, message: output });\n\t\t\tstream.end();\n\t\t} catch (error) {\n\t\t\tif (\n\t\t\t\tinputTransformations &&\n\t\t\t\t!output.diagnostics?.some((diagnostic) => diagnostic.type === \"anthropic_input_transformations\")\n\t\t\t) {\n\t\t\t\trecordInputTransformations(output, inputTransformations);\n\t\t\t}\n\t\t\tfor (const block of output.content) {\n\t\t\t\tdelete (block as { index?: number }).index;\n\t\t\t\t// partialJson is only a streaming scratch buffer; never persist it.\n\t\t\t\tdelete (block as { partialJson?: string }).partialJson;\n\t\t\t}\n\t\t\toutput.stopReason = options?.signal?.aborted ? \"aborted\" : \"error\";\n\t\t\toutput.errorMessage = error instanceof Error ? error.message : JSON.stringify(error);\n\t\t\tstream.push({ type: \"error\", reason: output.stopReason, error: output });\n\t\t\tstream.end();\n\t\t} finally {\n\t\t\tstreamDeadline.cleanup();\n\t\t}\n\t})();\n\n\treturn stream;\n};\n\n/**\n * Map ThinkingLevel to Anthropic effort levels for adaptive thinking.\n * Note: effort \"max\" is available on all adaptive-thinking Claude models, while native\n * \"xhigh\" is available on Opus 4.7, Opus 4.8, Opus 5, Sonnet 5, Fable 5, and Fable 5.1 —\n * the models the generator merges an `xhigh` mapping onto.\n */\nfunction mapThinkingLevelToEffort(\n\tmodel: Model<\"anthropic-messages\">,\n\tlevel: SimpleStreamOptions[\"reasoning\"],\n): AnthropicEffort {\n\tconst mapped = level ? model.thinkingLevelMap?.[level] : undefined;\n\tif (typeof mapped === \"string\") return mapped as AnthropicEffort;\n\n\tswitch (level) {\n\t\tcase \"minimal\":\n\t\tcase \"low\":\n\t\t\treturn \"low\";\n\t\tcase \"medium\":\n\t\t\treturn \"medium\";\n\t\tcase \"high\":\n\t\t\treturn \"high\";\n\t\tdefault:\n\t\t\treturn \"high\";\n\t}\n}\n\nexport const streamSimple: StreamFunction<\"anthropic-messages\", SimpleStreamOptions> = (\n\tmodel: Model<\"anthropic-messages\">,\n\tcontext: Context,\n\toptions?: SimpleStreamOptions,\n): AssistantMessageEventStream => {\n\tassertRequestAuth(model.provider, options?.apiKey, options?.headers);\n\n\tconst base = {\n\t\t...buildBaseOptions(model, context, options, options?.apiKey),\n\t\ttoolChoice: options?.toolChoice,\n\t} satisfies AnthropicOptions;\n\tif (!options?.reasoning) {\n\t\treturn stream(model, context, {\n\t\t\t...base,\n\t\t\tthinkingEnabled: false,\n\t\t} satisfies AnthropicOptions);\n\t}\n\n\t// For models with adaptive thinking: use an effort level.\n\t// For older models: use budget-based thinking.\n\tif (model.compat?.forceAdaptiveThinking === true) {\n\t\tconst effort = mapThinkingLevelToEffort(model, options.reasoning);\n\t\treturn stream(model, context, {\n\t\t\t...base,\n\t\t\tthinkingEnabled: true,\n\t\t\teffort,\n\t\t} satisfies AnthropicOptions);\n\t}\n\n\t// Undefined means the caller did not request an output cap; let the helper use the model cap.\n\t// Do not coerce to 0 here, or the thinking budget would become the entire max_tokens value.\n\tconst adjusted = adjustMaxTokensForThinking(\n\t\tbase.maxTokens,\n\t\tmodel.maxTokens,\n\t\toptions.reasoning,\n\t\toptions.thinkingBudgets,\n\t);\n\n\tconst maxTokens = clampMaxTokensToContext(model, context, adjusted.maxTokens);\n\n\treturn stream(model, context, {\n\t\t...base,\n\t\tmaxTokens,\n\t\tthinkingEnabled: true,\n\t\tthinkingBudgetTokens: Math.min(adjusted.thinkingBudget, Math.max(0, maxTokens - 1024)),\n\t} satisfies AnthropicOptions);\n};\n\nfunction isOAuthToken(apiKey: string): boolean {\n\treturn apiKey.includes(\"sk-ant-oat\");\n}\n\nfunction createClient(\n\tmodel: Model<\"anthropic-messages\">,\n\tapiKey: string | undefined,\n\tinterleavedThinking: boolean,\n\tuseFineGrainedToolStreamingBeta: boolean,\n\tuseServerSideFallbackBeta: boolean,\n\toptionsHeaders?: ProviderHeaders,\n\tfetch?: typeof globalThis.fetch,\n\tdynamicHeaders?: Record<string, string>,\n\tsessionId?: string,\n): { client: Anthropic; isOAuthToken: boolean } {\n\t// Adaptive thinking models have interleaved thinking built in, so skip the beta header.\n\tconst needsInterleavedBeta = interleavedThinking && model.compat?.forceAdaptiveThinking !== true;\n\tconst betaFeatures: string[] = [];\n\tif (useFineGrainedToolStreamingBeta) {\n\t\tbetaFeatures.push(FINE_GRAINED_TOOL_STREAMING_BETA);\n\t}\n\tif (needsInterleavedBeta) {\n\t\tbetaFeatures.push(INTERLEAVED_THINKING_BETA);\n\t}\n\tif (useServerSideFallbackBeta) {\n\t\tbetaFeatures.push(SERVER_SIDE_FALLBACK_BETA);\n\t}\n\tif (shouldUseThinkingBindingControlsBeta(model)) {\n\t\tbetaFeatures.push(THINKING_BINDING_CONTROLS_BETA);\n\t}\n\tif (supportsMidConvoEffort(model)) {\n\t\tbetaFeatures.push(MID_CONVERSATION_OUTPUT_CONFIG_BETA, THINKING_BINDING_CONTROLS_BETA);\n\t}\n\tconst uniqueBetaFeatures = [...new Set(betaFeatures)];\n\n\t// Copilot: Bearer auth, selective betas.\n\tif (model.provider === \"github-copilot\") {\n\t\tconst client = new Anthropic({\n\t\t\tapiKey: null,\n\t\t\tauthToken: apiKey ?? null,\n\t\t\tbaseURL: model.baseUrl,\n\t\t\tdangerouslyAllowBrowser: true,\n\t\t\tfetch,\n\t\t\tdefaultHeaders: mergeClientHeaders(\n\t\t\t\t{\n\t\t\t\t\taccept: \"application/json\",\n\t\t\t\t\t\"anthropic-dangerous-direct-browser-access\": \"true\",\n\t\t\t\t\t...(uniqueBetaFeatures.length > 0 ? { \"anthropic-beta\": uniqueBetaFeatures.join(\",\") } : {}),\n\t\t\t\t},\n\t\t\t\tmodel.headers,\n\t\t\t\tdynamicHeaders,\n\t\t\t\toptionsHeaders,\n\t\t\t),\n\t\t});\n\n\t\treturn { client, isOAuthToken: false };\n\t}\n\n\t// OAuth: Bearer auth, Claude Code identity headers\n\tif (apiKey && isOAuthToken(apiKey)) {\n\t\tconst client = new Anthropic({\n\t\t\tapiKey: null,\n\t\t\tauthToken: apiKey,\n\t\t\tbaseURL: model.baseUrl,\n\t\t\tdangerouslyAllowBrowser: true,\n\t\t\tfetch,\n\t\t\tdefaultHeaders: mergeClientHeaders(\n\t\t\t\t{\n\t\t\t\t\taccept: \"application/json\",\n\t\t\t\t\t\"anthropic-dangerous-direct-browser-access\": \"true\",\n\t\t\t\t\t\"anthropic-beta\": [\"claude-code-20250219\", \"oauth-2025-04-20\", ...uniqueBetaFeatures].join(\",\"),\n\t\t\t\t\t\"user-agent\": `claude-cli/${claudeCodeVersion}`,\n\t\t\t\t\t\"x-app\": \"cli\",\n\t\t\t\t},\n\t\t\t\tmodel.headers,\n\t\t\t\toptionsHeaders,\n\t\t\t),\n\t\t});\n\n\t\treturn { client, isOAuthToken: true };\n\t}\n\n\t// API key or header-owned auth.\n\tconst sessionAffinityHeaders: ProviderHeaders =\n\t\tsessionId && getAnthropicCompat(model).sendSessionAffinityHeaders ? { \"x-session-affinity\": sessionId } : {};\n\tconst defaultHeaders = mergeClientHeaders(\n\t\t{\n\t\t\taccept: \"application/json\",\n\t\t\t\"anthropic-dangerous-direct-browser-access\": \"true\",\n\t\t\t...(uniqueBetaFeatures.length > 0 ? { \"anthropic-beta\": uniqueBetaFeatures.join(\",\") } : {}),\n\t\t},\n\t\tsessionAffinityHeaders,\n\t\tmodel.headers,\n\t\toptionsHeaders,\n\t);\n\tconst client = new Anthropic({\n\t\tapiKey: apiKey ?? null,\n\t\tauthToken: null,\n\t\tbaseURL: model.baseUrl,\n\t\tdangerouslyAllowBrowser: true,\n\t\tfetch,\n\t\tdefaultHeaders,\n\t});\n\n\treturn { client, isOAuthToken: false };\n}\n\nfunction buildParams(\n\tmodel: Model<\"anthropic-messages\">,\n\tcontext: Context,\n\tisOAuthToken: boolean,\n\toptions?: AnthropicOptions,\n): MessageCreateParamsStreamingWithFallbacks {\n\tconst { cacheControl } = getCacheControl(model, options?.cacheRetention, options?.env);\n\tconst compat = getAnthropicCompat(model);\n\tconst transformedMessages = transformMessages(context.messages, model, normalizeToolCallId);\n\tconst normalizeToolName = isOAuthToken ? toClaudeCodeName : (name: string) => name;\n\tconst toolPlacement = splitDeferredTools(\n\t\t{ ...context, messages: transformedMessages },\n\t\tcompat.supportsToolReferences,\n\t\tnormalizeToolName,\n\t);\n\tlet immediateTools = toolPlacement.immediate;\n\tlet deferredTools = [...toolPlacement.deferred.values()];\n\tif (immediateTools.length === 0 && deferredTools.length > 0) {\n\t\timmediateTools = deferredTools;\n\t\tdeferredTools = [];\n\t}\n\tconst deferredToolNames = new Set(deferredTools.map((tool) => normalizeToolName(tool.name)));\n\tconst converted = convertMessages(\n\t\ttransformedMessages,\n\t\tisOAuthToken,\n\t\tcacheControl,\n\t\tcompat.allowEmptySignature,\n\t\tdeferredToolNames,\n\t\tnormalizeToolName,\n\t\tsupportsMidConvoEffort(model) ? model.provider : undefined,\n\t);\n\tconst activeEffort = options?.effort ?? \"high\";\n\tconst params: MessageCreateParamsStreamingWithFallbacks = {\n\t\tmodel: model.id,\n\t\tmessages: (supportsMidConvoEffort(model)\n\t\t\t? insertThinkingLevelMessages(converted, activeEffort)\n\t\t\t: converted.messages) as MessageParam[],\n\t\tmax_tokens: options?.maxTokens ?? model.maxTokens,\n\t\tstream: true,\n\t};\n\n\t// For OAuth tokens, we MUST include Claude Code identity\n\tif (isOAuthToken) {\n\t\tparams.system = [\n\t\t\t{\n\t\t\t\ttype: \"text\",\n\t\t\t\ttext: \"You are Claude Code, Anthropic's official CLI for Claude.\",\n\t\t\t\t...(cacheControl ? { cache_control: cacheControl } : {}),\n\t\t\t},\n\t\t];\n\t\tif (context.systemPrompt) {\n\t\t\tparams.system.push({\n\t\t\t\ttype: \"text\",\n\t\t\t\ttext: sanitizeSurrogates(context.systemPrompt),\n\t\t\t\t...(cacheControl ? { cache_control: cacheControl } : {}),\n\t\t\t});\n\t\t}\n\t} else if (context.systemPrompt) {\n\t\t// Add cache control to system prompt for non-OAuth tokens\n\t\tparams.system = [\n\t\t\t{\n\t\t\t\ttype: \"text\",\n\t\t\t\ttext: sanitizeSurrogates(context.systemPrompt),\n\t\t\t\t...(cacheControl ? { cache_control: cacheControl } : {}),\n\t\t\t},\n\t\t];\n\t}\n\n\t// Temperature is incompatible with extended thinking and unsupported on Claude Opus 4.7+.\n\tif (\n\t\toptions?.temperature !== undefined &&\n\t\t!options?.thinkingEnabled &&\n\t\t!supportsMidConvoEffort(model) &&\n\t\tcompat.supportsTemperature\n\t) {\n\t\tparams.temperature = options.temperature;\n\t}\n\n\tif (immediateTools.length > 0 || deferredTools.length > 0) {\n\t\tparams.tools = [\n\t\t\t...convertTools(\n\t\t\t\timmediateTools,\n\t\t\t\tisOAuthToken,\n\t\t\t\tcompat.supportsEagerToolInputStreaming,\n\t\t\t\tcompat.supportsStrictTools,\n\t\t\t\tcompat.supportsCacheControlOnTools ? cacheControl : undefined,\n\t\t\t),\n\t\t\t...convertTools(\n\t\t\t\tdeferredTools,\n\t\t\t\tisOAuthToken,\n\t\t\t\tcompat.supportsEagerToolInputStreaming,\n\t\t\t\tcompat.supportsStrictTools,\n\t\t\t\tundefined,\n\t\t\t\ttrue,\n\t\t\t),\n\t\t];\n\t}\n\n\t// Managed effort models always use adaptive thinking. Their per-turn markers\n\t// carry the actual effort; the top-level high value deliberately stays stable.\n\tif (supportsMidConvoEffort(model)) {\n\t\tparams.thinking = {\n\t\t\ttype: \"adaptive\",\n\t\t\tdisplay: options?.thinkingDisplay ?? \"summarized\",\n\t\t\tblock_binding: { prefix_mismatch_behavior: \"drop_block\" },\n\t\t} as unknown as NonNullable<MessageCreateParamsStreaming[\"thinking\"]>;\n\t\tparams.output_config = { effort: \"high\" };\n\t} else if (model.reasoning) {\n\t\tif (options?.thinkingEnabled) {\n\t\t\t// Default to \"summarized\" so Opus 4.7 and Mythos Preview behave like\n\t\t\t// older Claude 4 models (whose API default is also \"summarized\").\n\t\t\tconst display: AnthropicThinkingDisplay = options.thinkingDisplay ?? \"summarized\";\n\t\t\tif (model.compat?.forceAdaptiveThinking === true) {\n\t\t\t\t// Adaptive thinking: Claude decides when and how much to think.\n\t\t\t\tparams.thinking = { type: \"adaptive\", display };\n\t\t\t\tif (options.effort) {\n\t\t\t\t\t// The Anthropic SDK types can lag newly supported effort values such as \"xhigh\".\n\t\t\t\t\tparams.output_config =\n\t\t\t\t\t\toptions.effort === \"xhigh\"\n\t\t\t\t\t\t\t? ({ effort: options.effort } as unknown as NonNullable<\n\t\t\t\t\t\t\t\t\tMessageCreateParamsStreaming[\"output_config\"]\n\t\t\t\t\t\t\t\t>)\n\t\t\t\t\t\t\t: { effort: options.effort };\n\t\t\t\t}\n\t\t\t} else {\n\t\t\t\t// Budget-based thinking for older models\n\t\t\t\tparams.thinking = {\n\t\t\t\t\ttype: \"enabled\",\n\t\t\t\t\tbudget_tokens: options.thinkingBudgetTokens || 1024,\n\t\t\t\t\tdisplay,\n\t\t\t\t};\n\t\t\t}\n\t\t} else if (options?.thinkingEnabled === false && model.thinkingLevelMap?.off !== null) {\n\t\t\tparams.thinking = { type: \"disabled\" };\n\t\t}\n\n\t\t// `block_binding` must accompany the beta header on *every* request for a model that\n\t\t// enforces the conversation check, not only on reasoning turns. The header alone leaves\n\t\t// `prefix_mismatch_behavior` at its `\"error\"` default, which is the 400 this opts out of.\n\t\t// `streamSimple` sets `thinkingEnabled: false` whenever no reasoning level is requested,\n\t\t// and Claude Fable 5.1 denies thinking-off (`thinkingLevelMap.off === null`), so without\n\t\t// this the no-reasoning path would send the header with no field at all.\n\t\tif (shouldUseThinkingBindingControlsBeta(model)) {\n\t\t\tconst blockBinding = { block_binding: { prefix_mismatch_behavior: \"drop_block\" } };\n\t\t\t// The Anthropic SDK types lag the `thinking-binding-controls-2026-08-01` beta.\n\t\t\tif (params.thinking?.type === \"adaptive\" || params.thinking?.type === \"enabled\") {\n\t\t\t\t// Accepted alongside both thinking types. It only changes what happens to a block\n\t\t\t\t// replayed behind a changed prefix; the model check always drops regardless.\n\t\t\t\tparams.thinking = { ...params.thinking, ...blockBinding } as unknown as NonNullable<\n\t\t\t\t\tMessageCreateParamsStreaming[\"thinking\"]\n\t\t\t\t>;\n\t\t\t} else if (!params.thinking && model.compat?.forceAdaptiveThinking === true) {\n\t\t\t\t// Adaptive thinking is always on for these models, so omitting `thinking` and\n\t\t\t\t// sending `{type: \"adaptive\"}` are equivalent. `display` stays absent to keep the\n\t\t\t\t// API's `\"omitted\"` default, which is exactly what omitting `thinking` produced.\n\t\t\t\tparams.thinking = { type: \"adaptive\", ...blockBinding } as unknown as NonNullable<\n\t\t\t\t\tMessageCreateParamsStreaming[\"thinking\"]\n\t\t\t\t>;\n\t\t\t}\n\t\t}\n\t}\n\n\tif (options?.metadata) {\n\t\tconst userId = options.metadata.user_id;\n\t\tif (typeof userId === \"string\") {\n\t\t\tparams.metadata = { user_id: userId };\n\t\t}\n\t}\n\n\tif (options?.toolChoice) {\n\t\tconst requested = options.toolChoice;\n\t\tconst isForced = requested === \"any\" || (typeof requested !== \"string\" && requested.type === \"tool\");\n\t\tif (isForced && model.compat?.supportsForcedToolChoice === false) {\n\t\t\t// \"The exceptions are Claude Fable 5.1 and Claude Mythos 5.1, which reject forced tool\n\t\t\t// use on every request with a 400 error. On those models, use\n\t\t\t// `tool_choice: {\"type\": \"auto\"}` with strict tool use or structured outputs instead.\"\n\t\t\t// https://platform.claude.com/docs/en/build-with-claude/thinking\n\t\t\t//\n\t\t\t// Reject the request rather than rewriting it. Silently substituting `auto` would\n\t\t\t// discard an explicit caller instruction and make `AnthropicOptions.toolChoice`'s\n\t\t\t// declared shape a lie; the caller asked the model to call a tool, and quietly asking\n\t\t\t// it to decide instead is a different request. Failing here matches how this package\n\t\t\t// handles other explicitly requested capabilities a model cannot honor, and surfaces\n\t\t\t// the remedy before a round trip that would 400 anyway. Callers that want the\n\t\t\t// substitution can make it themselves, and can branch on\n\t\t\t// `model.compat.supportsForcedToolChoice` to decide.\n\t\t\tconst requestedLabel = typeof requested === \"string\" ? requested : `tool \"${requested.name}\"`;\n\t\t\tthrow new Error(\n\t\t\t\t`Model ${model.id} does not support forced tool choice (requested: ${requestedLabel}). ` +\n\t\t\t\t\t`Use toolChoice \"auto\" with strict tool use or structured outputs instead.`,\n\t\t\t);\n\t\t}\n\t\tif (typeof requested === \"string\") {\n\t\t\tparams.tool_choice = { type: requested };\n\t\t} else {\n\t\t\tparams.tool_choice = requested;\n\t\t}\n\t}\n\n\tconst allowedFallbackModels = model.compat?.allowedFallbackModels;\n\tif (allowedFallbackModels && allowedFallbackModels.length > 0) {\n\t\tparams.fallbacks = allowedFallbackModels.map((fallback) => ({ model: fallback.model }));\n\t}\n\n\treturn params;\n}\n\n// Normalize tool call IDs to match Anthropic's required pattern and length\nfunction normalizeToolCallId(id: string): string {\n\treturn id.replace(/[^a-zA-Z0-9_-]/g, \"_\").slice(0, 64);\n}\n\nfunction convertToolResult(\n\tmsg: ToolResultMessage,\n\tisOAuthToken: boolean,\n\tdeferredToolNames: ReadonlySet<string>,\n\tloadedToolNames: Set<string>,\n\tnormalizeToolName: (name: string) => string,\n): { toolResult: ContentBlockParam; siblingContent: ContentBlockParam[] } {\n\tconst references: Array<{ type: \"tool_reference\"; tool_name: string }> = [];\n\tfor (const name of msg.addedToolNames ?? []) {\n\t\tconst normalizedName = normalizeToolName(name);\n\t\tif (!deferredToolNames.has(normalizedName) || loadedToolNames.has(normalizedName)) continue;\n\t\tloadedToolNames.add(normalizedName);\n\t\treferences.push({\n\t\t\ttype: \"tool_reference\",\n\t\t\ttool_name: isOAuthToken ? toClaudeCodeName(name) : name,\n\t\t});\n\t}\n\tconst convertedContent = convertContentBlocks(msg.content);\n\t// Anthropic rejects tool references mixed with ordinary tool-result content.\n\treturn {\n\t\ttoolResult: {\n\t\t\ttype: \"tool_result\",\n\t\t\ttool_use_id: msg.toolCallId,\n\t\t\tcontent: references.length > 0 ? references : convertedContent,\n\t\t\tis_error: msg.isError,\n\t\t},\n\t\tsiblingContent:\n\t\t\treferences.length === 0\n\t\t\t\t? []\n\t\t\t\t: typeof convertedContent === \"string\"\n\t\t\t\t\t? [{ type: \"text\", text: convertedContent }]\n\t\t\t\t\t: convertedContent,\n\t};\n}\n\nfunction convertMessages(\n\ttransformedMessages: Message[],\n\tisOAuthToken: boolean,\n\tcacheControl?: CacheControlEphemeral,\n\tallowEmptySignature = false,\n\tdeferredToolNames: ReadonlySet<string> = new Set(),\n\tnormalizeToolName: (name: string) => string = (name) => name,\n\tmanagedProvider?: string,\n): ConvertedAnthropicMessages {\n\tconst params: MessageParam[] = [];\n\tconst assistantLevels = new Map<number, AnthropicEffort>();\n\tconst loadedToolNames = new Set<string>();\n\n\tfor (let i = 0; i < transformedMessages.length; i++) {\n\t\tconst msg = transformedMessages[i];\n\n\t\tif (msg.role === \"user\") {\n\t\t\tif (typeof msg.content === \"string\") {\n\t\t\t\tif (msg.content.trim().length > 0) {\n\t\t\t\t\tparams.push({\n\t\t\t\t\t\trole: \"user\",\n\t\t\t\t\t\tcontent: sanitizeSurrogates(msg.content),\n\t\t\t\t\t});\n\t\t\t\t}\n\t\t\t} else {\n\t\t\t\tconst blocks: ContentBlockParam[] = msg.content.map((item) => {\n\t\t\t\t\tif (item.type === \"text\") {\n\t\t\t\t\t\treturn {\n\t\t\t\t\t\t\ttype: \"text\",\n\t\t\t\t\t\t\ttext: sanitizeSurrogates(item.text),\n\t\t\t\t\t\t};\n\t\t\t\t\t}\n\t\t\t\t\tif (item.type === \"document\") {\n\t\t\t\t\t\t// `BetaBase64PDFSource`. PDFs ride Claude's vision path, so no beta header.\n\t\t\t\t\t\t// The media type is a fixed literal in that SDK type, so it is hardcoded here\n\t\t\t\t\t\t// and the block's own field is verified rather than read.\n\t\t\t\t\t\t// https://platform.claude.com/docs/en/build-with-claude/pdf-support\n\t\t\t\t\t\tassertSupportedDocumentMimeType(item);\n\t\t\t\t\t\treturn {\n\t\t\t\t\t\t\ttype: \"document\",\n\t\t\t\t\t\t\tsource: {\n\t\t\t\t\t\t\t\ttype: \"base64\",\n\t\t\t\t\t\t\t\tmedia_type: \"application/pdf\",\n\t\t\t\t\t\t\t\tdata: item.data,\n\t\t\t\t\t\t\t},\n\t\t\t\t\t\t\t...(item.name ? { title: item.name } : {}),\n\t\t\t\t\t\t};\n\t\t\t\t\t}\n\t\t\t\t\treturn {\n\t\t\t\t\t\ttype: \"image\",\n\t\t\t\t\t\tsource: {\n\t\t\t\t\t\t\ttype: \"base64\",\n\t\t\t\t\t\t\tmedia_type: item.mimeType as \"image/jpeg\" | \"image/png\" | \"image/gif\" | \"image/webp\",\n\t\t\t\t\t\t\tdata: item.data,\n\t\t\t\t\t\t},\n\t\t\t\t\t};\n\t\t\t\t});\n\t\t\t\tconst filteredBlocks = blocks.filter((b) => {\n\t\t\t\t\tif (b.type === \"text\") {\n\t\t\t\t\t\treturn b.text.trim().length > 0;\n\t\t\t\t\t}\n\t\t\t\t\treturn true;\n\t\t\t\t});\n\t\t\t\tif (filteredBlocks.length === 0) continue;\n\t\t\t\tparams.push({\n\t\t\t\t\trole: \"user\",\n\t\t\t\t\tcontent: filteredBlocks,\n\t\t\t\t});\n\t\t\t}\n\t\t} else if (msg.role === \"assistant\") {\n\t\t\tconst blocks: ContentBlockParam[] = [];\n\n\t\t\tfor (const block of msg.content) {\n\t\t\t\tif (block.type === \"text\") {\n\t\t\t\t\tif (block.text.trim().length === 0) continue;\n\t\t\t\t\tblocks.push({\n\t\t\t\t\t\ttype: \"text\",\n\t\t\t\t\t\ttext: sanitizeSurrogates(block.text),\n\t\t\t\t\t});\n\t\t\t\t} else if (block.type === \"thinking\") {\n\t\t\t\t\t// Redacted thinking: pass the opaque payload back as redacted_thinking\n\t\t\t\t\tif (block.redacted) {\n\t\t\t\t\t\tblocks.push({\n\t\t\t\t\t\t\ttype: \"redacted_thinking\",\n\t\t\t\t\t\t\tdata: block.thinkingSignature!,\n\t\t\t\t\t\t});\n\t\t\t\t\t\tcontinue;\n\t\t\t\t\t}\n\t\t\t\t\tconst thinkingSignature = block.thinkingSignature;\n\t\t\t\t\tconst hasThinkingSignature = !!thinkingSignature && thinkingSignature.trim().length > 0;\n\t\t\t\t\tif (block.thinking.trim().length === 0 && !hasThinkingSignature) continue;\n\t\t\t\t\t// If thinking signature is missing/empty (e.g., from aborted stream),\n\t\t\t\t\t// convert to plain text for Anthropic. Some compatible providers emit\n\t\t\t\t\t// and accept empty signatures, so let marked models preserve the block.\n\t\t\t\t\tif (!hasThinkingSignature) {\n\t\t\t\t\t\tblocks.push(\n\t\t\t\t\t\t\tallowEmptySignature\n\t\t\t\t\t\t\t\t? {\n\t\t\t\t\t\t\t\t\t\ttype: \"thinking\",\n\t\t\t\t\t\t\t\t\t\tthinking: sanitizeSurrogates(block.thinking),\n\t\t\t\t\t\t\t\t\t\tsignature: \"\",\n\t\t\t\t\t\t\t\t\t}\n\t\t\t\t\t\t\t\t: {\n\t\t\t\t\t\t\t\t\t\ttype: \"text\",\n\t\t\t\t\t\t\t\t\t\ttext: sanitizeSurrogates(block.thinking),\n\t\t\t\t\t\t\t\t\t},\n\t\t\t\t\t\t);\n\t\t\t\t\t} else {\n\t\t\t\t\t\tblocks.push({\n\t\t\t\t\t\t\ttype: \"thinking\",\n\t\t\t\t\t\t\tthinking: sanitizeSurrogates(block.thinking),\n\t\t\t\t\t\t\tsignature: thinkingSignature,\n\t\t\t\t\t\t});\n\t\t\t\t\t}\n\t\t\t\t} else if (block.type === \"toolCall\") {\n\t\t\t\t\tblocks.push({\n\t\t\t\t\t\ttype: \"tool_use\",\n\t\t\t\t\t\tid: block.id,\n\t\t\t\t\t\tname: isOAuthToken ? toClaudeCodeName(block.name) : block.name,\n\t\t\t\t\t\tinput: block.arguments ?? {},\n\t\t\t\t\t});\n\t\t\t\t} else if (block.type === \"fallback\") {\n\t\t\t\t\t// The server-side fallback boundary must go back on the wire at the position it\n\t\t\t\t\t// arrived: \"Keep it exactly where it appeared. The API uses its position to\n\t\t\t\t\t// validate the thinking blocks around it, so a request that echoes thinking\n\t\t\t\t\t// blocks from both sides of the boundary is rejected if the block is omitted or\n\t\t\t\t\t// moved.\" `transformMessages` drops the pre-boundary thinking; without this\n\t\t\t\t\t// branch the marker would be dropped too, which is the failure that rule names.\n\t\t\t\t\t// https://platform.claude.com/docs/en/build-with-claude/refusals-and-fallback\n\t\t\t\t\t//\n\t\t\t\t\t// The SDK types lag this block, matching the cast on the stream side.\n\t\t\t\t\tblocks.push({\n\t\t\t\t\t\ttype: \"fallback\",\n\t\t\t\t\t\tfrom: { model: block.fromModel },\n\t\t\t\t\t\tto: { model: block.toModel },\n\t\t\t\t\t} as unknown as ContentBlockParam);\n\t\t\t\t}\n\t\t\t}\n\t\t\tif (blocks.length === 0) continue;\n\t\t\tconst messageIndex = params.length;\n\t\t\tparams.push({\n\t\t\t\trole: \"assistant\",\n\t\t\t\tcontent: blocks,\n\t\t\t});\n\t\t\tif (\n\t\t\t\tmanagedProvider !== undefined &&\n\t\t\t\tmsg.api === \"anthropic-messages\" &&\n\t\t\t\tmsg.provider === managedProvider &&\n\t\t\t\tisAnthropicEffort(msg.providerThinkingLevel)\n\t\t\t) {\n\t\t\t\tassistantLevels.set(messageIndex, msg.providerThinkingLevel);\n\t\t\t}\n\t\t} else if (msg.role === \"toolResult\") {\n\t\t\t// Collect all consecutive toolResult messages, needed for z.ai Anthropic endpoint.\n\t\t\tconst toolResults: ContentBlockParam[] = [];\n\t\t\tconst siblingContent: ContentBlockParam[] = [];\n\t\t\tlet j = i;\n\t\t\twhile (j < transformedMessages.length && transformedMessages[j].role === \"toolResult\") {\n\t\t\t\tconst converted = convertToolResult(\n\t\t\t\t\ttransformedMessages[j] as ToolResultMessage,\n\t\t\t\t\tisOAuthToken,\n\t\t\t\t\tdeferredToolNames,\n\t\t\t\t\tloadedToolNames,\n\t\t\t\t\tnormalizeToolName,\n\t\t\t\t);\n\t\t\t\ttoolResults.push(converted.toolResult);\n\t\t\t\tsiblingContent.push(...converted.siblingContent);\n\t\t\t\tj++;\n\t\t\t}\n\n\t\t\t// Skip the messages we've already processed.\n\t\t\ti = j - 1;\n\n\t\t\t// Displaced reference-bearing results must follow every tool_result block.\n\t\t\tparams.push({\n\t\t\t\trole: \"user\",\n\t\t\t\tcontent: [...toolResults, ...siblingContent],\n\t\t\t});\n\t\t}\n\t}\n\n\t// Add cache_control to the last user message to cache conversation history\n\tif (cacheControl && params.length > 0) {\n\t\tconst lastMessage = params[params.length - 1];\n\t\tif (lastMessage.role === \"user\") {\n\t\t\tif (Array.isArray(lastMessage.content)) {\n\t\t\t\tconst lastBlock = lastMessage.content[lastMessage.content.length - 1];\n\t\t\t\tif (\n\t\t\t\t\tlastBlock &&\n\t\t\t\t\t(lastBlock.type === \"text\" || lastBlock.type === \"image\" || lastBlock.type === \"tool_result\")\n\t\t\t\t) {\n\t\t\t\t\t(lastBlock as any).cache_control = cacheControl;\n\t\t\t\t}\n\t\t\t} else if (typeof lastMessage.content === \"string\") {\n\t\t\t\tlastMessage.content = [\n\t\t\t\t\t{\n\t\t\t\t\t\ttype: \"text\",\n\t\t\t\t\t\ttext: lastMessage.content,\n\t\t\t\t\t\tcache_control: cacheControl,\n\t\t\t\t\t},\n\t\t\t\t] as any;\n\t\t\t}\n\t\t}\n\t}\n\n\treturn { messages: params, assistantLevels };\n}\n\ninterface ConvertedAnthropicMessages {\n\tmessages: MessageParam[];\n\tassistantLevels: Map<number, AnthropicEffort>;\n}\n\nfunction isAnthropicEffort(value: string | undefined): value is AnthropicEffort {\n\treturn value === \"low\" || value === \"medium\" || value === \"high\" || value === \"xhigh\" || value === \"max\";\n}\n\ninterface EffortMessage {\n\trole: \"system\";\n\tcontent: [];\n\toutput_config: { effort: AnthropicEffort };\n}\n\nfunction insertThinkingLevelMessages(\n\tconverted: ConvertedAnthropicMessages,\n\tactiveEffort: AnthropicEffort,\n): Array<MessageParam | EffortMessage> {\n\tconst messages: Array<MessageParam | EffortMessage> = [];\n\tfor (let index = 0; index < converted.messages.length; index++) {\n\t\tconst historicalEffort = converted.assistantLevels.get(index);\n\t\tif (historicalEffort !== undefined) {\n\t\t\tmessages.push({ role: \"system\", content: [], output_config: { effort: historicalEffort } });\n\t\t}\n\t\tmessages.push(converted.messages[index]);\n\t}\n\tmessages.push({ role: \"system\", content: [], output_config: { effort: activeEffort } });\n\treturn messages;\n}\n\nfunction shouldUseFineGrainedToolStreamingBeta(model: Model<\"anthropic-messages\">, context: Context): boolean {\n\treturn !!context.tools?.length && !getAnthropicCompat(model).supportsEagerToolInputStreaming;\n}\n\nfunction convertTools(\n\ttools: Tool[],\n\tisOAuthToken: boolean,\n\tsupportsEagerToolInputStreaming: boolean,\n\tsupportsStrictTools: boolean,\n\tcacheControl?: CacheControlEphemeral,\n\tdeferLoading = false,\n): Anthropic.Messages.Tool[] {\n\tif (!tools) return [];\n\n\treturn tools.map((tool, index) => {\n\t\tconst strict = resolveJsonSchemaStrictSampling(tool, supportsStrictTools);\n\t\tconst parameters = getJsonSchemaToolParameters(tool, strict);\n\t\tconst schema = parameters as { properties?: unknown; required?: string[] };\n\t\tconst legacyInputSchema = {\n\t\t\ttype: \"object\" as const,\n\t\t\tproperties: schema.properties ?? {},\n\t\t\trequired: schema.required ?? [],\n\t\t};\n\t\tconst inputSchema =\n\t\t\tstrict === true\n\t\t\t\t? {\n\t\t\t\t\t\t...(parameters as Record<string, unknown>),\n\t\t\t\t\t\t...legacyInputSchema,\n\t\t\t\t\t}\n\t\t\t\t: legacyInputSchema;\n\n\t\treturn {\n\t\t\tname: isOAuthToken ? toClaudeCodeName(tool.name) : tool.name,\n\t\t\tdescription: tool.description,\n\t\t\t...(supportsEagerToolInputStreaming ? { eager_input_streaming: true } : {}),\n\t\t\t...(strict === true ? { strict: true } : {}),\n\t\t\tinput_schema: inputSchema,\n\t\t\t...(deferLoading ? { defer_loading: true } : {}),\n\t\t\t...(cacheControl && index === tools.length - 1 ? { cache_control: cacheControl } : {}),\n\t\t};\n\t});\n}\n\nfunction mapStopReason(\n\treason: Anthropic.Messages.StopReason | string,\n\tstopDetails?: RefusalStopDetails | null,\n): { stopReason: StopReason; errorMessage?: string } {\n\tswitch (reason) {\n\t\tcase \"end_turn\":\n\t\t\treturn { stopReason: \"stop\" };\n\t\tcase \"max_tokens\":\n\t\t\treturn { stopReason: \"length\" };\n\t\tcase \"tool_use\":\n\t\t\treturn { stopReason: \"toolUse\" };\n\t\tcase \"refusal\":\n\t\t\treturn {\n\t\t\t\tstopReason: \"error\",\n\t\t\t\terrorMessage: stopDetails?.explanation || `The model refused to complete the request`,\n\t\t\t};\n\t\tcase \"pause_turn\": // Stop is good enough -> resubmit\n\t\t\treturn { stopReason: \"stop\" };\n\t\tcase \"stop_sequence\":\n\t\t\treturn { stopReason: \"stop\" }; // We don't supply stop sequences, so this should never happen\n\t\tcase \"sensitive\": // Content flagged by safety filters (not yet in SDK types)\n\t\t\treturn { stopReason: \"error\", errorMessage: \"Provider stopped with: sensitive\" };\n\t\tdefault:\n\t\t\t// Handle unknown stop reasons gracefully (API may add new values)\n\t\t\tthrow new Error(`Unhandled stop reason: ${reason}`);\n\t}\n}\n"]}
+{"version":3,"file":"anthropic-messages.d.ts","sourceRoot":"","sources":["../../src/api/anthropic-messages.ts"],"names":[],"mappings":"AAAA,OAAO,SAAS,MAAM,mBAAmB,CAAC;AAU1C,OAAO,KAAK,EAUX,mBAAmB,EAEnB,cAAc,EACd,aAAa,EAOb,MAAM,aAAa,CAAC;AA0JrB,MAAM,MAAM,eAAe,GAAG,KAAK,GAAG,QAAQ,GAAG,MAAM,GAAG,OAAO,GAAG,KAAK,CAAC;AAE1E,MAAM,MAAM,wBAAwB,GAAG,YAAY,GAAG,SAAS,CAAC;AAoLhE,MAAM,WAAW,gBAAiB,SAAQ,aAAa;IACtD;;;;;;OAMG;IACH,eAAe,CAAC,EAAE,OAAO,CAAC;IAC1B;;;;OAIG;IACH,oBAAoB,CAAC,EAAE,MAAM,CAAC;IAC9B;;;;;;;;;;;OAWG;IACH,MAAM,CAAC,EAAE,eAAe,CAAC;IACzB;;;;;;;;;;;OAWG;IACH,eAAe,CAAC,EAAE,wBAAwB,CAAC;IAC3C;;;;;OAKG;IACH,mBAAmB,CAAC,EAAE,OAAO,CAAC;IAC9B;;;;OAIG;IACH,UAAU,CAAC,EAAE,MAAM,GAAG,KAAK,GAAG,MAAM,GAAG;QAAE,IAAI,EAAE,MAAM,CAAC;QAAC,IAAI,EAAE,MAAM,CAAA;KAAE,CAAC;IACtE;;;;OAIG;IACH,MAAM,CAAC,EAAE,SAAS,CAAC;CACnB;AA6OD,eAAO,MAAM,MAAM,EAAE,cAAc,CAAC,oBAAoB,EAAE,gBAAgB,CAsWzE,CAAC;AA4BF,eAAO,MAAM,YAAY,EAAE,cAAc,CAAC,oBAAoB,EAAE,mBAAmB,CA8ClF,CAAC","sourcesContent":["import Anthropic from \"@anthropic-ai/sdk\";\nimport type {\n\tCacheControlEphemeral,\n\tContentBlockParam,\n\tMessageCreateParamsStreaming,\n\tMessageParam,\n\tRawMessageStreamEvent,\n\tRefusalStopDetails,\n} from \"@anthropic-ai/sdk/resources/messages.js\";\nimport { calculateCost } from \"../models.ts\";\nimport type {\n\tApi,\n\tAssistantMessage,\n\tCacheRetention,\n\tContext,\n\tImageContent,\n\tMessage,\n\tModel,\n\tProviderEnv,\n\tProviderHeaders,\n\tSimpleStreamOptions,\n\tStopReason,\n\tStreamFunction,\n\tStreamOptions,\n\tTextContent,\n\tThinkingContent,\n\tTool,\n\tToolCall,\n\tToolResultMessage,\n\tUsage,\n} from \"../types.ts\";\nimport { splitDeferredTools } from \"../utils/deferred-tools.ts\";\nimport { appendAssistantMessageDiagnostic } from \"../utils/diagnostics.ts\";\nimport { assertSupportedDocumentMimeType } from \"../utils/document-input.ts\";\nimport { AssistantMessageEventStream } from \"../utils/event-stream.ts\";\nimport { headersToRecord } from \"../utils/headers.ts\";\nimport { parseJsonWithRepair, parseStreamingJson } from \"../utils/json-parse.ts\";\nimport { getPiUserAgent } from \"../utils/pi-user-agent.ts\";\nimport { getProviderEnvValue } from \"../utils/provider-env.ts\";\nimport { retryProviderRequest } from \"../utils/provider-retry.ts\";\nimport { sanitizeSurrogates } from \"../utils/sanitize-unicode.ts\";\nimport { createStreamDeadline, withStreamDeadline } from \"../utils/stream-deadline.ts\";\n\nimport { getJsonSchemaToolParameters, resolveJsonSchemaStrictSampling } from \"./constrained-sampling.ts\";\nimport {\n\tbuildCopilotDynamicHeaders,\n\thasCopilotVisionInput,\n\tpreserveCopilotIntegrationHeader,\n} from \"./github-copilot-headers.ts\";\nimport { adjustMaxTokensForThinking, buildBaseOptions, clampMaxTokensToContext } from \"./simple-options.ts\";\nimport { transformMessages } from \"./transform-messages.ts\";\n\n/**\n * Resolve cache retention preference.\n * Defaults to \"short\" and uses PI_CACHE_RETENTION for backward compatibility.\n */\nfunction resolveCacheRetention(cacheRetention?: CacheRetention, env?: ProviderEnv): CacheRetention {\n\tif (cacheRetention) {\n\t\treturn cacheRetention;\n\t}\n\tif (getProviderEnvValue(\"PI_CACHE_RETENTION\", env) === \"long\") {\n\t\treturn \"long\";\n\t}\n\treturn \"short\";\n}\n\nfunction getCacheControl(\n\tmodel: Model<\"anthropic-messages\">,\n\tcacheRetention?: CacheRetention,\n\tenv?: ProviderEnv,\n): { retention: CacheRetention; cacheControl?: CacheControlEphemeral } {\n\tconst retention = resolveCacheRetention(cacheRetention, env);\n\tif (retention === \"none\") {\n\t\treturn { retention };\n\t}\n\tconst ttl = retention === \"long\" && getAnthropicCompat(model).supportsLongCacheRetention ? \"1h\" : undefined;\n\treturn {\n\t\tretention,\n\t\tcacheControl: { type: \"ephemeral\", ...(ttl && { ttl }) },\n\t};\n}\n\n// Stealth mode: Mimic Claude Code's tool naming exactly\n//\n// Anthropic gates newer models on the `claude-cli/<version>` user agent alone and rejects an\n// older one with `claude_code_version_too_old`. `claude-fable-5-1` requires >= 2.1.251, bisected\n// against the live API: 2.1.250 -> 400, 2.1.251 -> 200. This is pinned to that exact published\n// minimum rather than the newest release, and it is a strict superset of the previous 2.1.75 --\n// every model this provider ships answers 200 at 2.1.251. Raise it only when a model rejects\n// this value; a caller can override it for one client through the `headers` option, whose\n// lowercase `user-agent` key is merged last.\nconst claudeCodeVersion = \"2.1.251\";\n\n// Claude Code 2.x tool names (canonical casing)\n// Source: https://cchistory.mariozechner.at/data/prompts-2.1.11.md\n// To update: https://github.com/badlogic/cchistory\nconst claudeCodeTools = [\n\t\"Read\",\n\t\"Write\",\n\t\"Edit\",\n\t\"Bash\",\n\t\"Grep\",\n\t\"Glob\",\n\t\"AskUserQuestion\",\n\t\"EnterPlanMode\",\n\t\"ExitPlanMode\",\n\t\"KillShell\",\n\t\"NotebookEdit\",\n\t\"Skill\",\n\t\"Task\",\n\t\"TaskOutput\",\n\t\"TodoWrite\",\n\t\"WebFetch\",\n\t\"WebSearch\",\n];\n\nconst ccToolLookup = new Map(claudeCodeTools.map((t) => [t.toLowerCase(), t]));\n\n// Convert tool name to CC canonical casing if it matches (case-insensitive)\nconst toClaudeCodeName = (name: string) => ccToolLookup.get(name.toLowerCase()) ?? name;\nconst fromClaudeCodeName = (name: string, tools?: Tool[]) => {\n\tif (tools && tools.length > 0) {\n\t\tconst lowerName = name.toLowerCase();\n\t\tconst matchedTool = tools.find((tool) => tool.name.toLowerCase() === lowerName);\n\t\tif (matchedTool) return matchedTool.name;\n\t}\n\treturn name;\n};\n\n/**\n * Convert tool-result content blocks to Anthropic API format.\n *\n * Tool results carry text and images only. `DocumentContent` appears exclusively in user\n * messages, which `convertMessages` serializes on its own path.\n */\nfunction convertContentBlocks(content: (TextContent | ImageContent)[]):\n\t| string\n\t| Array<\n\t\t\t| { type: \"text\"; text: string }\n\t\t\t| {\n\t\t\t\t\ttype: \"image\";\n\t\t\t\t\tsource: {\n\t\t\t\t\t\ttype: \"base64\";\n\t\t\t\t\t\tmedia_type: \"image/jpeg\" | \"image/png\" | \"image/gif\" | \"image/webp\";\n\t\t\t\t\t\tdata: string;\n\t\t\t\t\t};\n\t\t\t  }\n\t  > {\n\t// If only text blocks, return as concatenated string for simplicity\n\tconst hasImages = content.some((c) => c.type === \"image\");\n\tif (!hasImages) {\n\t\treturn sanitizeSurrogates(content.map((c) => (c as TextContent).text).join(\"\\n\"));\n\t}\n\n\t// If we have images, convert to content block array\n\tconst blocks = content.map((block) => {\n\t\tif (block.type === \"text\") {\n\t\t\treturn {\n\t\t\t\ttype: \"text\" as const,\n\t\t\t\ttext: sanitizeSurrogates(block.text),\n\t\t\t};\n\t\t}\n\t\treturn {\n\t\t\ttype: \"image\" as const,\n\t\t\tsource: {\n\t\t\t\ttype: \"base64\" as const,\n\t\t\t\tmedia_type: block.mimeType as \"image/jpeg\" | \"image/png\" | \"image/gif\" | \"image/webp\",\n\t\t\t\tdata: block.data,\n\t\t\t},\n\t\t};\n\t});\n\n\t// If only images (no text), add placeholder text block\n\tconst hasText = blocks.some((b) => b.type === \"text\");\n\tif (!hasText) {\n\t\tblocks.unshift({\n\t\t\ttype: \"text\" as const,\n\t\t\ttext: \"(see attached image)\",\n\t\t});\n\t}\n\n\treturn blocks;\n}\n\nexport type AnthropicEffort = \"low\" | \"medium\" | \"high\" | \"xhigh\" | \"max\";\n\nexport type AnthropicThinkingDisplay = \"summarized\" | \"omitted\";\n\ntype MessageCreateParamsStreamingWithFallbacks = MessageCreateParamsStreaming & {\n\tfallbacks?: readonly { model: string }[];\n};\n\nconst FINE_GRAINED_TOOL_STREAMING_BETA = \"fine-grained-tool-streaming-2025-05-14\";\nconst INTERLEAVED_THINKING_BETA = \"interleaved-thinking-2025-05-14\";\nconst SERVER_SIDE_FALLBACK_BETA = \"server-side-fallback-2026-07-01\";\nconst THINKING_BINDING_CONTROLS_BETA = \"thinking-binding-controls-2026-08-01\";\nconst MID_CONVERSATION_OUTPUT_CONFIG_BETA = \"mid-conversation-output-config-2026-07-01\";\n\nfunction shouldUseServerSideFallbackBeta(model: Model<\"anthropic-messages\">): boolean {\n\treturn (model.compat?.allowedFallbackModels?.length ?? 0) > 0;\n}\n\n/**\n * Claude Fable 5.1 binds each thinking block to the conversation prefix that produced it and,\n * for Anthropic accounts created on or after 2026-08-31, rejects a replay behind a changed\n * `system` prompt, `tools` array, or earlier message with a 400 `invalid_request_error`.\n * Atomic rebuilds those inputs between turns (dynamic system prompt, tool availability changes,\n * model switches), so it opts into `prefix_mismatch_behavior: \"drop_block\"`: the API discards\n * the affected thinking blocks and answers the turn instead of failing the session.\n *\n * This covers live prefix mismatches *between* compaction boundaries. It is not what handles\n * compaction itself: Atomic's client-side `preserve_recent` tail compaction serializes the\n * protected tail into a single boundary message, so no signed thinking block survives a\n * boundary to be replayed behind it. That is Anthropic's documented keep-tail remedy applied\n * structurally rather than a case `drop_block` has to catch.\n * https://platform.claude.com/docs/en/build-with-claude/preserved-thinking\n */\nfunction shouldUseThinkingBindingControlsBeta(model: Model<\"anthropic-messages\">): boolean {\n\treturn model.compat?.enforcesPreservedThinkingBinding === true;\n}\n\n/**\n * Per-turn effort and preserved-thinking binding are independent capabilities.\n * `supportsMidConvoEffort` adds provider effort markers and requires binding controls;\n * Atomic's older `enforcesPreservedThinkingBinding` remains the broader opt-in for\n * transports that need drop recovery without supporting effort-only system messages.\n */\nfunction supportsMidConvoEffort(model: Model<\"anthropic-messages\">): boolean {\n\treturn model.compat?.supportsMidConvoEffort === true;\n}\n\n/** One entry of the `input_transformations` array the block-binding controls beta adds. */\ninterface AnthropicInputTransformation {\n\ttype?: string;\n\tpath?: string;\n\treason?: string;\n}\n\nfunction getInputTransformations(message: unknown): AnthropicInputTransformation[] | undefined {\n\tconst transformations = (message as { input_transformations?: AnthropicInputTransformation[] })\n\t\t?.input_transformations;\n\treturn Array.isArray(transformations) ? transformations : undefined;\n}\n\n/** Record thinking blocks the API dropped from this request without exposing block contents. */\nfunction recordInputTransformations(output: AssistantMessage, transformations: AnthropicInputTransformation[]): void {\n\tif (transformations.length === 0) return;\n\tappendAssistantMessageDiagnostic(output, {\n\t\ttype: \"anthropic_input_transformations\",\n\t\ttimestamp: Date.now(),\n\t\tdetails: {\n\t\t\tdroppedBlockCount: transformations.length,\n\t\t\treasons: [...new Set(transformations.map((entry) => entry.reason).filter((r): r is string => !!r))],\n\t\t\tpaths: transformations.map((entry) => entry.path).filter((p): p is string => !!p),\n\t\t},\n\t});\n}\n\n/** One entry of the per-attempt `usage.iterations` array a server-side fallback response carries. */\ninterface AnthropicUsageIteration {\n\ttype?: string;\n\tmodel?: string;\n\tinput_tokens?: number;\n\toutput_tokens?: number;\n\tcache_read_input_tokens?: number;\n\t/** The aggregate cache-write count. `cache_creation` splits the same tokens by TTL. */\n\tcache_creation_input_tokens?: number;\n\t/** \"Breakdown of cached tokens by TTL\", per `BetaMessageIterationUsage`. */\n\tcache_creation?: { ephemeral_1h_input_tokens?: number; ephemeral_5m_input_tokens?: number } | null;\n}\n\n/**\n * Bill the attempts that ran *before* the one which produced the returned message.\n *\n * \"Every attempt that produced output, including one that declined partway through its response,\n * is billed separately at the rates of the model that ran it. The `usage.iterations` array is the\n * per-attempt record of what you're billed. The top-level `usage` counts describe only the attempt\n * that produced the returned message. Tokens from different models are never summed into one\n * field.\" https://platform.claude.com/docs/en/build-with-claude/refusals-and-fallback\n *\n * Two rules follow, and both matter for not double-counting:\n *\n * - Only `type: \"message\"` entries are added. The `fallback_message` entry *is* the serving\n *   attempt, and its tokens are already in the top-level usage that `calculateCost` just priced.\n * - Only entries with output are added: \"An attempt that declined before producing any output is\n *   not billed: its tokens are reported on its `usage.iterations` entry but not charged.\"\n *\n * Token counts are deliberately left alone — the docs forbid summing across models and `Usage` has\n * no per-attempt shape — so this contributes to `usage.cost` only.\n */\nfunction addEarlierAttemptCosts(\n\toutput: AssistantMessage,\n\tmodel: Model<\"anthropic-messages\">,\n\tservingModel: Model<\"anthropic-messages\">,\n\tevent: unknown,\n): void {\n\tconst iterations = (event as { usage?: { iterations?: AnthropicUsageIteration[] } })?.usage?.iterations;\n\tif (!Array.isArray(iterations) || iterations.length === 0) return;\n\n\tfor (const iteration of iterations) {\n\t\tif (iteration.type !== \"message\") continue;\n\t\tif (!iteration.output_tokens) continue;\n\n\t\t// Price at the rates of the model that ran the attempt, which is not the serving model.\n\t\tconst attemptCost =\n\t\t\titeration.model === model.id\n\t\t\t\t? model.cost\n\t\t\t\t: iteration.model === servingModel.id\n\t\t\t\t\t? servingModel.cost\n\t\t\t\t\t: model.compat?.allowedFallbackModels?.find(\n\t\t\t\t\t\t\t(fallback) => fallback.provider === model.provider && fallback.model === iteration.model,\n\t\t\t\t\t\t)?.cost;\n\t\tif (!attemptCost) continue;\n\n\t\tconst attemptUsage: Usage = {\n\t\t\tinput: iteration.input_tokens ?? 0,\n\t\t\toutput: iteration.output_tokens,\n\t\t\tcacheRead: iteration.cache_read_input_tokens ?? 0,\n\t\t\t// `cacheWrite` stays the aggregate: `calculateCost` derives the 5-minute share by\n\t\t\t// subtracting `cacheWrite1h` from it, so setting this to the 5-minute count instead\n\t\t\t// would under-charge and adding the two together would double-charge. Without the 1h\n\t\t\t// split, an hour-long write bills at the 5-minute rate rather than 2x base input.\n\t\t\tcacheWrite: iteration.cache_creation_input_tokens ?? 0,\n\t\t\tcacheWrite1h: iteration.cache_creation?.ephemeral_1h_input_tokens ?? 0,\n\t\t\ttotalTokens: 0,\n\t\t\tcost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0, total: 0 },\n\t\t};\n\t\tcalculateCost({ ...model, id: iteration.model ?? model.id, cost: attemptCost }, attemptUsage);\n\n\t\toutput.usage.cost.input += attemptUsage.cost.input;\n\t\toutput.usage.cost.output += attemptUsage.cost.output;\n\t\toutput.usage.cost.cacheRead += attemptUsage.cost.cacheRead;\n\t\toutput.usage.cost.cacheWrite += attemptUsage.cost.cacheWrite;\n\t\toutput.usage.cost.total += attemptUsage.cost.total;\n\t}\n}\n\nfunction getAnthropicCompat(model: Model<\"anthropic-messages\">) {\n\tconst isOpenRouter = model.provider === \"openrouter\" || model.baseUrl.includes(\"openrouter.ai\");\n\treturn {\n\t\tsupportsEagerToolInputStreaming: model.compat?.supportsEagerToolInputStreaming ?? true,\n\t\tsupportsLongCacheRetention: model.compat?.supportsLongCacheRetention ?? true,\n\t\tsendSessionAffinityHeaders: model.compat?.sendSessionAffinityHeaders ?? isOpenRouter,\n\t\tsessionAffinityFormat: model.compat?.sessionAffinityFormat ?? (isOpenRouter ? \"openrouter\" : undefined),\n\t\tsupportsCacheControlOnTools: model.compat?.supportsCacheControlOnTools ?? true,\n\t\tsupportsTemperature: model.compat?.supportsTemperature ?? true,\n\t\tallowEmptySignature: model.compat?.allowEmptySignature ?? false,\n\t\tsupportsStrictTools: model.compat?.supportsStrictTools ?? false,\n\t\tsupportsToolReferences: model.compat?.supportsToolReferences ?? defaultSupportsToolReferences(model),\n\t};\n}\n\n/**\n * Default for `supportsToolReferences`: first-party Anthropic models except\n * Haiku (rejects client-side tool_reference blocks) and models that predate\n * tool search (Claude 3.x, Opus/Sonnet 4.0, Opus 4.1).\n */\nfunction defaultSupportsToolReferences(model: Model<\"anthropic-messages\">): boolean {\n\tif (model.provider !== \"anthropic\" || model.id.includes(\"haiku\")) return false;\n\tconst version = model.id.match(/^claude-(?:opus|sonnet|fable)-(\\d+)(?:-(\\d+))?(?:-|$)/);\n\tif (!version) return false;\n\tconst major = Number(version[1]);\n\tconst minor = version[2] && version[2].length < 8 ? Number(version[2]) : 0;\n\treturn major > 4 || (major === 4 && minor >= 5);\n}\n\nexport interface AnthropicOptions extends StreamOptions {\n\t/**\n\t * Enable extended thinking.\n\t * For adaptive thinking models: the model decides when/how much to think.\n\t * For older models: uses budget-based thinking with thinkingBudgetTokens.\n\t * Default: undefined (thinking is omitted unless `streamSimple()` maps\n\t * a simple reasoning level to this option, or callers set it explicitly).\n\t */\n\tthinkingEnabled?: boolean;\n\t/**\n\t * Token budget for extended thinking (older models only).\n\t * Ignored for adaptive thinking models.\n\t * Default: 1024 when `thinkingEnabled` is true and no budget is provided.\n\t */\n\tthinkingBudgetTokens?: number;\n\t/**\n\t * Effort level for adaptive thinking models.\n\t * Controls how much thinking Claude allocates:\n\t * - \"max\": Always thinks with no constraints (Opus 4.6 only)\n\t * - \"xhigh\": Highest reasoning level (Opus 4.7+, Fable 5)\n\t * - \"high\": Always thinks, deep reasoning\n\t * - \"medium\": Moderate thinking, may skip for simple queries\n\t * - \"low\": Minimal thinking, skips for simple tasks\n\t * Ignored for older models.\n\t * Default: omitted unless `streamSimple()` maps a simple reasoning\n\t * level to this option.\n\t */\n\teffort?: AnthropicEffort;\n\t/**\n\t * Controls how thinking content is returned in API responses.\n\t * - \"summarized\": Thinking blocks contain summarized thinking text.\n\t * - \"omitted\": Thinking blocks return an empty thinking field; the encrypted\n\t *   signature still travels back for multi-turn continuity. Use for faster\n\t *   time-to-first-text-token when your UI does not surface thinking.\n\t *\n\t * Note: Anthropic's API default for Claude Opus 4.7 and Claude Mythos Preview\n\t * is \"omitted\". We default to \"summarized\" here to keep behavior consistent\n\t * with older Claude 4 models. Set this explicitly to \"omitted\" to opt in.\n\t * Default: \"summarized\" when thinking is enabled.\n\t */\n\tthinkingDisplay?: AnthropicThinkingDisplay;\n\t/**\n\t * Whether to request the interleaved thinking beta header for non-adaptive\n\t * thinking models. Adaptive thinking models have interleaved thinking built in,\n\t * so the header is skipped for them regardless of this setting.\n\t * Default: true.\n\t */\n\tinterleavedThinking?: boolean;\n\t/**\n\t * Anthropic tool choice behavior. String values map to Anthropic's built-in\n\t * choices; `{ type: \"tool\", name }` forces a specific tool.\n\t * Default: omitted (Anthropic default behavior, currently equivalent to auto).\n\t */\n\ttoolChoice?: \"auto\" | \"any\" | \"none\" | { type: \"tool\"; name: string };\n\t/**\n\t * Pre-built Anthropic client instance. When provided, skips internal client\n\t * construction entirely. Use this to inject alternative SDK clients such as\n\t * `AnthropicVertex` that shares the same messaging API.\n\t */\n\tclient?: Anthropic;\n}\n\nfunction mergeHeaders(...headerSources: (ProviderHeaders | undefined)[]): ProviderHeaders {\n\tconst merged: ProviderHeaders = {};\n\tfor (const headers of headerSources) {\n\t\tif (headers) {\n\t\t\tObject.assign(merged, headers);\n\t\t}\n\t}\n\treturn merged;\n}\n\nfunction mergeClientHeaders(...headerSources: (ProviderHeaders | undefined)[]): ProviderHeaders {\n\treturn mergeHeaders({ \"User-Agent\": getPiUserAgent() }, ...headerSources);\n}\n\nfunction hasHeader(headers: ProviderHeaders | undefined, name: string): boolean {\n\tif (!headers) return false;\n\tconst expected = name.toLowerCase();\n\tfor (const [key, value] of Object.entries(headers)) {\n\t\tif (key.toLowerCase() === expected && value !== null && value.trim().length > 0) return true;\n\t}\n\treturn false;\n}\n\nfunction assertRequestAuth(provider: string, apiKey: string | undefined, headers: ProviderHeaders | undefined): void {\n\tif (apiKey) return;\n\tif (\n\t\thasHeader(headers, \"authorization\") ||\n\t\thasHeader(headers, \"x-api-key\") ||\n\t\thasHeader(headers, \"cf-aig-authorization\")\n\t) {\n\t\treturn;\n\t}\n\tthrow new Error(`No API key for provider: ${provider}`);\n}\n\ninterface ServerSentEvent {\n\tevent: string | null;\n\tdata: string;\n\traw: string[];\n}\n\ninterface SseDecoderState {\n\tevent: string | null;\n\tdata: string[];\n\traw: string[];\n}\n\nconst ANTHROPIC_MESSAGE_EVENTS: ReadonlySet<string> = new Set([\n\t\"message_start\",\n\t\"message_delta\",\n\t\"message_stop\",\n\t\"content_block_start\",\n\t\"content_block_delta\",\n\t\"content_block_stop\",\n]);\n\nfunction flushSseEvent(state: SseDecoderState): ServerSentEvent | null {\n\tif (!state.event && state.data.length === 0) {\n\t\treturn null;\n\t}\n\n\tconst event: ServerSentEvent = {\n\t\tevent: state.event,\n\t\tdata: state.data.join(\"\\n\"),\n\t\traw: [...state.raw],\n\t};\n\tstate.event = null;\n\tstate.data = [];\n\tstate.raw = [];\n\treturn event;\n}\n\nfunction decodeSseLine(line: string, state: SseDecoderState): ServerSentEvent | null {\n\tif (line === \"\") {\n\t\treturn flushSseEvent(state);\n\t}\n\n\tstate.raw.push(line);\n\tif (line.startsWith(\":\")) {\n\t\treturn null;\n\t}\n\n\tconst delimiterIndex = line.indexOf(\":\");\n\tconst fieldName = delimiterIndex === -1 ? line : line.slice(0, delimiterIndex);\n\tlet value = delimiterIndex === -1 ? \"\" : line.slice(delimiterIndex + 1);\n\tif (value.startsWith(\" \")) {\n\t\tvalue = value.slice(1);\n\t}\n\n\tif (fieldName === \"event\") {\n\t\tstate.event = value;\n\t} else if (fieldName === \"data\") {\n\t\tstate.data.push(value);\n\t}\n\n\treturn null;\n}\n\nfunction nextLineBreakIndex(text: string): number {\n\tconst carriageReturnIndex = text.indexOf(\"\\r\");\n\tconst newlineIndex = text.indexOf(\"\\n\");\n\tif (carriageReturnIndex === -1) {\n\t\treturn newlineIndex;\n\t}\n\tif (newlineIndex === -1) {\n\t\treturn carriageReturnIndex;\n\t}\n\treturn Math.min(carriageReturnIndex, newlineIndex);\n}\n\nfunction consumeLine(text: string): { line: string; rest: string } | null {\n\tconst lineBreakIndex = nextLineBreakIndex(text);\n\tif (lineBreakIndex === -1) {\n\t\treturn null;\n\t}\n\n\tlet nextIndex = lineBreakIndex + 1;\n\tif (text[lineBreakIndex] === \"\\r\" && text[nextIndex] === \"\\n\") {\n\t\tnextIndex += 1;\n\t}\n\n\treturn {\n\t\tline: text.slice(0, lineBreakIndex),\n\t\trest: text.slice(nextIndex),\n\t};\n}\n\nasync function* iterateSseMessages(\n\tbody: ReadableStream<Uint8Array>,\n\tsignal?: AbortSignal,\n): AsyncGenerator<ServerSentEvent> {\n\tconst reader = body.getReader();\n\tconst decoder = new TextDecoder();\n\tconst state: SseDecoderState = { event: null, data: [], raw: [] };\n\tlet buffer = \"\";\n\tconst onAbort = () => {\n\t\tvoid reader.cancel().catch(() => {});\n\t};\n\tsignal?.addEventListener(\"abort\", onAbort, { once: true });\n\n\ttry {\n\t\twhile (true) {\n\t\t\tif (signal?.aborted) {\n\t\t\t\tthrow new Error(\"Request was aborted\");\n\t\t\t}\n\n\t\t\tconst { value, done } = await reader.read();\n\t\t\tif (done) {\n\t\t\t\tbreak;\n\t\t\t}\n\n\t\t\tbuffer += decoder.decode(value, { stream: true });\n\t\t\tlet consumed = consumeLine(buffer);\n\t\t\twhile (consumed) {\n\t\t\t\tbuffer = consumed.rest;\n\t\t\t\tconst event = decodeSseLine(consumed.line, state);\n\t\t\t\tif (event) {\n\t\t\t\t\tyield event;\n\t\t\t\t}\n\t\t\t\tconsumed = consumeLine(buffer);\n\t\t\t}\n\t\t}\n\n\t\tbuffer += decoder.decode();\n\t\tlet consumed = consumeLine(buffer);\n\t\twhile (consumed) {\n\t\t\tbuffer = consumed.rest;\n\t\t\tconst event = decodeSseLine(consumed.line, state);\n\t\t\tif (event) {\n\t\t\t\tyield event;\n\t\t\t}\n\t\t\tconsumed = consumeLine(buffer);\n\t\t}\n\n\t\tif (buffer.length > 0) {\n\t\t\tconst event = decodeSseLine(buffer, state);\n\t\t\tif (event) {\n\t\t\t\tyield event;\n\t\t\t}\n\t\t}\n\n\t\tconst trailingEvent = flushSseEvent(state);\n\t\tif (trailingEvent) {\n\t\t\tyield trailingEvent;\n\t\t}\n\t} finally {\n\t\tsignal?.removeEventListener(\"abort\", onAbort);\n\t\ttry {\n\t\t\tawait reader.cancel();\n\t\t} catch {}\n\t\treader.releaseLock();\n\t}\n}\n\nasync function* iterateAnthropicEvents(\n\tresponse: Response,\n\tsignal?: AbortSignal,\n): AsyncGenerator<RawMessageStreamEvent> {\n\tif (!response.body) {\n\t\tthrow new Error(\"Attempted to iterate over an Anthropic response with no body\");\n\t}\n\n\tlet sawMessageStart = false;\n\tlet sawMessageEnd = false;\n\n\tfor await (const sse of iterateSseMessages(response.body, signal)) {\n\t\tif (sse.event === \"error\") {\n\t\t\tthrow new Error(sse.data);\n\t\t}\n\n\t\tif (!ANTHROPIC_MESSAGE_EVENTS.has(sse.event ?? \"\")) {\n\t\t\tcontinue;\n\t\t}\n\n\t\ttry {\n\t\t\tconst event = parseJsonWithRepair<RawMessageStreamEvent>(sse.data);\n\t\t\tif (event.type === \"message_start\") {\n\t\t\t\tsawMessageStart = true;\n\t\t\t} else if (event.type === \"message_stop\") {\n\t\t\t\tsawMessageEnd = true;\n\t\t\t}\n\t\t\tyield event;\n\t\t} catch (error) {\n\t\t\tconst message = error instanceof Error ? error.message : String(error);\n\t\t\tthrow new Error(\n\t\t\t\t`Could not parse Anthropic SSE event ${sse.event}: ${message}; data=${sse.data}; raw=${sse.raw.join(\"\\\\n\")}`,\n\t\t\t);\n\t\t}\n\t}\n\n\tif (sawMessageStart && !sawMessageEnd) {\n\t\tthrow new Error(\"Anthropic stream ended before message_stop\");\n\t}\n}\n\nexport const stream: StreamFunction<\"anthropic-messages\", AnthropicOptions> = (\n\tmodel: Model<\"anthropic-messages\">,\n\tcontext: Context,\n\toptions?: AnthropicOptions,\n): AssistantMessageEventStream => {\n\tconst stream = new AssistantMessageEventStream();\n\n\t(async () => {\n\t\tconst providerThinkingLevel = supportsMidConvoEffort(model) ? (options?.effort ?? \"high\") : undefined;\n\t\tconst output: AssistantMessage = {\n\t\t\trole: \"assistant\",\n\t\t\tcontent: [],\n\t\t\tapi: model.api as Api,\n\t\t\tprovider: model.provider,\n\t\t\tmodel: model.id,\n\t\t\t...(providerThinkingLevel === undefined ? {} : { providerThinkingLevel }),\n\t\t\tusage: {\n\t\t\t\tinput: 0,\n\t\t\t\toutput: 0,\n\t\t\t\tcacheRead: 0,\n\t\t\t\tcacheWrite: 0,\n\t\t\t\ttotalTokens: 0,\n\t\t\t\tcost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0, total: 0 },\n\t\t\t},\n\t\t\tstopReason: \"pending\",\n\t\t\ttimestamp: Date.now(),\n\t\t};\n\n\t\tconst streamDeadline = createStreamDeadline(options?.streamDeadlineMs, options?.signal);\n\n\t\tlet inputTransformations: AnthropicInputTransformation[] | undefined;\n\n\t\ttry {\n\t\t\tlet client: Anthropic;\n\t\t\tlet isOAuth: boolean;\n\t\t\tlet usageModel = model;\n\n\t\t\tif (options?.client) {\n\t\t\t\tclient = options.client;\n\t\t\t\tisOAuth = false;\n\t\t\t} else {\n\t\t\t\tconst apiKey = options?.apiKey;\n\t\t\t\tassertRequestAuth(model.provider, apiKey, options?.headers);\n\n\t\t\t\tlet copilotDynamicHeaders: Record<string, string> | undefined;\n\t\t\t\tif (model.provider === \"github-copilot\") {\n\t\t\t\t\tconst hasImages = hasCopilotVisionInput(context.messages);\n\t\t\t\t\tcopilotDynamicHeaders = preserveCopilotIntegrationHeader(\n\t\t\t\t\t\tmodel.headers,\n\t\t\t\t\t\tbuildCopilotDynamicHeaders({\n\t\t\t\t\t\t\tmessages: context.messages,\n\t\t\t\t\t\t\thasImages,\n\t\t\t\t\t\t\tapiKey,\n\t\t\t\t\t\t}),\n\t\t\t\t\t);\n\t\t\t\t}\n\n\t\t\t\tconst cacheRetention = resolveCacheRetention(options?.cacheRetention, options?.env);\n\t\t\t\tconst cacheSessionId = cacheRetention === \"none\" ? undefined : options?.sessionId;\n\n\t\t\t\tconst created = createClient(\n\t\t\t\t\tmodel,\n\t\t\t\t\tapiKey,\n\t\t\t\t\tmodel.reasoning && options?.thinkingEnabled === true && (options.interleavedThinking ?? true),\n\t\t\t\t\tshouldUseFineGrainedToolStreamingBeta(model, context),\n\t\t\t\t\tshouldUseServerSideFallbackBeta(model),\n\t\t\t\t\toptions?.headers,\n\t\t\t\t\toptions?.fetch,\n\t\t\t\t\tcopilotDynamicHeaders,\n\t\t\t\t\tcacheSessionId,\n\t\t\t\t);\n\t\t\t\tclient = created.client;\n\t\t\t\tisOAuth = created.isOAuthToken;\n\t\t\t}\n\t\t\tlet params = buildParams(model, context, isOAuth, options);\n\t\t\tconst nextParams = await options?.onPayload?.(params, model);\n\t\t\tif (nextParams !== undefined) {\n\t\t\t\tparams = nextParams as MessageCreateParamsStreaming;\n\t\t\t}\n\t\t\tconst requestOptions = {\n\t\t\t\t...(streamDeadline.signal ? { signal: streamDeadline.signal } : {}),\n\t\t\t\t...(options?.timeoutMs !== undefined ? { timeout: options.timeoutMs } : {}),\n\t\t\t\tmaxRetries: 0,\n\t\t\t};\n\t\t\tconst response = await retryProviderRequest(\n\t\t\t\t() => client.messages.create({ ...params, stream: true }, requestOptions).asResponse(),\n\t\t\t\t{\n\t\t\t\t\tmaxRetries: options?.maxRetries,\n\t\t\t\t\tmaxRetryDelayMs: options?.maxRetryDelayMs,\n\t\t\t\t\tsignal: streamDeadline.signal,\n\t\t\t\t},\n\t\t\t);\n\t\t\tawait options?.onResponse?.({ status: response.status, headers: headersToRecord(response.headers) }, model);\n\t\t\tstream.push({ type: \"start\", partial: output });\n\n\t\t\ttype Block = (ThinkingContent | TextContent | (ToolCall & { partialJson: string })) & { index: number };\n\t\t\tconst blocks = output.content as Block[];\n\n\t\t\tfor await (const event of withStreamDeadline(\n\t\t\t\titerateAnthropicEvents(response, streamDeadline.signal),\n\t\t\t\tstreamDeadline.deadlineMs,\n\t\t\t\tstreamDeadline.abort,\n\t\t\t)) {\n\t\t\t\tif (event.type === \"message_start\") {\n\t\t\t\t\toutput.responseId = event.message.id;\n\t\t\t\t\toutput.model = event.message.model;\n\t\t\t\t\tconst fallbackCost =\n\t\t\t\t\t\toutput.model === model.id\n\t\t\t\t\t\t\t? undefined\n\t\t\t\t\t\t\t: model.compat?.allowedFallbackModels?.find(\n\t\t\t\t\t\t\t\t\t(fallback) => fallback.provider === model.provider && fallback.model === output.model,\n\t\t\t\t\t\t\t\t)?.cost;\n\t\t\t\t\tusageModel = fallbackCost ? { ...model, id: output.model, cost: fallbackCost } : model;\n\t\t\t\t\t// Capture initial token usage from message_start event\n\t\t\t\t\t// This ensures we have input token counts even if the stream is aborted early\n\t\t\t\t\toutput.usage.input = event.message.usage.input_tokens || 0;\n\t\t\t\t\toutput.usage.output = event.message.usage.output_tokens || 0;\n\t\t\t\t\toutput.usage.cacheRead = event.message.usage.cache_read_input_tokens || 0;\n\t\t\t\t\toutput.usage.cacheWrite = event.message.usage.cache_creation_input_tokens || 0;\n\t\t\t\t\toutput.usage.cacheWrite1h = event.message.usage.cache_creation?.ephemeral_1h_input_tokens || 0;\n\t\t\t\t\t// Anthropic doesn't provide total_tokens, compute from components\n\t\t\t\t\toutput.usage.totalTokens =\n\t\t\t\t\t\toutput.usage.input + output.usage.output + output.usage.cacheRead + output.usage.cacheWrite;\n\t\t\t\t\tcalculateCost(usageModel, output.usage);\n\t\t\t\t\tinputTransformations = getInputTransformations(event.message) ?? inputTransformations;\n\t\t\t\t} else if (event.type === \"content_block_start\") {\n\t\t\t\t\tif (event.content_block.type === \"text\") {\n\t\t\t\t\t\tconst block: Block = {\n\t\t\t\t\t\t\ttype: \"text\",\n\t\t\t\t\t\t\ttext: event.content_block.text ?? \"\",\n\t\t\t\t\t\t\tindex: event.index,\n\t\t\t\t\t\t};\n\t\t\t\t\t\toutput.content.push(block);\n\t\t\t\t\t\tstream.push({ type: \"text_start\", contentIndex: output.content.length - 1, partial: output });\n\t\t\t\t\t} else if (event.content_block.type === \"thinking\") {\n\t\t\t\t\t\tconst block: Block = {\n\t\t\t\t\t\t\ttype: \"thinking\",\n\t\t\t\t\t\t\tthinking: event.content_block.thinking ?? \"\",\n\t\t\t\t\t\t\tthinkingSignature: event.content_block.signature ?? \"\",\n\t\t\t\t\t\t\tindex: event.index,\n\t\t\t\t\t\t};\n\t\t\t\t\t\toutput.content.push(block);\n\t\t\t\t\t\tstream.push({ type: \"thinking_start\", contentIndex: output.content.length - 1, partial: output });\n\t\t\t\t\t} else if (event.content_block.type === \"redacted_thinking\") {\n\t\t\t\t\t\tconst block: Block = {\n\t\t\t\t\t\t\ttype: \"thinking\",\n\t\t\t\t\t\t\tthinking: \"[Reasoning redacted]\",\n\t\t\t\t\t\t\tthinkingSignature: event.content_block.data,\n\t\t\t\t\t\t\tredacted: true,\n\t\t\t\t\t\t\tindex: event.index,\n\t\t\t\t\t\t};\n\t\t\t\t\t\toutput.content.push(block);\n\t\t\t\t\t\tstream.push({ type: \"thinking_start\", contentIndex: output.content.length - 1, partial: output });\n\t\t\t\t\t} else if (event.content_block.type === \"tool_use\") {\n\t\t\t\t\t\tconst block: Block = {\n\t\t\t\t\t\t\ttype: \"toolCall\",\n\t\t\t\t\t\t\tid: event.content_block.id,\n\t\t\t\t\t\t\tname: isOAuth\n\t\t\t\t\t\t\t\t? fromClaudeCodeName(event.content_block.name, context.tools)\n\t\t\t\t\t\t\t\t: event.content_block.name,\n\t\t\t\t\t\t\targuments: (event.content_block.input as Record<string, any>) ?? {},\n\t\t\t\t\t\t\tpartialJson: \"\",\n\t\t\t\t\t\t\tindex: event.index,\n\t\t\t\t\t\t};\n\t\t\t\t\t\toutput.content.push(block);\n\t\t\t\t\t\tstream.push({ type: \"toolcall_start\", contentIndex: output.content.length - 1, partial: output });\n\t\t\t\t\t} else if ((event.content_block as { type?: string }).type === \"fallback\") {\n\t\t\t\t\t\t// Server-side fallback boundary: \"the `fallback` block (an ordinary\n\t\t\t\t\t\t// `content_block_start` and `content_block_stop` pair with no deltas) marks the\n\t\t\t\t\t\t// boundary\", and clients must \"Keep it exactly where it appeared. The API uses\n\t\t\t\t\t\t// its position to validate the thinking blocks around it\".\n\t\t\t\t\t\t// https://platform.claude.com/docs/en/build-with-claude/refusals-and-fallback\n\t\t\t\t\t\t//\n\t\t\t\t\t\t// The SDK types lag this block, so read it through a narrow cast.\n\t\t\t\t\t\tconst fallbackBlock = event.content_block as unknown as {\n\t\t\t\t\t\t\tfrom?: { model?: string };\n\t\t\t\t\t\t\tto?: { model?: string };\n\t\t\t\t\t\t};\n\t\t\t\t\t\tconst toModel = fallbackBlock.to?.model;\n\t\t\t\t\t\toutput.content.push({\n\t\t\t\t\t\t\ttype: \"fallback\",\n\t\t\t\t\t\t\tfromModel: fallbackBlock.from?.model ?? output.model,\n\t\t\t\t\t\t\ttoModel: toModel ?? output.model,\n\t\t\t\t\t\t});\n\t\t\t\t\t\t// `message_start` named the requested model, so on a mid-output fallback the\n\t\t\t\t\t\t// serving model is only knowable here. Re-derive pricing the same way the\n\t\t\t\t\t\t// `message_start` branch does, so the returned message is costed at the rates\n\t\t\t\t\t\t// of the model that actually produced it.\n\t\t\t\t\t\tif (toModel && toModel !== output.model) {\n\t\t\t\t\t\t\toutput.model = toModel;\n\t\t\t\t\t\t\tconst fallbackCost = model.compat?.allowedFallbackModels?.find(\n\t\t\t\t\t\t\t\t(fallback) => fallback.provider === model.provider && fallback.model === toModel,\n\t\t\t\t\t\t\t)?.cost;\n\t\t\t\t\t\t\tusageModel = fallbackCost ? { ...model, id: toModel, cost: fallbackCost } : model;\n\t\t\t\t\t\t}\n\t\t\t\t\t}\n\t\t\t\t} else if (event.type === \"content_block_delta\") {\n\t\t\t\t\tif (event.delta.type === \"text_delta\") {\n\t\t\t\t\t\tconst index = blocks.findIndex((b) => b.index === event.index);\n\t\t\t\t\t\tconst block = blocks[index];\n\t\t\t\t\t\tif (block && block.type === \"text\") {\n\t\t\t\t\t\t\tblock.text += event.delta.text;\n\t\t\t\t\t\t\tstream.push({\n\t\t\t\t\t\t\t\ttype: \"text_delta\",\n\t\t\t\t\t\t\t\tcontentIndex: index,\n\t\t\t\t\t\t\t\tdelta: event.delta.text,\n\t\t\t\t\t\t\t\tpartial: output,\n\t\t\t\t\t\t\t});\n\t\t\t\t\t\t}\n\t\t\t\t\t} else if (event.delta.type === \"thinking_delta\") {\n\t\t\t\t\t\tconst index = blocks.findIndex((b) => b.index === event.index);\n\t\t\t\t\t\tconst block = blocks[index];\n\t\t\t\t\t\tif (block && block.type === \"thinking\") {\n\t\t\t\t\t\t\tblock.thinking += event.delta.thinking;\n\t\t\t\t\t\t\tstream.push({\n\t\t\t\t\t\t\t\ttype: \"thinking_delta\",\n\t\t\t\t\t\t\t\tcontentIndex: index,\n\t\t\t\t\t\t\t\tdelta: event.delta.thinking,\n\t\t\t\t\t\t\t\tpartial: output,\n\t\t\t\t\t\t\t});\n\t\t\t\t\t\t}\n\t\t\t\t\t} else if (event.delta.type === \"input_json_delta\") {\n\t\t\t\t\t\tconst index = blocks.findIndex((b) => b.index === event.index);\n\t\t\t\t\t\tconst block = blocks[index];\n\t\t\t\t\t\tif (block && block.type === \"toolCall\") {\n\t\t\t\t\t\t\tblock.partialJson += event.delta.partial_json;\n\t\t\t\t\t\t\tblock.arguments = parseStreamingJson(block.partialJson);\n\t\t\t\t\t\t\tstream.push({\n\t\t\t\t\t\t\t\ttype: \"toolcall_delta\",\n\t\t\t\t\t\t\t\tcontentIndex: index,\n\t\t\t\t\t\t\t\tdelta: event.delta.partial_json,\n\t\t\t\t\t\t\t\tpartial: output,\n\t\t\t\t\t\t\t});\n\t\t\t\t\t\t}\n\t\t\t\t\t} else if (event.delta.type === \"signature_delta\") {\n\t\t\t\t\t\tconst index = blocks.findIndex((b) => b.index === event.index);\n\t\t\t\t\t\tconst block = blocks[index];\n\t\t\t\t\t\tif (block && block.type === \"thinking\") {\n\t\t\t\t\t\t\tblock.thinkingSignature = block.thinkingSignature || \"\";\n\t\t\t\t\t\t\tblock.thinkingSignature += event.delta.signature;\n\t\t\t\t\t\t}\n\t\t\t\t\t}\n\t\t\t\t} else if (event.type === \"content_block_stop\") {\n\t\t\t\t\tconst index = blocks.findIndex((b) => b.index === event.index);\n\t\t\t\t\tconst block = blocks[index];\n\t\t\t\t\tif (block) {\n\t\t\t\t\t\tdelete (block as any).index;\n\t\t\t\t\t\tif (block.type === \"text\") {\n\t\t\t\t\t\t\tstream.push({\n\t\t\t\t\t\t\t\ttype: \"text_end\",\n\t\t\t\t\t\t\t\tcontentIndex: index,\n\t\t\t\t\t\t\t\tcontent: block.text,\n\t\t\t\t\t\t\t\tpartial: output,\n\t\t\t\t\t\t\t});\n\t\t\t\t\t\t} else if (block.type === \"thinking\") {\n\t\t\t\t\t\t\tstream.push({\n\t\t\t\t\t\t\t\ttype: \"thinking_end\",\n\t\t\t\t\t\t\t\tcontentIndex: index,\n\t\t\t\t\t\t\t\tcontent: block.thinking,\n\t\t\t\t\t\t\t\tpartial: output,\n\t\t\t\t\t\t\t});\n\t\t\t\t\t\t} else if (block.type === \"toolCall\") {\n\t\t\t\t\t\t\tblock.arguments = parseStreamingJson(block.partialJson);\n\t\t\t\t\t\t\t// Finalize in-place and strip the scratch buffer so replay only\n\t\t\t\t\t\t\t// carries parsed arguments.\n\t\t\t\t\t\t\tdelete (block as { partialJson?: string }).partialJson;\n\t\t\t\t\t\t\tstream.push({\n\t\t\t\t\t\t\t\ttype: \"toolcall_end\",\n\t\t\t\t\t\t\t\tcontentIndex: index,\n\t\t\t\t\t\t\t\ttoolCall: block,\n\t\t\t\t\t\t\t\tpartial: output,\n\t\t\t\t\t\t\t});\n\t\t\t\t\t\t}\n\t\t\t\t\t}\n\t\t\t\t} else if (event.type === \"message_delta\") {\n\t\t\t\t\tif (event.delta.stop_reason) {\n\t\t\t\t\t\toutput.rawStopReason = event.delta.stop_reason;\n\t\t\t\t\t\tconst stopReasonResult = mapStopReason(event.delta.stop_reason, event.delta.stop_details);\n\t\t\t\t\t\toutput.stopReason = stopReasonResult.stopReason;\n\t\t\t\t\t\tif (stopReasonResult.errorMessage) {\n\t\t\t\t\t\t\toutput.errorMessage = stopReasonResult.errorMessage;\n\t\t\t\t\t\t}\n\t\t\t\t\t}\n\t\t\t\t\t// Only update usage fields if present (not null).\n\t\t\t\t\t// Preserves input_tokens from message_start when proxies omit it in message_delta.\n\t\t\t\t\tif (event.usage) {\n\t\t\t\t\t\tif (event.usage.input_tokens != null) {\n\t\t\t\t\t\t\toutput.usage.input = event.usage.input_tokens;\n\t\t\t\t\t\t}\n\t\t\t\t\t\tif (event.usage.output_tokens != null) {\n\t\t\t\t\t\t\toutput.usage.output = event.usage.output_tokens;\n\t\t\t\t\t\t}\n\t\t\t\t\t\tif (event.usage.cache_read_input_tokens != null) {\n\t\t\t\t\t\t\toutput.usage.cacheRead = event.usage.cache_read_input_tokens;\n\t\t\t\t\t\t}\n\t\t\t\t\t\tif (event.usage.cache_creation_input_tokens != null) {\n\t\t\t\t\t\t\toutput.usage.cacheWrite = event.usage.cache_creation_input_tokens;\n\t\t\t\t\t\t}\n\t\t\t\t\t\t// Anthropic reports reasoning tokens in `output_tokens_details.thinking_tokens` on the\n\t\t\t\t\t\t// final message_delta usage (a subset of output_tokens). SDK 0.91.1 omits the field from\n\t\t\t\t\t\t// its Usage type, so read it through a narrow cast. Verified against the live API.\n\t\t\t\t\t\tconst thinkingTokens = (event.usage as { output_tokens_details?: { thinking_tokens?: number } })\n\t\t\t\t\t\t\t.output_tokens_details?.thinking_tokens;\n\t\t\t\t\t\tif (thinkingTokens != null) {\n\t\t\t\t\t\t\toutput.usage.reasoning = thinkingTokens;\n\t\t\t\t\t\t}\n\t\t\t\t\t}\n\t\t\t\t\t// Anthropic doesn't provide total_tokens, compute from components\n\t\t\t\t\toutput.usage.totalTokens =\n\t\t\t\t\t\toutput.usage.input + output.usage.output + output.usage.cacheRead + output.usage.cacheWrite;\n\t\t\t\t\tcalculateCost(usageModel, output.usage);\n\t\t\t\t\t// A non-empty final fallback report supersedes the request-start report with\n\t\t\t\t\t// the serving model's transformations. An empty report must not erase drops\n\t\t\t\t\t// already reported at message_start.\n\t\t\t\t\tconst servingTransformations = getInputTransformations(event);\n\t\t\t\t\tif (servingTransformations && servingTransformations.length > 0) {\n\t\t\t\t\t\tinputTransformations = servingTransformations;\n\t\t\t\t\t}\n\t\t\t\t\taddEarlierAttemptCosts(output, model, usageModel, event);\n\t\t\t\t}\n\t\t\t}\n\n\t\t\tif (inputTransformations) recordInputTransformations(output, inputTransformations);\n\t\t\tif (options?.signal?.aborted) {\n\t\t\t\tthrow new Error(\"Request was aborted\");\n\t\t\t}\n\n\t\t\tif (output.stopReason === \"pending\") {\n\t\t\t\tthrow new Error(\"Anthropic stream ended without a stop reason\");\n\t\t\t}\n\t\t\tif (output.stopReason === \"aborted\" || output.stopReason === \"error\") {\n\t\t\t\tthrow new Error(output.errorMessage || \"An unknown error occurred\");\n\t\t\t}\n\n\t\t\tstream.push({ type: \"done\", reason: output.stopReason, message: output });\n\t\t\tstream.end();\n\t\t} catch (error) {\n\t\t\tif (\n\t\t\t\tinputTransformations &&\n\t\t\t\t!output.diagnostics?.some((diagnostic) => diagnostic.type === \"anthropic_input_transformations\")\n\t\t\t) {\n\t\t\t\trecordInputTransformations(output, inputTransformations);\n\t\t\t}\n\t\t\tfor (const block of output.content) {\n\t\t\t\tdelete (block as { index?: number }).index;\n\t\t\t\t// partialJson is only a streaming scratch buffer; never persist it.\n\t\t\t\tdelete (block as { partialJson?: string }).partialJson;\n\t\t\t}\n\t\t\toutput.stopReason = options?.signal?.aborted ? \"aborted\" : \"error\";\n\t\t\toutput.errorMessage = error instanceof Error ? error.message : JSON.stringify(error);\n\t\t\tstream.push({ type: \"error\", reason: output.stopReason, error: output });\n\t\t\tstream.end();\n\t\t} finally {\n\t\t\tstreamDeadline.cleanup();\n\t\t}\n\t})();\n\n\treturn stream;\n};\n\n/**\n * Map ThinkingLevel to Anthropic effort levels for adaptive thinking.\n * Note: effort \"max\" is available on all adaptive-thinking Claude models, while native\n * \"xhigh\" is available on Opus 4.7, Opus 4.8, Opus 5, Sonnet 5, Fable 5, and Fable 5.1 —\n * the models the generator merges an `xhigh` mapping onto.\n */\nfunction mapThinkingLevelToEffort(\n\tmodel: Model<\"anthropic-messages\">,\n\tlevel: SimpleStreamOptions[\"reasoning\"],\n): AnthropicEffort {\n\tconst mapped = level ? model.thinkingLevelMap?.[level] : undefined;\n\tif (typeof mapped === \"string\") return mapped as AnthropicEffort;\n\n\tswitch (level) {\n\t\tcase \"minimal\":\n\t\tcase \"low\":\n\t\t\treturn \"low\";\n\t\tcase \"medium\":\n\t\t\treturn \"medium\";\n\t\tcase \"high\":\n\t\t\treturn \"high\";\n\t\tdefault:\n\t\t\treturn \"high\";\n\t}\n}\n\nexport const streamSimple: StreamFunction<\"anthropic-messages\", SimpleStreamOptions> = (\n\tmodel: Model<\"anthropic-messages\">,\n\tcontext: Context,\n\toptions?: SimpleStreamOptions,\n): AssistantMessageEventStream => {\n\tassertRequestAuth(model.provider, options?.apiKey, options?.headers);\n\n\tconst base = {\n\t\t...buildBaseOptions(model, context, options, options?.apiKey),\n\t\ttoolChoice: options?.toolChoice,\n\t} satisfies AnthropicOptions;\n\tif (!options?.reasoning) {\n\t\treturn stream(model, context, {\n\t\t\t...base,\n\t\t\tthinkingEnabled: false,\n\t\t} satisfies AnthropicOptions);\n\t}\n\n\t// For models with adaptive thinking: use an effort level.\n\t// For older models: use budget-based thinking.\n\tif (model.compat?.forceAdaptiveThinking === true) {\n\t\tconst effort = mapThinkingLevelToEffort(model, options.reasoning);\n\t\treturn stream(model, context, {\n\t\t\t...base,\n\t\t\tthinkingEnabled: true,\n\t\t\teffort,\n\t\t} satisfies AnthropicOptions);\n\t}\n\n\t// Undefined means the caller did not request an output cap; let the helper use the model cap.\n\t// Do not coerce to 0 here, or the thinking budget would become the entire max_tokens value.\n\tconst adjusted = adjustMaxTokensForThinking(\n\t\tbase.maxTokens,\n\t\tmodel.maxTokens,\n\t\toptions.reasoning,\n\t\toptions.thinkingBudgets,\n\t);\n\n\tconst maxTokens = clampMaxTokensToContext(model, context, adjusted.maxTokens);\n\n\treturn stream(model, context, {\n\t\t...base,\n\t\tmaxTokens,\n\t\tthinkingEnabled: true,\n\t\tthinkingBudgetTokens: Math.min(adjusted.thinkingBudget, Math.max(0, maxTokens - 1024)),\n\t} satisfies AnthropicOptions);\n};\n\nfunction isOAuthToken(apiKey: string): boolean {\n\treturn apiKey.includes(\"sk-ant-oat\");\n}\n\nfunction createClient(\n\tmodel: Model<\"anthropic-messages\">,\n\tapiKey: string | undefined,\n\tinterleavedThinking: boolean,\n\tuseFineGrainedToolStreamingBeta: boolean,\n\tuseServerSideFallbackBeta: boolean,\n\toptionsHeaders?: ProviderHeaders,\n\tfetch?: typeof globalThis.fetch,\n\tdynamicHeaders?: Record<string, string>,\n\tsessionId?: string,\n): { client: Anthropic; isOAuthToken: boolean } {\n\t// Adaptive thinking models have interleaved thinking built in, so skip the beta header.\n\tconst needsInterleavedBeta = interleavedThinking && model.compat?.forceAdaptiveThinking !== true;\n\tconst betaFeatures: string[] = [];\n\tif (useFineGrainedToolStreamingBeta) {\n\t\tbetaFeatures.push(FINE_GRAINED_TOOL_STREAMING_BETA);\n\t}\n\tif (needsInterleavedBeta) {\n\t\tbetaFeatures.push(INTERLEAVED_THINKING_BETA);\n\t}\n\tif (useServerSideFallbackBeta) {\n\t\tbetaFeatures.push(SERVER_SIDE_FALLBACK_BETA);\n\t}\n\tif (shouldUseThinkingBindingControlsBeta(model)) {\n\t\tbetaFeatures.push(THINKING_BINDING_CONTROLS_BETA);\n\t}\n\tif (supportsMidConvoEffort(model)) {\n\t\tbetaFeatures.push(MID_CONVERSATION_OUTPUT_CONFIG_BETA, THINKING_BINDING_CONTROLS_BETA);\n\t}\n\tconst uniqueBetaFeatures = [...new Set(betaFeatures)];\n\n\t// Copilot: Bearer auth, selective betas.\n\tif (model.provider === \"github-copilot\") {\n\t\tconst client = new Anthropic({\n\t\t\tapiKey: null,\n\t\t\tauthToken: apiKey ?? null,\n\t\t\tbaseURL: model.baseUrl,\n\t\t\tdangerouslyAllowBrowser: true,\n\t\t\tfetch,\n\t\t\tdefaultHeaders: mergeClientHeaders(\n\t\t\t\t{\n\t\t\t\t\taccept: \"application/json\",\n\t\t\t\t\t\"anthropic-dangerous-direct-browser-access\": \"true\",\n\t\t\t\t\t...(uniqueBetaFeatures.length > 0 ? { \"anthropic-beta\": uniqueBetaFeatures.join(\",\") } : {}),\n\t\t\t\t},\n\t\t\t\tmodel.headers,\n\t\t\t\tdynamicHeaders,\n\t\t\t\toptionsHeaders,\n\t\t\t),\n\t\t});\n\n\t\treturn { client, isOAuthToken: false };\n\t}\n\n\t// OAuth: Bearer auth, Claude Code identity headers\n\tif (apiKey && isOAuthToken(apiKey)) {\n\t\tconst client = new Anthropic({\n\t\t\tapiKey: null,\n\t\t\tauthToken: apiKey,\n\t\t\tbaseURL: model.baseUrl,\n\t\t\tdangerouslyAllowBrowser: true,\n\t\t\tfetch,\n\t\t\tdefaultHeaders: mergeClientHeaders(\n\t\t\t\t{\n\t\t\t\t\taccept: \"application/json\",\n\t\t\t\t\t\"anthropic-dangerous-direct-browser-access\": \"true\",\n\t\t\t\t\t\"anthropic-beta\": [\"claude-code-20250219\", \"oauth-2025-04-20\", ...uniqueBetaFeatures].join(\",\"),\n\t\t\t\t\t\"user-agent\": `claude-cli/${claudeCodeVersion}`,\n\t\t\t\t\t\"x-app\": \"cli\",\n\t\t\t\t},\n\t\t\t\tmodel.headers,\n\t\t\t\toptionsHeaders,\n\t\t\t),\n\t\t});\n\n\t\treturn { client, isOAuthToken: true };\n\t}\n\n\t// API key or header-owned auth.\n\tconst compat = getAnthropicCompat(model);\n\tconst sessionAffinityHeaders: ProviderHeaders = {};\n\tif (sessionId && compat.sendSessionAffinityHeaders) {\n\t\tconst header = compat.sessionAffinityFormat === \"openrouter\" ? \"x-session-id\" : \"x-session-affinity\";\n\t\tsessionAffinityHeaders[header] = sessionId;\n\t}\n\tconst defaultHeaders = mergeClientHeaders(\n\t\t{\n\t\t\taccept: \"application/json\",\n\t\t\t\"anthropic-dangerous-direct-browser-access\": \"true\",\n\t\t\t...(uniqueBetaFeatures.length > 0 ? { \"anthropic-beta\": uniqueBetaFeatures.join(\",\") } : {}),\n\t\t},\n\t\tsessionAffinityHeaders,\n\t\tmodel.headers,\n\t\toptionsHeaders,\n\t);\n\tconst client = new Anthropic({\n\t\tapiKey: apiKey ?? null,\n\t\tauthToken: null,\n\t\tbaseURL: model.baseUrl,\n\t\tdangerouslyAllowBrowser: true,\n\t\tfetch,\n\t\tdefaultHeaders,\n\t});\n\n\treturn { client, isOAuthToken: false };\n}\n\nfunction buildParams(\n\tmodel: Model<\"anthropic-messages\">,\n\tcontext: Context,\n\tisOAuthToken: boolean,\n\toptions?: AnthropicOptions,\n): MessageCreateParamsStreamingWithFallbacks {\n\tconst { cacheControl } = getCacheControl(model, options?.cacheRetention, options?.env);\n\tconst compat = getAnthropicCompat(model);\n\tconst transformedMessages = transformMessages(context.messages, model, normalizeToolCallId);\n\tconst normalizeToolName = isOAuthToken ? toClaudeCodeName : (name: string) => name;\n\tconst toolPlacement = splitDeferredTools(\n\t\t{ ...context, messages: transformedMessages },\n\t\tcompat.supportsToolReferences,\n\t\tnormalizeToolName,\n\t);\n\tlet immediateTools = toolPlacement.immediate;\n\tlet deferredTools = [...toolPlacement.deferred.values()];\n\tif (immediateTools.length === 0 && deferredTools.length > 0) {\n\t\timmediateTools = deferredTools;\n\t\tdeferredTools = [];\n\t}\n\tconst deferredToolNames = new Set(deferredTools.map((tool) => normalizeToolName(tool.name)));\n\tconst converted = convertMessages(\n\t\ttransformedMessages,\n\t\tisOAuthToken,\n\t\tcacheControl,\n\t\tcompat.allowEmptySignature,\n\t\tdeferredToolNames,\n\t\tnormalizeToolName,\n\t\tsupportsMidConvoEffort(model) ? model.provider : undefined,\n\t);\n\tconst activeEffort = options?.effort ?? \"high\";\n\tconst params: MessageCreateParamsStreamingWithFallbacks = {\n\t\tmodel: model.id,\n\t\tmessages: (supportsMidConvoEffort(model)\n\t\t\t? insertThinkingLevelMessages(converted, activeEffort)\n\t\t\t: converted.messages) as MessageParam[],\n\t\tmax_tokens: options?.maxTokens ?? model.maxTokens,\n\t\tstream: true,\n\t};\n\n\t// For OAuth tokens, we MUST include Claude Code identity\n\tif (isOAuthToken) {\n\t\tparams.system = [\n\t\t\t{\n\t\t\t\ttype: \"text\",\n\t\t\t\ttext: \"You are Claude Code, Anthropic's official CLI for Claude.\",\n\t\t\t\t...(cacheControl ? { cache_control: cacheControl } : {}),\n\t\t\t},\n\t\t];\n\t\tif (context.systemPrompt) {\n\t\t\tparams.system.push({\n\t\t\t\ttype: \"text\",\n\t\t\t\ttext: sanitizeSurrogates(context.systemPrompt),\n\t\t\t\t...(cacheControl ? { cache_control: cacheControl } : {}),\n\t\t\t});\n\t\t}\n\t} else if (context.systemPrompt) {\n\t\t// Add cache control to system prompt for non-OAuth tokens\n\t\tparams.system = [\n\t\t\t{\n\t\t\t\ttype: \"text\",\n\t\t\t\ttext: sanitizeSurrogates(context.systemPrompt),\n\t\t\t\t...(cacheControl ? { cache_control: cacheControl } : {}),\n\t\t\t},\n\t\t];\n\t}\n\n\t// Temperature is incompatible with extended thinking and unsupported on Claude Opus 4.7+.\n\tif (\n\t\toptions?.temperature !== undefined &&\n\t\t!options?.thinkingEnabled &&\n\t\t!supportsMidConvoEffort(model) &&\n\t\tcompat.supportsTemperature\n\t) {\n\t\tparams.temperature = options.temperature;\n\t}\n\n\tif (immediateTools.length > 0 || deferredTools.length > 0) {\n\t\tparams.tools = [\n\t\t\t...convertTools(\n\t\t\t\timmediateTools,\n\t\t\t\tisOAuthToken,\n\t\t\t\tcompat.supportsEagerToolInputStreaming,\n\t\t\t\tcompat.supportsStrictTools,\n\t\t\t\tcompat.supportsCacheControlOnTools ? cacheControl : undefined,\n\t\t\t),\n\t\t\t...convertTools(\n\t\t\t\tdeferredTools,\n\t\t\t\tisOAuthToken,\n\t\t\t\tcompat.supportsEagerToolInputStreaming,\n\t\t\t\tcompat.supportsStrictTools,\n\t\t\t\tundefined,\n\t\t\t\ttrue,\n\t\t\t),\n\t\t];\n\t}\n\n\t// Managed effort models always use adaptive thinking. Their per-turn markers\n\t// carry the actual effort; the top-level high value deliberately stays stable.\n\tif (supportsMidConvoEffort(model)) {\n\t\tparams.thinking = {\n\t\t\ttype: \"adaptive\",\n\t\t\tdisplay: options?.thinkingDisplay ?? \"summarized\",\n\t\t\tblock_binding: { prefix_mismatch_behavior: \"drop_block\" },\n\t\t} as unknown as NonNullable<MessageCreateParamsStreaming[\"thinking\"]>;\n\t\tparams.output_config = { effort: \"high\" };\n\t} else if (model.reasoning) {\n\t\tif (options?.thinkingEnabled) {\n\t\t\t// Default to \"summarized\" so Opus 4.7 and Mythos Preview behave like\n\t\t\t// older Claude 4 models (whose API default is also \"summarized\").\n\t\t\tconst display: AnthropicThinkingDisplay = options.thinkingDisplay ?? \"summarized\";\n\t\t\tif (model.compat?.forceAdaptiveThinking === true) {\n\t\t\t\t// Adaptive thinking: Claude decides when and how much to think.\n\t\t\t\tparams.thinking = { type: \"adaptive\", display };\n\t\t\t\tif (options.effort) {\n\t\t\t\t\t// The Anthropic SDK types can lag newly supported effort values such as \"xhigh\".\n\t\t\t\t\tparams.output_config =\n\t\t\t\t\t\toptions.effort === \"xhigh\"\n\t\t\t\t\t\t\t? ({ effort: options.effort } as unknown as NonNullable<\n\t\t\t\t\t\t\t\t\tMessageCreateParamsStreaming[\"output_config\"]\n\t\t\t\t\t\t\t\t>)\n\t\t\t\t\t\t\t: { effort: options.effort };\n\t\t\t\t}\n\t\t\t} else {\n\t\t\t\t// Budget-based thinking for older models\n\t\t\t\tparams.thinking = {\n\t\t\t\t\ttype: \"enabled\",\n\t\t\t\t\tbudget_tokens: options.thinkingBudgetTokens || 1024,\n\t\t\t\t\tdisplay,\n\t\t\t\t};\n\t\t\t}\n\t\t} else if (options?.thinkingEnabled === false && model.thinkingLevelMap?.off !== null) {\n\t\t\tparams.thinking = { type: \"disabled\" };\n\t\t}\n\n\t\t// `block_binding` must accompany the beta header on *every* request for a model that\n\t\t// enforces the conversation check, not only on reasoning turns. The header alone leaves\n\t\t// `prefix_mismatch_behavior` at its `\"error\"` default, which is the 400 this opts out of.\n\t\t// `streamSimple` sets `thinkingEnabled: false` whenever no reasoning level is requested,\n\t\t// and Claude Fable 5.1 denies thinking-off (`thinkingLevelMap.off === null`), so without\n\t\t// this the no-reasoning path would send the header with no field at all.\n\t\tif (shouldUseThinkingBindingControlsBeta(model)) {\n\t\t\tconst blockBinding = { block_binding: { prefix_mismatch_behavior: \"drop_block\" } };\n\t\t\t// The Anthropic SDK types lag the `thinking-binding-controls-2026-08-01` beta.\n\t\t\tif (params.thinking?.type === \"adaptive\" || params.thinking?.type === \"enabled\") {\n\t\t\t\t// Accepted alongside both thinking types. It only changes what happens to a block\n\t\t\t\t// replayed behind a changed prefix; the model check always drops regardless.\n\t\t\t\tparams.thinking = { ...params.thinking, ...blockBinding } as unknown as NonNullable<\n\t\t\t\t\tMessageCreateParamsStreaming[\"thinking\"]\n\t\t\t\t>;\n\t\t\t} else if (!params.thinking && model.compat?.forceAdaptiveThinking === true) {\n\t\t\t\t// Adaptive thinking is always on for these models, so omitting `thinking` and\n\t\t\t\t// sending `{type: \"adaptive\"}` are equivalent. `display` stays absent to keep the\n\t\t\t\t// API's `\"omitted\"` default, which is exactly what omitting `thinking` produced.\n\t\t\t\tparams.thinking = { type: \"adaptive\", ...blockBinding } as unknown as NonNullable<\n\t\t\t\t\tMessageCreateParamsStreaming[\"thinking\"]\n\t\t\t\t>;\n\t\t\t}\n\t\t}\n\t}\n\n\tif (options?.metadata) {\n\t\tconst userId = options.metadata.user_id;\n\t\tif (typeof userId === \"string\") {\n\t\t\tparams.metadata = { user_id: userId };\n\t\t}\n\t}\n\n\tif (options?.toolChoice) {\n\t\tconst requested = options.toolChoice;\n\t\tconst isForced = requested === \"any\" || (typeof requested !== \"string\" && requested.type === \"tool\");\n\t\tif (isForced && model.compat?.supportsForcedToolChoice === false) {\n\t\t\t// \"The exceptions are Claude Fable 5.1 and Claude Mythos 5.1, which reject forced tool\n\t\t\t// use on every request with a 400 error. On those models, use\n\t\t\t// `tool_choice: {\"type\": \"auto\"}` with strict tool use or structured outputs instead.\"\n\t\t\t// https://platform.claude.com/docs/en/build-with-claude/thinking\n\t\t\t//\n\t\t\t// Reject the request rather than rewriting it. Silently substituting `auto` would\n\t\t\t// discard an explicit caller instruction and make `AnthropicOptions.toolChoice`'s\n\t\t\t// declared shape a lie; the caller asked the model to call a tool, and quietly asking\n\t\t\t// it to decide instead is a different request. Failing here matches how this package\n\t\t\t// handles other explicitly requested capabilities a model cannot honor, and surfaces\n\t\t\t// the remedy before a round trip that would 400 anyway. Callers that want the\n\t\t\t// substitution can make it themselves, and can branch on\n\t\t\t// `model.compat.supportsForcedToolChoice` to decide.\n\t\t\tconst requestedLabel = typeof requested === \"string\" ? requested : `tool \"${requested.name}\"`;\n\t\t\tthrow new Error(\n\t\t\t\t`Model ${model.id} does not support forced tool choice (requested: ${requestedLabel}). ` +\n\t\t\t\t\t`Use toolChoice \"auto\" with strict tool use or structured outputs instead.`,\n\t\t\t);\n\t\t}\n\t\tif (typeof requested === \"string\") {\n\t\t\tparams.tool_choice = { type: requested };\n\t\t} else {\n\t\t\tparams.tool_choice = requested;\n\t\t}\n\t}\n\n\tconst allowedFallbackModels = model.compat?.allowedFallbackModels;\n\tif (allowedFallbackModels && allowedFallbackModels.length > 0) {\n\t\tparams.fallbacks = allowedFallbackModels.map((fallback) => ({ model: fallback.model }));\n\t}\n\n\treturn params;\n}\n\n// Normalize tool call IDs to match Anthropic's required pattern and length\nfunction normalizeToolCallId(id: string): string {\n\treturn id.replace(/[^a-zA-Z0-9_-]/g, \"_\").slice(0, 64);\n}\n\nfunction convertToolResult(\n\tmsg: ToolResultMessage,\n\tisOAuthToken: boolean,\n\tdeferredToolNames: ReadonlySet<string>,\n\tloadedToolNames: Set<string>,\n\tnormalizeToolName: (name: string) => string,\n): { toolResult: ContentBlockParam; siblingContent: ContentBlockParam[] } {\n\tconst references: Array<{ type: \"tool_reference\"; tool_name: string }> = [];\n\tfor (const name of msg.addedToolNames ?? []) {\n\t\tconst normalizedName = normalizeToolName(name);\n\t\tif (!deferredToolNames.has(normalizedName) || loadedToolNames.has(normalizedName)) continue;\n\t\tloadedToolNames.add(normalizedName);\n\t\treferences.push({\n\t\t\ttype: \"tool_reference\",\n\t\t\ttool_name: isOAuthToken ? toClaudeCodeName(name) : name,\n\t\t});\n\t}\n\tconst convertedContent = convertContentBlocks(msg.content);\n\t// Anthropic rejects tool references mixed with ordinary tool-result content.\n\treturn {\n\t\ttoolResult: {\n\t\t\ttype: \"tool_result\",\n\t\t\ttool_use_id: msg.toolCallId,\n\t\t\tcontent: references.length > 0 ? references : convertedContent,\n\t\t\tis_error: msg.isError,\n\t\t},\n\t\tsiblingContent:\n\t\t\treferences.length === 0\n\t\t\t\t? []\n\t\t\t\t: typeof convertedContent === \"string\"\n\t\t\t\t\t? [{ type: \"text\", text: convertedContent }]\n\t\t\t\t\t: convertedContent,\n\t};\n}\n\nfunction convertMessages(\n\ttransformedMessages: Message[],\n\tisOAuthToken: boolean,\n\tcacheControl?: CacheControlEphemeral,\n\tallowEmptySignature = false,\n\tdeferredToolNames: ReadonlySet<string> = new Set(),\n\tnormalizeToolName: (name: string) => string = (name) => name,\n\tmanagedProvider?: string,\n): ConvertedAnthropicMessages {\n\tconst params: MessageParam[] = [];\n\tconst assistantLevels = new Map<number, AnthropicEffort>();\n\tconst loadedToolNames = new Set<string>();\n\n\tfor (let i = 0; i < transformedMessages.length; i++) {\n\t\tconst msg = transformedMessages[i];\n\n\t\tif (msg.role === \"user\") {\n\t\t\tif (typeof msg.content === \"string\") {\n\t\t\t\tif (msg.content.trim().length > 0) {\n\t\t\t\t\tparams.push({\n\t\t\t\t\t\trole: \"user\",\n\t\t\t\t\t\tcontent: sanitizeSurrogates(msg.content),\n\t\t\t\t\t});\n\t\t\t\t}\n\t\t\t} else {\n\t\t\t\tconst blocks: ContentBlockParam[] = msg.content.map((item) => {\n\t\t\t\t\tif (item.type === \"text\") {\n\t\t\t\t\t\treturn {\n\t\t\t\t\t\t\ttype: \"text\",\n\t\t\t\t\t\t\ttext: sanitizeSurrogates(item.text),\n\t\t\t\t\t\t};\n\t\t\t\t\t}\n\t\t\t\t\tif (item.type === \"document\") {\n\t\t\t\t\t\t// `BetaBase64PDFSource`. PDFs ride Claude's vision path, so no beta header.\n\t\t\t\t\t\t// The media type is a fixed literal in that SDK type, so it is hardcoded here\n\t\t\t\t\t\t// and the block's own field is verified rather than read.\n\t\t\t\t\t\t// https://platform.claude.com/docs/en/build-with-claude/pdf-support\n\t\t\t\t\t\tassertSupportedDocumentMimeType(item);\n\t\t\t\t\t\treturn {\n\t\t\t\t\t\t\ttype: \"document\",\n\t\t\t\t\t\t\tsource: {\n\t\t\t\t\t\t\t\ttype: \"base64\",\n\t\t\t\t\t\t\t\tmedia_type: \"application/pdf\",\n\t\t\t\t\t\t\t\tdata: item.data,\n\t\t\t\t\t\t\t},\n\t\t\t\t\t\t\t...(item.name ? { title: item.name } : {}),\n\t\t\t\t\t\t};\n\t\t\t\t\t}\n\t\t\t\t\treturn {\n\t\t\t\t\t\ttype: \"image\",\n\t\t\t\t\t\tsource: {\n\t\t\t\t\t\t\ttype: \"base64\",\n\t\t\t\t\t\t\tmedia_type: item.mimeType as \"image/jpeg\" | \"image/png\" | \"image/gif\" | \"image/webp\",\n\t\t\t\t\t\t\tdata: item.data,\n\t\t\t\t\t\t},\n\t\t\t\t\t};\n\t\t\t\t});\n\t\t\t\tconst filteredBlocks = blocks.filter((b) => {\n\t\t\t\t\tif (b.type === \"text\") {\n\t\t\t\t\t\treturn b.text.trim().length > 0;\n\t\t\t\t\t}\n\t\t\t\t\treturn true;\n\t\t\t\t});\n\t\t\t\tif (filteredBlocks.length === 0) continue;\n\t\t\t\tparams.push({\n\t\t\t\t\trole: \"user\",\n\t\t\t\t\tcontent: filteredBlocks,\n\t\t\t\t});\n\t\t\t}\n\t\t} else if (msg.role === \"assistant\") {\n\t\t\tconst blocks: ContentBlockParam[] = [];\n\n\t\t\tfor (const block of msg.content) {\n\t\t\t\tif (block.type === \"text\") {\n\t\t\t\t\tif (block.text.trim().length === 0) continue;\n\t\t\t\t\tblocks.push({\n\t\t\t\t\t\ttype: \"text\",\n\t\t\t\t\t\ttext: sanitizeSurrogates(block.text),\n\t\t\t\t\t});\n\t\t\t\t} else if (block.type === \"thinking\") {\n\t\t\t\t\t// Redacted thinking: pass the opaque payload back as redacted_thinking\n\t\t\t\t\tif (block.redacted) {\n\t\t\t\t\t\tblocks.push({\n\t\t\t\t\t\t\ttype: \"redacted_thinking\",\n\t\t\t\t\t\t\tdata: block.thinkingSignature!,\n\t\t\t\t\t\t});\n\t\t\t\t\t\tcontinue;\n\t\t\t\t\t}\n\t\t\t\t\tconst thinkingSignature = block.thinkingSignature;\n\t\t\t\t\tconst hasThinkingSignature = !!thinkingSignature && thinkingSignature.trim().length > 0;\n\t\t\t\t\tif (block.thinking.trim().length === 0 && !hasThinkingSignature) continue;\n\t\t\t\t\t// If thinking signature is missing/empty (e.g., from aborted stream),\n\t\t\t\t\t// convert to plain text for Anthropic. Some compatible providers emit\n\t\t\t\t\t// and accept empty signatures, so let marked models preserve the block.\n\t\t\t\t\tif (!hasThinkingSignature) {\n\t\t\t\t\t\tblocks.push(\n\t\t\t\t\t\t\tallowEmptySignature\n\t\t\t\t\t\t\t\t? {\n\t\t\t\t\t\t\t\t\t\ttype: \"thinking\",\n\t\t\t\t\t\t\t\t\t\tthinking: sanitizeSurrogates(block.thinking),\n\t\t\t\t\t\t\t\t\t\tsignature: \"\",\n\t\t\t\t\t\t\t\t\t}\n\t\t\t\t\t\t\t\t: {\n\t\t\t\t\t\t\t\t\t\ttype: \"text\",\n\t\t\t\t\t\t\t\t\t\ttext: sanitizeSurrogates(block.thinking),\n\t\t\t\t\t\t\t\t\t},\n\t\t\t\t\t\t);\n\t\t\t\t\t} else {\n\t\t\t\t\t\tblocks.push({\n\t\t\t\t\t\t\ttype: \"thinking\",\n\t\t\t\t\t\t\tthinking: sanitizeSurrogates(block.thinking),\n\t\t\t\t\t\t\tsignature: thinkingSignature,\n\t\t\t\t\t\t});\n\t\t\t\t\t}\n\t\t\t\t} else if (block.type === \"toolCall\") {\n\t\t\t\t\tblocks.push({\n\t\t\t\t\t\ttype: \"tool_use\",\n\t\t\t\t\t\tid: block.id,\n\t\t\t\t\t\tname: isOAuthToken ? toClaudeCodeName(block.name) : block.name,\n\t\t\t\t\t\tinput: block.arguments ?? {},\n\t\t\t\t\t});\n\t\t\t\t} else if (block.type === \"fallback\") {\n\t\t\t\t\t// The server-side fallback boundary must go back on the wire at the position it\n\t\t\t\t\t// arrived: \"Keep it exactly where it appeared. The API uses its position to\n\t\t\t\t\t// validate the thinking blocks around it, so a request that echoes thinking\n\t\t\t\t\t// blocks from both sides of the boundary is rejected if the block is omitted or\n\t\t\t\t\t// moved.\" `transformMessages` drops the pre-boundary thinking; without this\n\t\t\t\t\t// branch the marker would be dropped too, which is the failure that rule names.\n\t\t\t\t\t// https://platform.claude.com/docs/en/build-with-claude/refusals-and-fallback\n\t\t\t\t\t//\n\t\t\t\t\t// The SDK types lag this block, matching the cast on the stream side.\n\t\t\t\t\tblocks.push({\n\t\t\t\t\t\ttype: \"fallback\",\n\t\t\t\t\t\tfrom: { model: block.fromModel },\n\t\t\t\t\t\tto: { model: block.toModel },\n\t\t\t\t\t} as unknown as ContentBlockParam);\n\t\t\t\t}\n\t\t\t}\n\t\t\tif (blocks.length === 0) continue;\n\t\t\tconst messageIndex = params.length;\n\t\t\tparams.push({\n\t\t\t\trole: \"assistant\",\n\t\t\t\tcontent: blocks,\n\t\t\t});\n\t\t\tif (\n\t\t\t\tmanagedProvider !== undefined &&\n\t\t\t\tmsg.api === \"anthropic-messages\" &&\n\t\t\t\tmsg.provider === managedProvider &&\n\t\t\t\tisAnthropicEffort(msg.providerThinkingLevel)\n\t\t\t) {\n\t\t\t\tassistantLevels.set(messageIndex, msg.providerThinkingLevel);\n\t\t\t}\n\t\t} else if (msg.role === \"toolResult\") {\n\t\t\t// Collect all consecutive toolResult messages, needed for z.ai Anthropic endpoint.\n\t\t\tconst toolResults: ContentBlockParam[] = [];\n\t\t\tconst siblingContent: ContentBlockParam[] = [];\n\t\t\tlet j = i;\n\t\t\twhile (j < transformedMessages.length && transformedMessages[j].role === \"toolResult\") {\n\t\t\t\tconst converted = convertToolResult(\n\t\t\t\t\ttransformedMessages[j] as ToolResultMessage,\n\t\t\t\t\tisOAuthToken,\n\t\t\t\t\tdeferredToolNames,\n\t\t\t\t\tloadedToolNames,\n\t\t\t\t\tnormalizeToolName,\n\t\t\t\t);\n\t\t\t\ttoolResults.push(converted.toolResult);\n\t\t\t\tsiblingContent.push(...converted.siblingContent);\n\t\t\t\tj++;\n\t\t\t}\n\n\t\t\t// Skip the messages we've already processed.\n\t\t\ti = j - 1;\n\n\t\t\t// Displaced reference-bearing results must follow every tool_result block.\n\t\t\tparams.push({\n\t\t\t\trole: \"user\",\n\t\t\t\tcontent: [...toolResults, ...siblingContent],\n\t\t\t});\n\t\t}\n\t}\n\n\t// Add cache_control to the last user message to cache conversation history\n\tif (cacheControl && params.length > 0) {\n\t\tconst lastMessage = params[params.length - 1];\n\t\tif (lastMessage.role === \"user\") {\n\t\t\tif (Array.isArray(lastMessage.content)) {\n\t\t\t\tconst lastBlock = lastMessage.content[lastMessage.content.length - 1];\n\t\t\t\tif (\n\t\t\t\t\tlastBlock &&\n\t\t\t\t\t(lastBlock.type === \"text\" || lastBlock.type === \"image\" || lastBlock.type === \"tool_result\")\n\t\t\t\t) {\n\t\t\t\t\t(lastBlock as any).cache_control = cacheControl;\n\t\t\t\t}\n\t\t\t} else if (typeof lastMessage.content === \"string\") {\n\t\t\t\tlastMessage.content = [\n\t\t\t\t\t{\n\t\t\t\t\t\ttype: \"text\",\n\t\t\t\t\t\ttext: lastMessage.content,\n\t\t\t\t\t\tcache_control: cacheControl,\n\t\t\t\t\t},\n\t\t\t\t] as any;\n\t\t\t}\n\t\t}\n\t}\n\n\treturn { messages: params, assistantLevels };\n}\n\ninterface ConvertedAnthropicMessages {\n\tmessages: MessageParam[];\n\tassistantLevels: Map<number, AnthropicEffort>;\n}\n\nfunction isAnthropicEffort(value: string | undefined): value is AnthropicEffort {\n\treturn value === \"low\" || value === \"medium\" || value === \"high\" || value === \"xhigh\" || value === \"max\";\n}\n\ninterface EffortMessage {\n\trole: \"system\";\n\tcontent: [];\n\toutput_config: { effort: AnthropicEffort };\n}\n\nfunction insertThinkingLevelMessages(\n\tconverted: ConvertedAnthropicMessages,\n\tactiveEffort: AnthropicEffort,\n): Array<MessageParam | EffortMessage> {\n\tconst messages: Array<MessageParam | EffortMessage> = [];\n\tfor (let index = 0; index < converted.messages.length; index++) {\n\t\tconst historicalEffort = converted.assistantLevels.get(index);\n\t\tif (historicalEffort !== undefined) {\n\t\t\tmessages.push({ role: \"system\", content: [], output_config: { effort: historicalEffort } });\n\t\t}\n\t\tmessages.push(converted.messages[index]);\n\t}\n\tmessages.push({ role: \"system\", content: [], output_config: { effort: activeEffort } });\n\treturn messages;\n}\n\nfunction shouldUseFineGrainedToolStreamingBeta(model: Model<\"anthropic-messages\">, context: Context): boolean {\n\treturn !!context.tools?.length && !getAnthropicCompat(model).supportsEagerToolInputStreaming;\n}\n\ntype ToolSchemaObject = {\n\ttype?: string | string[];\n\tproperties?: Record<string, unknown>;\n\trequired?: string[];\n\tanyOf?: ToolSchemaObject[];\n};\n\nfunction projectObjectUnionForAnthropic(\n\tschema: ToolSchemaObject,\n): { properties: Record<string, unknown>; required: string[] } | undefined {\n\tif (schema.properties !== undefined || !Array.isArray(schema.anyOf) || schema.anyOf.length === 0) return undefined;\n\tconst branches = schema.anyOf;\n\tif (\n\t\t!branches.every(\n\t\t\t(branch) =>\n\t\t\t\t(branch.type === undefined ||\n\t\t\t\t\tbranch.type === \"object\" ||\n\t\t\t\t\t(Array.isArray(branch.type) &&\n\t\t\t\t\t\tbranch.type.length > 0 &&\n\t\t\t\t\t\tbranch.type.every((type) => type === \"object\"))) &&\n\t\t\t\tbranch.properties !== undefined,\n\t\t)\n\t)\n\t\treturn undefined;\n\n\tconst variantsByKey = new Map<string, unknown[]>();\n\tfor (const branch of branches) {\n\t\tfor (const [key, propertySchema] of Object.entries(branch.properties ?? {})) {\n\t\t\tconst variants = variantsByKey.get(key);\n\t\t\tif (variants === undefined) {\n\t\t\t\tvariantsByKey.set(key, [propertySchema]);\n\t\t\t\tcontinue;\n\t\t\t}\n\t\t\tconst serialized = JSON.stringify(propertySchema);\n\t\t\tif (!variants.some((variant) => variant === propertySchema || JSON.stringify(variant) === serialized)) {\n\t\t\t\tvariants.push(propertySchema);\n\t\t\t}\n\t\t}\n\t}\n\n\tconst properties: Record<string, unknown> = Object.fromEntries(\n\t\t[...variantsByKey].flatMap(([key, variants]) => {\n\t\t\tconst first = variants[0];\n\t\t\treturn first === undefined ? [] : [[key, variants.length === 1 ? first : { anyOf: variants }]];\n\t\t}),\n\t);\n\n\treturn {\n\t\tproperties,\n\t\trequired: (branches[0]?.required ?? []).filter((key) =>\n\t\t\tbranches.every((branch) => (branch.required ?? []).includes(key)),\n\t\t),\n\t};\n}\n\nfunction convertTools(\n\ttools: Tool[],\n\tisOAuthToken: boolean,\n\tsupportsEagerToolInputStreaming: boolean,\n\tsupportsStrictTools: boolean,\n\tcacheControl?: CacheControlEphemeral,\n\tdeferLoading = false,\n): Anthropic.Messages.Tool[] {\n\tif (!tools) return [];\n\n\treturn tools.map((tool, index) => {\n\t\tconst strict = resolveJsonSchemaStrictSampling(tool, supportsStrictTools);\n\t\tconst parameters = getJsonSchemaToolParameters(tool, strict);\n\t\tconst schema = parameters as ToolSchemaObject;\n\t\t// Anthropic rejects top-level combinators, so project object-union fields for advertising only.\n\t\t// Local runtime validation still enforces the complete authored union without mutating it.\n\t\tconst projectedUnion = projectObjectUnionForAnthropic(schema);\n\t\tconst legacyInputSchema = {\n\t\t\ttype: \"object\" as const,\n\t\t\tproperties: projectedUnion?.properties ?? schema.properties ?? {},\n\t\t\trequired: projectedUnion?.required ?? schema.required ?? [],\n\t\t};\n\t\tconst inputSchema =\n\t\t\tstrict === true && projectedUnion === undefined\n\t\t\t\t? {\n\t\t\t\t\t\t...(parameters as Record<string, unknown>),\n\t\t\t\t\t\t...legacyInputSchema,\n\t\t\t\t\t}\n\t\t\t\t: legacyInputSchema;\n\n\t\treturn {\n\t\t\tname: isOAuthToken ? toClaudeCodeName(tool.name) : tool.name,\n\t\t\tdescription: tool.description,\n\t\t\t...(supportsEagerToolInputStreaming ? { eager_input_streaming: true } : {}),\n\t\t\t...(strict === true ? { strict: true } : {}),\n\t\t\tinput_schema: inputSchema,\n\t\t\t...(deferLoading ? { defer_loading: true } : {}),\n\t\t\t...(cacheControl && index === tools.length - 1 ? { cache_control: cacheControl } : {}),\n\t\t};\n\t});\n}\n\nfunction mapStopReason(\n\treason: Anthropic.Messages.StopReason | string,\n\tstopDetails?: RefusalStopDetails | null,\n): { stopReason: StopReason; errorMessage?: string } {\n\tswitch (reason) {\n\t\tcase \"end_turn\":\n\t\t\treturn { stopReason: \"stop\" };\n\t\tcase \"max_tokens\":\n\t\t\treturn { stopReason: \"length\" };\n\t\tcase \"tool_use\":\n\t\t\treturn { stopReason: \"toolUse\" };\n\t\tcase \"refusal\":\n\t\t\treturn {\n\t\t\t\tstopReason: \"error\",\n\t\t\t\terrorMessage: stopDetails?.explanation || `The model refused to complete the request`,\n\t\t\t};\n\t\tcase \"pause_turn\": // Stop is good enough -> resubmit\n\t\t\treturn { stopReason: \"stop\" };\n\t\tcase \"stop_sequence\":\n\t\t\treturn { stopReason: \"stop\" }; // We don't supply stop sequences, so this should never happen\n\t\tcase \"sensitive\": // Content flagged by safety filters (not yet in SDK types)\n\t\t\treturn { stopReason: \"error\", errorMessage: \"Provider stopped with: sensitive\" };\n\t\tdefault:\n\t\t\t// Handle unknown stop reasons gracefully (API may add new values)\n\t\t\tthrow new Error(`Unhandled stop reason: ${reason}`);\n\t}\n}\n"]}

package/dist/api/anthropic-messages.js

@@ -234,10 +234,12 @@ function addEarlierAttemptCosts(output, model, servingModel, event) {
     }
 }
 function getAnthropicCompat(model) {
+    const isOpenRouter = model.provider === "openrouter" || model.baseUrl.includes("openrouter.ai");
     return {
         supportsEagerToolInputStreaming: model.compat?.supportsEagerToolInputStreaming ?? true,
         supportsLongCacheRetention: model.compat?.supportsLongCacheRetention ?? true,
-        sendSessionAffinityHeaders: model.compat?.sendSessionAffinityHeaders ?? false,
+        sendSessionAffinityHeaders: model.compat?.sendSessionAffinityHeaders ?? isOpenRouter,
+        sessionAffinityFormat: model.compat?.sessionAffinityFormat ?? (isOpenRouter ? "openrouter" : undefined),
         supportsCacheControlOnTools: model.compat?.supportsCacheControlOnTools ?? true,
         supportsTemperature: model.compat?.supportsTemperature ?? true,
         allowEmptySignature: model.compat?.allowEmptySignature ?? false,
@@ -892,7 +894,12 @@ function createClient(model, apiKey, interleavedThinking, useFineGrainedToolStre
         return { client, isOAuthToken: true };
     }
     // API key or header-owned auth.
-    const sessionAffinityHeaders = sessionId && getAnthropicCompat(model).sendSessionAffinityHeaders ? { "x-session-affinity": sessionId } : {};
+    const compat = getAnthropicCompat(model);
+    const sessionAffinityHeaders = {};
+    if (sessionId && compat.sendSessionAffinityHeaders) {
+        const header = compat.sessionAffinityFormat === "openrouter" ? "x-session-id" : "x-session-affinity";
+        sessionAffinityHeaders[header] = sessionId;
+    }
     const defaultHeaders = mergeClientHeaders({
         accept: "application/json",
         "anthropic-dangerous-direct-browser-access": "true",
@@ -1313,6 +1320,40 @@ function insertThinkingLevelMessages(converted, activeEffort) {
 function shouldUseFineGrainedToolStreamingBeta(model, context) {
     return !!context.tools?.length && !getAnthropicCompat(model).supportsEagerToolInputStreaming;
 }
+function projectObjectUnionForAnthropic(schema) {
+    if (schema.properties !== undefined || !Array.isArray(schema.anyOf) || schema.anyOf.length === 0)
+        return undefined;
+    const branches = schema.anyOf;
+    if (!branches.every((branch) => (branch.type === undefined ||
+        branch.type === "object" ||
+        (Array.isArray(branch.type) &&
+            branch.type.length > 0 &&
+            branch.type.every((type) => type === "object"))) &&
+        branch.properties !== undefined))
+        return undefined;
+    const variantsByKey = new Map();
+    for (const branch of branches) {
+        for (const [key, propertySchema] of Object.entries(branch.properties ?? {})) {
+            const variants = variantsByKey.get(key);
+            if (variants === undefined) {
+                variantsByKey.set(key, [propertySchema]);
+                continue;
+            }
+            const serialized = JSON.stringify(propertySchema);
+            if (!variants.some((variant) => variant === propertySchema || JSON.stringify(variant) === serialized)) {
+                variants.push(propertySchema);
+            }
+        }
+    }
+    const properties = Object.fromEntries([...variantsByKey].flatMap(([key, variants]) => {
+        const first = variants[0];
+        return first === undefined ? [] : [[key, variants.length === 1 ? first : { anyOf: variants }]];
+    }));
+    return {
+        properties,
+        required: (branches[0]?.required ?? []).filter((key) => branches.every((branch) => (branch.required ?? []).includes(key))),
+    };
+}
 function convertTools(tools, isOAuthToken, supportsEagerToolInputStreaming, supportsStrictTools, cacheControl, deferLoading = false) {
     if (!tools)
         return [];
@@ -1320,12 +1361,15 @@ function convertTools(tools, isOAuthToken, supportsEagerToolInputStreaming, supp
         const strict = resolveJsonSchemaStrictSampling(tool, supportsStrictTools);
         const parameters = getJsonSchemaToolParameters(tool, strict);
         const schema = parameters;
+        // Anthropic rejects top-level combinators, so project object-union fields for advertising only.
+        // Local runtime validation still enforces the complete authored union without mutating it.
+        const projectedUnion = projectObjectUnionForAnthropic(schema);
         const legacyInputSchema = {
             type: "object",
-            properties: schema.properties ?? {},
-            required: schema.required ?? [],
+            properties: projectedUnion?.properties ?? schema.properties ?? {},
+            required: projectedUnion?.required ?? schema.required ?? [],
         };
-        const inputSchema = strict === true
+        const inputSchema = strict === true && projectedUnion === undefined
             ? {
                 ...parameters,
                 ...legacyInputSchema,