@warlock.js/ai-openai 5.30.0 → 5.31.0

package/esm/config.type.d.mts

@@ -138,9 +138,11 @@ type OpenAIModelConfig = ModelConfig & {
    */
   reasoning?: boolean;
   /**
-   * Stable request bucket used by OpenAI to improve prompt-cache matching.
-   * Sent only by models whose SDK provider label is `"openai"`; compatible
-   * wrapper endpoints deliberately never receive this OpenAI-only field.
+   * Stable request bucket used to improve prompt-cache matching, sent as
+   * `prompt_cache_key`. Sent when the SDK provider label is `"openai"`,
+   * `"xai"` or `"mistral"` — the endpoints that document the field. Other
+   * labels (Groq and DeepSeek cache automatically; custom gateways) never
+   * receive it.
    */
   promptCacheKey?: string;
   /**

package/esm/config.type.d.mts.map

@@ -1 +1 @@
-{"version":3,"file":"config.type.d.mts","names":[],"sources":["../../../../../../../packages/ai-openai/src/config.type.ts"],"mappings":";;;;;AAyCA;;;;;;;;;;;;;;;;AAeuC;AAiBvC;;;;AAAgC;AAahC;;;;AAAqB;AAcrB;;KA3DY,eAAA,GAAkB,IAAA,CAAK,aAAA;EA2DH;;;;;;;;EAlD9B,QAAA;EAwDM;;;;;EAlDN,OAAA,GAAU,MAAA,SAAe,YAAA;AAAA;;;;;;;;;;AA4HpB;AAaP;;;;KAxHY,oBAAA;AAsIZ;;;;AAAgD;AAahD;;;;AAAkD;AAalD;AA1BA,KAzHY,SAAA;;;AAmJoD;;;;;;;;;;KArIpD,iBAAA,GAAoB,WAAA;;;;;;EAM9B,GAAA,GAAM,SAAA;;;;;;;;EAQN,MAAA;;;;;;;;;;;;;;EAcA,cAAA,GAAiB,oBAAA;;;;;;;;;EASjB,gBAAA;;;;;;;;;;;EAWA,SAAA;;;;;;EAMA,cAAA;;;;;;EAMA,oBAAA,GAAuB,OAAA,CACrB,MAAA,CAAO,IAAA,CAAK,WAAA,CAAY,0BAAA;;;;;;;;;EAW1B,GAAA;;;;;;;;EAQA,KAAA;AAAA;;;;;;;;;;;KAaU,oBAAA,GAAuB,cAAc;;;;;;;;;;;;;KAcrC,iBAAA,GAAoB,gBAAgB;;;;;;;;;;;;KAapC,kBAAA,GAAqB,iBAAiB;;;;;;;;;;;;KAatC,yBAAA,GAA4B,wBAAwB"}
+{"version":3,"file":"config.type.d.mts","names":[],"sources":["../../../../../../../packages/ai-openai/src/config.type.ts"],"mappings":";;;;;AAyCA;;;;;;;;;;;;;;;;AAeuC;AAiBvC;;;;AAAgC;AAahC;;;;AAAqB;AAcrB;;KA3DY,eAAA,GAAkB,IAAA,CAAK,aAAA;EA2DH;;;;;;;;EAlD9B,QAAA;EAwDM;;;;;EAlDN,OAAA,GAAU,MAAA,SAAe,YAAA;AAAA;;;;;;;;;;AA8HpB;AAaP;;;;KA1HY,oBAAA;AAwIZ;;;;AAAgD;AAahD;;;;AAAkD;AAalD;AA1BA,KA3HY,SAAA;;;AAqJoD;;;;;;;;;;KAvIpD,iBAAA,GAAoB,WAAA;;;;;;EAM9B,GAAA,GAAM,SAAA;;;;;;;;EAQN,MAAA;;;;;;;;;;;;;;EAcA,cAAA,GAAiB,oBAAA;;;;;;;;;EASjB,gBAAA;;;;;;;;;;;EAWA,SAAA;;;;;;;;EAQA,cAAA;;;;;;EAMA,oBAAA,GAAuB,OAAA,CACrB,MAAA,CAAO,IAAA,CAAK,WAAA,CAAY,0BAAA;;;;;;;;;EAW1B,GAAA;;;;;;;;EAQA,KAAA;AAAA;;;;;;;;;;;KAaU,oBAAA,GAAuB,cAAc;;;;;;;;;;;;;KAcrC,iBAAA,GAAoB,gBAAgB;;;;;;;;;;;;KAapC,kBAAA,GAAqB,iBAAiB;;;;;;;;;;;;KAatC,yBAAA,GAA4B,wBAAwB"}

package/esm/embedder.d.mts.map

@@ -1 +1 @@
-{"version":3,"file":"embedder.d.mts","names":[],"sources":["../../../../../../../packages/ai-openai/src/embedder.ts"],"mappings":";;;;;;;AA2CA;;;;;;;;;;;;;;;;;;;;;;;cAAa,cAAA,YAA0B,gBAAA;EAAA,SACrB,IAAA;EAAA,SACA,QAAA;EACT,UAAA;EAAA,iBAEU,MAAA;EA8CJ;;;;;EAAA,iBAvCI,oBAAA;EAAA,iBACA,MAAA;cAEE,MAAA,EAAQ,MAAA,EAAQ,MAAA,EAAQ,oBAAA;EAO9B,KAAA,CAAM,KAAA,WAAgB,OAAA,CAAQ,eAAA;EA6B9B,SAAA,CAAU,MAAA,aAAmB,OAAA,CAAQ,oBAAA;;;;;;;UAgBpC,OAAA;AAAA"}
+{"version":3,"file":"embedder.d.mts","names":[],"sources":["../../../../../../../packages/ai-openai/src/embedder.ts"],"mappings":";;;;;;;AAyCA;;;;;;;;;;;;;;;;;;;;;;;cAAa,cAAA,YAA0B,gBAAA;EAAA,SACrB,IAAA;EAAA,SACA,QAAA;EACT,UAAA;EAAA,iBAEU,MAAA;EA8CJ;;;;;EAAA,iBAvCI,oBAAA;EAAA,iBACA,MAAA;cAEE,MAAA,EAAQ,MAAA,EAAQ,MAAA,EAAQ,oBAAA;EAO9B,KAAA,CAAM,KAAA,WAAgB,OAAA,CAAQ,eAAA;EA6B9B,SAAA,CAAU,MAAA,aAAmB,OAAA,CAAQ,oBAAA;;;;;;;UAgBpC,OAAA;AAAA"}

package/esm/embedder.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"embedder.mjs","names":[],"sources":["../../../../../../../packages/ai-openai/src/embedder.ts"],"sourcesContent":["import {\n  EmbeddingVectorCountMismatchError,\n  type EmbeddingBatchResult,\n  type EmbeddingResult,\n  type EmbeddingUsage,\n  type EmbedderContract,\n} from \"@warlock.js/ai\";\nimport { log, type Logger } from \"@warlock.js/logger\";\nimport type OpenAI from \"openai\";\nimport type { OpenAIEmbedderConfig } from \"./config.type\";\nimport { wrapOpenAIError } from \"./utils\";\n\nconst LOG_MODULE = \"ai.openai\";\n\ntype EmbeddingsResponse = Awaited<\n  ReturnType<OpenAI[\"embeddings\"][\"create\"]>\n>;\n\n/**\n * OpenAI-backed implementation of `EmbedderContract`.\n *\n * **Role.** Converts text (or a batch of texts) into floating-point\n * vectors via OpenAI's Embeddings API. Standalone primitive — no\n * relationship to chat completions, tools, or the agent loop.\n *\n * **Dimensions.** When no `dimensions` override is supplied in config,\n * `this.dimensions` starts at `0` and is populated from the first\n * response's vector length, then cached for all subsequent calls —\n * even if a later response were to return a different length, the\n * first value wins so batches stay dimensionally consistent. Passing\n * `dimensions` in config both forwards the truncation hint to the API\n * (for models like `text-embedding-3-*`) and sets the initial value.\n *\n * **Error handling.** Raw OpenAI SDK errors are wrapped into the\n * typed `@warlock.js/ai` `AIError` hierarchy via `wrapOpenAIError` —\n * callers catch `AIError` subclasses (`ProviderRateLimitError`,\n * `ProviderAuthError`, etc.) instead of OpenAI's own classes.\n *\n * @example\n * const embedder = new OpenAIEmbedder(client, { name: \"text-embedding-3-small\" });\n * const { vector, dimensions, usage } = await embedder.embed(\"Hello world\");\n * const { vectors } = await embedder.embedMany([\"doc 1\", \"doc 2\"]);\n */\nexport class OpenAIEmbedder implements EmbedderContract {\n  public readonly name: string;\n  public readonly provider = \"openai\";\n  public dimensions: number;\n\n  private readonly client: OpenAI;\n\n  /**\n   * User-specified truncation hint, or `undefined` if omitted.\n   * Forwarded to the API on every call so OpenAI can truncate the\n   * embedding server-side for models that support it.\n   */\n  private readonly configuredDimensions: number | undefined;\n  private readonly logger: Logger = log;\n\n  public constructor(client: OpenAI, config: OpenAIEmbedderConfig) {\n    this.client = client;\n    this.name = config.name;\n    this.configuredDimensions = config.dimensions;\n    this.dimensions = config.dimensions ?? 0;\n  }\n\n  public async embed(input: string): Promise<EmbeddingResult> {\n    const { response, usage } = await this.request(input);\n\n    const [first] = response.data;\n\n    if (first === undefined) {\n      // A 200 whose `data` array is empty. Before this guard the next line\n      // read `.embedding` off `undefined` and the caller got a bare\n      // `TypeError: Cannot read properties of undefined`, naming neither the\n      // provider nor what it actually returned.\n      //\n      // Same typed error the RAG and skill-catalog paths raise for this\n      // contract violation, so one shortfall has one name wherever it\n      // surfaces. Expected 1, received 0.\n      throw new EmbeddingVectorCountMismatchError({\n        provider: this.provider,\n        expectedCount: 1,\n        receivedCount: response.data.length,\n        record: input,\n      });\n    }\n\n    return {\n      vector: first.embedding,\n      dimensions: this.dimensions,\n      usage,\n    };\n  }\n\n  public async embedMany(inputs: string[]): Promise<EmbeddingBatchResult> {\n    const { response, usage } = await this.request(inputs);\n\n    return {\n      vectors: response.data.map((d) => d.embedding),\n      dimensions: this.dimensions,\n      usage,\n    };\n  }\n\n  /**\n   * Shared transport for both `embed()` and `embedMany()` — issues the\n   * `embeddings.create` call, wraps provider errors, caches dimensions\n   * on the first successful response, and returns the raw response\n   * plus a camelCase usage object for the caller to shape.\n   */\n  private async request(input: string | string[]): Promise<{\n    response: EmbeddingsResponse;\n    usage: EmbeddingUsage;\n  }> {\n    this.logger.debug(LOG_MODULE, \"embedder.request\", \"embeddings.create\", {\n      model: this.name,\n      batch: Array.isArray(input),\n      count: Array.isArray(input) ? input.length : 1,\n    });\n\n    let response: EmbeddingsResponse;\n\n    try {\n      response = await this.client.embeddings.create({\n        model: this.name,\n        input,\n        ...(this.configuredDimensions !== undefined\n          ? { dimensions: this.configuredDimensions }\n          : {}),\n      });\n    } catch (thrown) {\n      const wrapped = wrapOpenAIError(thrown);\n\n      this.logger.error(LOG_MODULE, \"embedder.error\", wrapped.message, {\n        code: wrapped.code,\n        context: wrapped.context,\n      });\n\n      throw wrapped;\n    }\n\n    this.logger.debug(LOG_MODULE, \"embedder.response\", \"embeddings.create returned\", {\n      dimensions: response.data[0]?.embedding.length,\n      usage: {\n        promptTokens: response.usage.prompt_tokens,\n        totalTokens: response.usage.total_tokens,\n      },\n    });\n\n    // Cache dimensions on the first response. Once set, stays set —\n    // we trust the first call to define the shape for this embedder.\n    //\n    // Reads the first entry optionally rather than asserting it: this runs for\n    // BOTH embed() and embedMany(), and an empty `data` array is a real\n    // provider response, not an impossible one. Leaving `dimensions` at 0 for\n    // an empty response is correct — the callers each raise their own typed\n    // error about the shortfall, and cacheing a dimension off a response that\n    // carried no vectors would be inventing a shape from nothing.\n    if (this.dimensions === 0) {\n      const firstEmbedding = response.data[0]?.embedding;\n\n      if (firstEmbedding !== undefined) {\n        this.dimensions = firstEmbedding.length;\n      }\n    }\n\n    const usage: EmbeddingUsage = {\n      promptTokens: response.usage.prompt_tokens,\n      totalTokens: response.usage.total_tokens,\n    };\n\n    return { response, usage };\n  }\n}\n"],"mappings":";;;;;;AAYA,MAAM,aAAa;;;;;;;;;;;;;;;;;;;;;;;;;;AA+BnB,IAAa,iBAAb,MAAwD;CAetD,AAAO,YAAY,QAAgB,QAA8B;kBAbtC;gBAWO;EAGhC,KAAK,SAAS;EACd,KAAK,OAAO,OAAO;EACnB,KAAK,uBAAuB,OAAO;EACnC,KAAK,aAAa,OAAO,cAAc;CACzC;CAEA,MAAa,MAAM,OAAyC;EAC1D,MAAM,EAAE,UAAU,UAAU,MAAM,KAAK,QAAQ,KAAK;EAEpD,MAAM,CAAC,SAAS,SAAS;EAEzB,IAAI,UAAU,QASZ,MAAM,IAAI,kCAAkC;GAC1C,UAAU,KAAK;GACf,eAAe;GACf,eAAe,SAAS,KAAK;GAC7B,QAAQ;EACV,CAAC;EAGH,OAAO;GACL,QAAQ,MAAM;GACd,YAAY,KAAK;GACjB;EACF;CACF;CAEA,MAAa,UAAU,QAAiD;EACtE,MAAM,EAAE,UAAU,UAAU,MAAM,KAAK,QAAQ,MAAM;EAErD,OAAO;GACL,SAAS,SAAS,KAAK,KAAK,MAAM,EAAE,SAAS;GAC7C,YAAY,KAAK;GACjB;EACF;CACF;;;;;;;CAQA,MAAc,QAAQ,OAGnB;EACD,KAAK,OAAO,MAAM,YAAY,oBAAoB,qBAAqB;GACrE,OAAO,KAAK;GACZ,OAAO,MAAM,QAAQ,KAAK;GAC1B,OAAO,MAAM,QAAQ,KAAK,IAAI,MAAM,SAAS;EAC/C,CAAC;EAED,IAAI;EAEJ,IAAI;GACF,WAAW,MAAM,KAAK,OAAO,WAAW,OAAO;IAC7C,OAAO,KAAK;IACZ;IACA,GAAI,KAAK,yBAAyB,SAC9B,EAAE,YAAY,KAAK,qBAAqB,IACxC,CAAC;GACP,CAAC;EACH,SAAS,QAAQ;GACf,MAAM,UAAU,gBAAgB,MAAM;GAEtC,KAAK,OAAO,MAAM,YAAY,kBAAkB,QAAQ,SAAS;IAC/D,MAAM,QAAQ;IACd,SAAS,QAAQ;GACnB,CAAC;GAED,MAAM;EACR;EAEA,KAAK,OAAO,MAAM,YAAY,qBAAqB,8BAA8B;GAC/E,YAAY,SAAS,KAAK,EAAE,EAAE,UAAU;GACxC,OAAO;IACL,cAAc,SAAS,MAAM;IAC7B,aAAa,SAAS,MAAM;GAC9B;EACF,CAAC;EAWD,IAAI,KAAK,eAAe,GAAG;GACzB,MAAM,iBAAiB,SAAS,KAAK,EAAE,EAAE;GAEzC,IAAI,mBAAmB,QACrB,KAAK,aAAa,eAAe;EAErC;EAEA,MAAM,QAAwB;GAC5B,cAAc,SAAS,MAAM;GAC7B,aAAa,SAAS,MAAM;EAC9B;EAEA,OAAO;GAAE;GAAU;EAAM;CAC3B;AACF"}
+{"version":3,"file":"embedder.mjs","names":[],"sources":["../../../../../../../packages/ai-openai/src/embedder.ts"],"sourcesContent":["import {\n  EmbeddingVectorCountMismatchError,\n  type EmbeddingBatchResult,\n  type EmbeddingResult,\n  type EmbeddingUsage,\n  type EmbedderContract,\n} from \"@warlock.js/ai\";\nimport { log, type Logger } from \"@warlock.js/logger\";\nimport type OpenAI from \"openai\";\nimport type { OpenAIEmbedderConfig } from \"./config.type\";\nimport { wrapOpenAIError } from \"./utils\";\n\nconst LOG_MODULE = \"ai.openai\";\n\ntype EmbeddingsResponse = Awaited<ReturnType<OpenAI[\"embeddings\"][\"create\"]>>;\n\n/**\n * OpenAI-backed implementation of `EmbedderContract`.\n *\n * **Role.** Converts text (or a batch of texts) into floating-point\n * vectors via OpenAI's Embeddings API. Standalone primitive — no\n * relationship to chat completions, tools, or the agent loop.\n *\n * **Dimensions.** When no `dimensions` override is supplied in config,\n * `this.dimensions` starts at `0` and is populated from the first\n * response's vector length, then cached for all subsequent calls —\n * even if a later response were to return a different length, the\n * first value wins so batches stay dimensionally consistent. Passing\n * `dimensions` in config both forwards the truncation hint to the API\n * (for models like `text-embedding-3-*`) and sets the initial value.\n *\n * **Error handling.** Raw OpenAI SDK errors are wrapped into the\n * typed `@warlock.js/ai` `AIError` hierarchy via `wrapOpenAIError` —\n * callers catch `AIError` subclasses (`ProviderRateLimitError`,\n * `ProviderAuthError`, etc.) instead of OpenAI's own classes.\n *\n * @example\n * const embedder = new OpenAIEmbedder(client, { name: \"text-embedding-3-small\" });\n * const { vector, dimensions, usage } = await embedder.embed(\"Hello world\");\n * const { vectors } = await embedder.embedMany([\"doc 1\", \"doc 2\"]);\n */\nexport class OpenAIEmbedder implements EmbedderContract {\n  public readonly name: string;\n  public readonly provider = \"openai\";\n  public dimensions: number;\n\n  private readonly client: OpenAI;\n\n  /**\n   * User-specified truncation hint, or `undefined` if omitted.\n   * Forwarded to the API on every call so OpenAI can truncate the\n   * embedding server-side for models that support it.\n   */\n  private readonly configuredDimensions: number | undefined;\n  private readonly logger: Logger = log;\n\n  public constructor(client: OpenAI, config: OpenAIEmbedderConfig) {\n    this.client = client;\n    this.name = config.name;\n    this.configuredDimensions = config.dimensions;\n    this.dimensions = config.dimensions ?? 0;\n  }\n\n  public async embed(input: string): Promise<EmbeddingResult> {\n    const { response, usage } = await this.request(input);\n\n    const [first] = response.data;\n\n    if (first === undefined) {\n      // A 200 whose `data` array is empty. Before this guard the next line\n      // read `.embedding` off `undefined` and the caller got a bare\n      // `TypeError: Cannot read properties of undefined`, naming neither the\n      // provider nor what it actually returned.\n      //\n      // Same typed error the RAG and skill-catalog paths raise for this\n      // contract violation, so one shortfall has one name wherever it\n      // surfaces. Expected 1, received 0.\n      throw new EmbeddingVectorCountMismatchError({\n        provider: this.provider,\n        expectedCount: 1,\n        receivedCount: response.data.length,\n        record: input,\n      });\n    }\n\n    return {\n      vector: first.embedding,\n      dimensions: this.dimensions,\n      usage,\n    };\n  }\n\n  public async embedMany(inputs: string[]): Promise<EmbeddingBatchResult> {\n    const { response, usage } = await this.request(inputs);\n\n    return {\n      vectors: response.data.map((d) => d.embedding),\n      dimensions: this.dimensions,\n      usage,\n    };\n  }\n\n  /**\n   * Shared transport for both `embed()` and `embedMany()` — issues the\n   * `embeddings.create` call, wraps provider errors, caches dimensions\n   * on the first successful response, and returns the raw response\n   * plus a camelCase usage object for the caller to shape.\n   */\n  private async request(input: string | string[]): Promise<{\n    response: EmbeddingsResponse;\n    usage: EmbeddingUsage;\n  }> {\n    this.logger.debug(LOG_MODULE, \"embedder.request\", \"embeddings.create\", {\n      model: this.name,\n      batch: Array.isArray(input),\n      count: Array.isArray(input) ? input.length : 1,\n    });\n\n    let response: EmbeddingsResponse;\n\n    try {\n      response = await this.client.embeddings.create({\n        model: this.name,\n        input,\n        ...(this.configuredDimensions !== undefined\n          ? { dimensions: this.configuredDimensions }\n          : {}),\n      });\n    } catch (thrown) {\n      const wrapped = wrapOpenAIError(thrown);\n\n      this.logger.error(LOG_MODULE, \"embedder.error\", wrapped.message, {\n        code: wrapped.code,\n        context: wrapped.context,\n      });\n\n      throw wrapped;\n    }\n\n    this.logger.debug(LOG_MODULE, \"embedder.response\", \"embeddings.create returned\", {\n      dimensions: response.data[0]?.embedding.length,\n      usage: {\n        promptTokens: response.usage.prompt_tokens,\n        totalTokens: response.usage.total_tokens,\n      },\n    });\n\n    // Cache dimensions on the first response. Once set, stays set —\n    // we trust the first call to define the shape for this embedder.\n    //\n    // Reads the first entry optionally rather than asserting it: this runs for\n    // BOTH embed() and embedMany(), and an empty `data` array is a real\n    // provider response, not an impossible one. Leaving `dimensions` at 0 for\n    // an empty response is correct — the callers each raise their own typed\n    // error about the shortfall, and cacheing a dimension off a response that\n    // carried no vectors would be inventing a shape from nothing.\n    if (this.dimensions === 0) {\n      const firstEmbedding = response.data[0]?.embedding;\n\n      if (firstEmbedding !== undefined) {\n        this.dimensions = firstEmbedding.length;\n      }\n    }\n\n    const usage: EmbeddingUsage = {\n      promptTokens: response.usage.prompt_tokens,\n      totalTokens: response.usage.total_tokens,\n    };\n\n    return { response, usage };\n  }\n}\n"],"mappings":";;;;;;AAYA,MAAM,aAAa;;;;;;;;;;;;;;;;;;;;;;;;;;AA6BnB,IAAa,iBAAb,MAAwD;CAetD,AAAO,YAAY,QAAgB,QAA8B;kBAbtC;gBAWO;EAGhC,KAAK,SAAS;EACd,KAAK,OAAO,OAAO;EACnB,KAAK,uBAAuB,OAAO;EACnC,KAAK,aAAa,OAAO,cAAc;CACzC;CAEA,MAAa,MAAM,OAAyC;EAC1D,MAAM,EAAE,UAAU,UAAU,MAAM,KAAK,QAAQ,KAAK;EAEpD,MAAM,CAAC,SAAS,SAAS;EAEzB,IAAI,UAAU,QASZ,MAAM,IAAI,kCAAkC;GAC1C,UAAU,KAAK;GACf,eAAe;GACf,eAAe,SAAS,KAAK;GAC7B,QAAQ;EACV,CAAC;EAGH,OAAO;GACL,QAAQ,MAAM;GACd,YAAY,KAAK;GACjB;EACF;CACF;CAEA,MAAa,UAAU,QAAiD;EACtE,MAAM,EAAE,UAAU,UAAU,MAAM,KAAK,QAAQ,MAAM;EAErD,OAAO;GACL,SAAS,SAAS,KAAK,KAAK,MAAM,EAAE,SAAS;GAC7C,YAAY,KAAK;GACjB;EACF;CACF;;;;;;;CAQA,MAAc,QAAQ,OAGnB;EACD,KAAK,OAAO,MAAM,YAAY,oBAAoB,qBAAqB;GACrE,OAAO,KAAK;GACZ,OAAO,MAAM,QAAQ,KAAK;GAC1B,OAAO,MAAM,QAAQ,KAAK,IAAI,MAAM,SAAS;EAC/C,CAAC;EAED,IAAI;EAEJ,IAAI;GACF,WAAW,MAAM,KAAK,OAAO,WAAW,OAAO;IAC7C,OAAO,KAAK;IACZ;IACA,GAAI,KAAK,yBAAyB,SAC9B,EAAE,YAAY,KAAK,qBAAqB,IACxC,CAAC;GACP,CAAC;EACH,SAAS,QAAQ;GACf,MAAM,UAAU,gBAAgB,MAAM;GAEtC,KAAK,OAAO,MAAM,YAAY,kBAAkB,QAAQ,SAAS;IAC/D,MAAM,QAAQ;IACd,SAAS,QAAQ;GACnB,CAAC;GAED,MAAM;EACR;EAEA,KAAK,OAAO,MAAM,YAAY,qBAAqB,8BAA8B;GAC/E,YAAY,SAAS,KAAK,EAAE,EAAE,UAAU;GACxC,OAAO;IACL,cAAc,SAAS,MAAM;IAC7B,aAAa,SAAS,MAAM;GAC9B;EACF,CAAC;EAWD,IAAI,KAAK,eAAe,GAAG;GACzB,MAAM,iBAAiB,SAAS,KAAK,EAAE,EAAE;GAEzC,IAAI,mBAAmB,QACrB,KAAK,aAAa,eAAe;EAErC;EAEA,MAAM,QAAwB;GAC5B,cAAc,SAAS,MAAM;GAC7B,aAAa,SAAS,MAAM;EAC9B;EAEA,OAAO;GAAE;GAAU;EAAM;CAC3B;AACF"}

package/esm/index.d.mts

@@ -1,7 +1,9 @@
 import { OpenAIEmbedderConfig, OpenAIImageConfig, OpenAISDKConfig, OpenAISpeechConfig, OpenAITranscriptionConfig } from "./config.type.mjs";
 import { OpenAISDK } from "./sdk.mjs";
+import { OPENAI_MODEL_PRICING, openAIPricingFor } from "./known-pricing.mjs";
 import { OpenAIEmbedder } from "./embedder.mjs";
 import { OpenAIImageModel } from "./image.mjs";
 import { OpenAISpeechModel } from "./speech.mjs";
 import { OpenAITranscriptionModel } from "./transcription.mjs";
-export { OpenAIEmbedder, type OpenAIEmbedderConfig, type OpenAIImageConfig, OpenAIImageModel, OpenAISDK, type OpenAISDKConfig, type OpenAISpeechConfig, OpenAISpeechModel, type OpenAITranscriptionConfig, OpenAITranscriptionModel };
+import { OpenAICallOptions } from "./provider-options.mjs";
+export { OPENAI_MODEL_PRICING, type OpenAICallOptions, OpenAIEmbedder, type OpenAIEmbedderConfig, type OpenAIImageConfig, OpenAIImageModel, OpenAISDK, type OpenAISDKConfig, type OpenAISpeechConfig, OpenAISpeechModel, type OpenAITranscriptionConfig, OpenAITranscriptionModel, openAIPricingFor };

package/esm/index.mjs

@@ -1,7 +1,8 @@
+import { OPENAI_MODEL_PRICING, openAIPricingFor } from "./known-pricing.mjs";
 import { OpenAIEmbedder } from "./embedder.mjs";
 import { OpenAIImageModel } from "./image.mjs";
 import { OpenAISpeechModel } from "./speech.mjs";
 import { OpenAITranscriptionModel } from "./transcription.mjs";
 import { OpenAISDK } from "./sdk.mjs";
 
-export { OpenAIEmbedder, OpenAIImageModel, OpenAISDK, OpenAISpeechModel, OpenAITranscriptionModel };
+export { OPENAI_MODEL_PRICING, OpenAIEmbedder, OpenAIImageModel, OpenAISDK, OpenAISpeechModel, OpenAITranscriptionModel, openAIPricingFor };

package/esm/known-pricing.d.mts

@@ -0,0 +1,35 @@
+import { ModelPricing } from "@warlock.js/ai";
+
+//#region ../../packages/ai-openai/src/known-pricing.d.ts
+/**
+ * Built-in OpenAI API list prices, USD per 1,000,000 tokens, keyed by
+ * exact model id. `cachedInput` is the cached-input rate. None of these
+ * models charges for cache writes.
+ *
+ * Source: https://developers.openai.com/api/docs/pricing (the target of
+ * platform.openai.com/docs/pricing), Standard tier, fetched 2026-10-17.
+ *
+ * Applied only when the SDK's provider label is `"openai"`, and only on
+ * an exact id match — dated snapshots can carry other prices. Left out on
+ * purpose, so pass `pricing` yourself for them:
+ *
+ * - Models priced by prompt length (a higher rate above 272K input
+ *   tokens): the gpt-6 / gpt-5.6 / gpt-5.5 / gpt-5.4 families and the
+ *   `-pro` models. A flat `ModelPricing` cannot express the tiers.
+ * - Promotional, Batch, Flex and Priority pricing, and the regional
+ *   (data-residency) uplift.
+ */
+declare const OPENAI_MODEL_PRICING: Readonly<Record<string, ModelPricing>>;
+/**
+ * Look up the built-in list price for an OpenAI model id (exact match).
+ * Returns `undefined` for unknown ids, so cost stays unknown rather than
+ * wrong.
+ *
+ * @example
+ * openAIPricingFor("gpt-4o-mini"); // { input: 0.15, output: 0.6, cachedInput: 0.075 }
+ * openAIPricingFor("gpt-5.5"); // undefined — priced by prompt length
+ */
+declare function openAIPricingFor(modelName: string): ModelPricing | undefined;
+//#endregion
+export { OPENAI_MODEL_PRICING, openAIPricingFor };
+//# sourceMappingURL=known-pricing.d.mts.map

package/esm/known-pricing.d.mts.map

@@ -0,0 +1 @@
+{"version":3,"file":"known-pricing.d.mts","names":[],"sources":["../../../../../../../packages/ai-openai/src/known-pricing.ts"],"mappings":";;;;;AAoBA;;;;;;;;;;;;AAAuE;AA4BvE;;;cA5Ba,oBAAA,EAAsB,QAAA,CAAS,MAAA,SAAe,YAAA;AA4BM;;;;;;;;;AAAA,iBAAjD,gBAAA,CAAiB,SAAA,WAAoB,YAAY"}

package/esm/known-pricing.mjs

@@ -0,0 +1,117 @@
+//#region ../../packages/ai-openai/src/known-pricing.ts
+/**
+* Built-in OpenAI API list prices, USD per 1,000,000 tokens, keyed by
+* exact model id. `cachedInput` is the cached-input rate. None of these
+* models charges for cache writes.
+*
+* Source: https://developers.openai.com/api/docs/pricing (the target of
+* platform.openai.com/docs/pricing), Standard tier, fetched 2026-10-17.
+*
+* Applied only when the SDK's provider label is `"openai"`, and only on
+* an exact id match — dated snapshots can carry other prices. Left out on
+* purpose, so pass `pricing` yourself for them:
+*
+* - Models priced by prompt length (a higher rate above 272K input
+*   tokens): the gpt-6 / gpt-5.6 / gpt-5.5 / gpt-5.4 families and the
+*   `-pro` models. A flat `ModelPricing` cannot express the tiers.
+* - Promotional, Batch, Flex and Priority pricing, and the regional
+*   (data-residency) uplift.
+*/
+const OPENAI_MODEL_PRICING = {
+	"gpt-5.4-mini": {
+		input: .75,
+		output: 4.5,
+		cachedInput: .075
+	},
+	"gpt-5.4-nano": {
+		input: .2,
+		output: 1.25,
+		cachedInput: .02
+	},
+	"gpt-5.2": {
+		input: 1.75,
+		output: 14,
+		cachedInput: .175
+	},
+	"gpt-5.1": {
+		input: 1.25,
+		output: 10,
+		cachedInput: .125
+	},
+	"gpt-5": {
+		input: 1.25,
+		output: 10,
+		cachedInput: .125
+	},
+	"gpt-5-mini": {
+		input: .25,
+		output: 2,
+		cachedInput: .025
+	},
+	"gpt-5-nano": {
+		input: .05,
+		output: .4,
+		cachedInput: .005
+	},
+	"gpt-4.1": {
+		input: 2,
+		output: 8,
+		cachedInput: .5
+	},
+	"gpt-4.1-mini": {
+		input: .4,
+		output: 1.6,
+		cachedInput: .1
+	},
+	"gpt-4.1-nano": {
+		input: .1,
+		output: .4,
+		cachedInput: .025
+	},
+	"gpt-4o": {
+		input: 2.5,
+		output: 10,
+		cachedInput: 1.25
+	},
+	"gpt-4o-mini": {
+		input: .15,
+		output: .6,
+		cachedInput: .075
+	},
+	o3: {
+		input: 2,
+		output: 8,
+		cachedInput: .5
+	},
+	"o4-mini": {
+		input: 1.1,
+		output: 4.4,
+		cachedInput: .275
+	},
+	"o3-mini": {
+		input: 1.1,
+		output: 4.4,
+		cachedInput: .55
+	},
+	o1: {
+		input: 15,
+		output: 60,
+		cachedInput: 7.5
+	}
+};
+/**
+* Look up the built-in list price for an OpenAI model id (exact match).
+* Returns `undefined` for unknown ids, so cost stays unknown rather than
+* wrong.
+*
+* @example
+* openAIPricingFor("gpt-4o-mini"); // { input: 0.15, output: 0.6, cachedInput: 0.075 }
+* openAIPricingFor("gpt-5.5"); // undefined — priced by prompt length
+*/
+function openAIPricingFor(modelName) {
+	return OPENAI_MODEL_PRICING[modelName];
+}
+
+//#endregion
+export { OPENAI_MODEL_PRICING, openAIPricingFor };
+//# sourceMappingURL=known-pricing.mjs.map

package/esm/known-pricing.mjs.map

@@ -0,0 +1 @@
+{"version":3,"file":"known-pricing.mjs","names":[],"sources":["../../../../../../../packages/ai-openai/src/known-pricing.ts"],"sourcesContent":["import type { ModelPricing } from \"@warlock.js/ai\";\n\n/**\n * Built-in OpenAI API list prices, USD per 1,000,000 tokens, keyed by\n * exact model id. `cachedInput` is the cached-input rate. None of these\n * models charges for cache writes.\n *\n * Source: https://developers.openai.com/api/docs/pricing (the target of\n * platform.openai.com/docs/pricing), Standard tier, fetched 2026-10-17.\n *\n * Applied only when the SDK's provider label is `\"openai\"`, and only on\n * an exact id match — dated snapshots can carry other prices. Left out on\n * purpose, so pass `pricing` yourself for them:\n *\n * - Models priced by prompt length (a higher rate above 272K input\n *   tokens): the gpt-6 / gpt-5.6 / gpt-5.5 / gpt-5.4 families and the\n *   `-pro` models. A flat `ModelPricing` cannot express the tiers.\n * - Promotional, Batch, Flex and Priority pricing, and the regional\n *   (data-residency) uplift.\n */\nexport const OPENAI_MODEL_PRICING: Readonly<Record<string, ModelPricing>> = {\n  \"gpt-5.4-mini\": { input: 0.75, output: 4.5, cachedInput: 0.075 },\n  \"gpt-5.4-nano\": { input: 0.2, output: 1.25, cachedInput: 0.02 },\n  \"gpt-5.2\": { input: 1.75, output: 14, cachedInput: 0.175 },\n  \"gpt-5.1\": { input: 1.25, output: 10, cachedInput: 0.125 },\n  \"gpt-5\": { input: 1.25, output: 10, cachedInput: 0.125 },\n  \"gpt-5-mini\": { input: 0.25, output: 2, cachedInput: 0.025 },\n  \"gpt-5-nano\": { input: 0.05, output: 0.4, cachedInput: 0.005 },\n  \"gpt-4.1\": { input: 2, output: 8, cachedInput: 0.5 },\n  \"gpt-4.1-mini\": { input: 0.4, output: 1.6, cachedInput: 0.1 },\n  \"gpt-4.1-nano\": { input: 0.1, output: 0.4, cachedInput: 0.025 },\n  \"gpt-4o\": { input: 2.5, output: 10, cachedInput: 1.25 },\n  \"gpt-4o-mini\": { input: 0.15, output: 0.6, cachedInput: 0.075 },\n  o3: { input: 2, output: 8, cachedInput: 0.5 },\n  \"o4-mini\": { input: 1.1, output: 4.4, cachedInput: 0.275 },\n  \"o3-mini\": { input: 1.1, output: 4.4, cachedInput: 0.55 },\n  o1: { input: 15, output: 60, cachedInput: 7.5 },\n};\n\n/**\n * Look up the built-in list price for an OpenAI model id (exact match).\n * Returns `undefined` for unknown ids, so cost stays unknown rather than\n * wrong.\n *\n * @example\n * openAIPricingFor(\"gpt-4o-mini\"); // { input: 0.15, output: 0.6, cachedInput: 0.075 }\n * openAIPricingFor(\"gpt-5.5\"); // undefined — priced by prompt length\n */\nexport function openAIPricingFor(modelName: string): ModelPricing | undefined {\n  return OPENAI_MODEL_PRICING[modelName];\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;AAoBA,MAAa,uBAA+D;CAC1E,gBAAgB;EAAE,OAAO;EAAM,QAAQ;EAAK,aAAa;CAAM;CAC/D,gBAAgB;EAAE,OAAO;EAAK,QAAQ;EAAM,aAAa;CAAK;CAC9D,WAAW;EAAE,OAAO;EAAM,QAAQ;EAAI,aAAa;CAAM;CACzD,WAAW;EAAE,OAAO;EAAM,QAAQ;EAAI,aAAa;CAAM;CACzD,SAAS;EAAE,OAAO;EAAM,QAAQ;EAAI,aAAa;CAAM;CACvD,cAAc;EAAE,OAAO;EAAM,QAAQ;EAAG,aAAa;CAAM;CAC3D,cAAc;EAAE,OAAO;EAAM,QAAQ;EAAK,aAAa;CAAM;CAC7D,WAAW;EAAE,OAAO;EAAG,QAAQ;EAAG,aAAa;CAAI;CACnD,gBAAgB;EAAE,OAAO;EAAK,QAAQ;EAAK,aAAa;CAAI;CAC5D,gBAAgB;EAAE,OAAO;EAAK,QAAQ;EAAK,aAAa;CAAM;CAC9D,UAAU;EAAE,OAAO;EAAK,QAAQ;EAAI,aAAa;CAAK;CACtD,eAAe;EAAE,OAAO;EAAM,QAAQ;EAAK,aAAa;CAAM;CAC9D,IAAI;EAAE,OAAO;EAAG,QAAQ;EAAG,aAAa;CAAI;CAC5C,WAAW;EAAE,OAAO;EAAK,QAAQ;EAAK,aAAa;CAAM;CACzD,WAAW;EAAE,OAAO;EAAK,QAAQ;EAAK,aAAa;CAAK;CACxD,IAAI;EAAE,OAAO;EAAI,QAAQ;EAAI,aAAa;CAAI;AAChD;;;;;;;;;;AAWA,SAAgB,iBAAiB,WAA6C;CAC5E,OAAO,qBAAqB;AAC9B"}

package/esm/model.mjs

@@ -7,9 +7,10 @@ import { inferStructuredOutput, planStructuredOutput } from "./utils/structured-
 import { toOpenAIMessages } from "./utils/to-openai-messages.mjs";
 import { toOpenAITools } from "./utils/to-openai-tools.mjs";
 import "./utils/index.mjs";
+import { mergeProviderBody } from "./provider-options.mjs";
 import { inferReasoningCapability } from "./known-reasoning-models.mjs";
 import { inferVisionCapability } from "./known-vision-models.mjs";
-import { InvalidRequestError, ProviderError, safeJsonParse } from "@warlock.js/ai";
+import { InvalidRequestError, ProviderError, resolveStreamChunks, safeJsonParse, selectProviderTools } from "@warlock.js/ai";
 import { log } from "@warlock.js/logger";
 
 //#region ../../packages/ai-openai/src/model.ts
@@ -88,15 +89,15 @@ var OpenAIModel = class {
 		});
 		let response;
 		try {
-			response = await this.client.chat.completions.create({
+			response = await this.client.chat.completions.create(mergeProviderBody({
 				model: this.name,
 				messages: toOpenAIMessages(messages, this.promptCacheBreakpoints(options)),
 				...this.buildSamplingParams(options),
 				...buildPromptCacheParams(this.provider, this.config),
-				tools: toOpenAITools(options?.tools),
+				...this.buildTools(options),
 				...this.buildResponseFormat(options?.responseSchema),
 				...this.buildReasoningParams(options?.reasoning, Boolean(options?.tools?.length))
-			}, options?.signal ? { signal: options.signal } : void 0);
+			}, options), options?.signal ? { signal: options.signal } : void 0);
 		} catch (thrown) {
 			const wrapped = wrapOpenAIError(thrown);
 			this.logger.error(LOG_MODULE, "error", wrapped.message, {
@@ -129,6 +130,11 @@ var OpenAIModel = class {
 	* `delta` for text tokens, `tool-call` when the model requests a tool,
 	* and a terminal `done` carrying the final finish reason + usage totals.
 	* Callers consume it with `for await`.
+	*
+	* With `options.streamChunks` it also yields `tool-args` for each
+	* `tool_calls[].function.arguments` fragment and `reasoning` for a
+	* vendor's `reasoning_content` / `reasoning` delta field. Usage arrives
+	* only on the final chunk here, so no `usage` chunk is emitted.
 	*/
 	async *stream(messages, options) {
 		this.logger.debug(LOG_MODULE, "request", "Starting streaming call to chat.completions", {
@@ -139,17 +145,17 @@ var OpenAIModel = class {
 		});
 		let stream;
 		try {
-			stream = await this.client.chat.completions.create({
+			stream = await this.client.chat.completions.create(mergeProviderBody({
 				model: this.name,
 				messages: toOpenAIMessages(messages, this.promptCacheBreakpoints(options)),
 				...this.buildSamplingParams(options),
 				...buildPromptCacheParams(this.provider, this.config),
-				tools: toOpenAITools(options?.tools),
+				...this.buildTools(options),
 				stream: true,
 				stream_options: { include_usage: true },
 				...this.buildResponseFormat(options?.responseSchema),
 				...this.buildReasoningParams(options?.reasoning, Boolean(options?.tools?.length))
-			}, options?.signal ? { signal: options.signal } : void 0);
+			}, options), options?.signal ? { signal: options.signal } : void 0);
 		} catch (thrown) {
 			const wrapped = wrapOpenAIError(thrown);
 			this.logger.error(LOG_MODULE, "error", wrapped.message, {
@@ -159,16 +165,24 @@ var OpenAIModel = class {
 			throw wrapped;
 		}
 		let rawFinishReason = "stop";
-		const usage = {
+		let usage = {
 			input: 0,
 			output: 0,
 			total: 0
 		};
+		const want = resolveStreamChunks(options);
 		const toolCallAccum = /* @__PURE__ */ new Map();
 		try {
 			for await (const chunk of stream) {
 				const delta = chunk.choices[0]?.delta;
 				const finish = chunk.choices[0]?.finish_reason;
+				if (want.reasoning && delta) {
+					const reasoning = readReasoningDelta(delta);
+					if (reasoning) yield {
+						type: "reasoning",
+						delta: reasoning
+					};
+				}
 				if (delta?.content) yield {
 					type: "delta",
 					content: delta.content
@@ -178,23 +192,25 @@ var OpenAIModel = class {
 					if (!toolCallAccum.has(idx)) toolCallAccum.set(idx, {
 						id: "",
 						name: "",
-						arguments: ""
+						arguments: "",
+						streamed: 0
 					});
 					const acc = toolCallAccum.get(idx);
 					if (toolCall.id) acc.id = toolCall.id;
 					if (toolCall.function?.name) acc.name = toolCall.function.name;
 					if (toolCall.function?.arguments) acc.arguments += toolCall.function.arguments;
+					if (want.toolArgs && acc.name && acc.arguments.length > acc.streamed) {
+						yield {
+							type: "tool-args",
+							callId: acc.id,
+							name: acc.name,
+							delta: acc.arguments.slice(acc.streamed)
+						};
+						acc.streamed = acc.arguments.length;
+					}
 				}
 				if (finish) rawFinishReason = finish;
-				if (chunk.usage) {
-					usage.input = chunk.usage.prompt_tokens ?? 0;
-					usage.output = chunk.usage.completion_tokens ?? 0;
-					usage.total = chunk.usage.total_tokens ?? 0;
-					const cached = chunk.usage.prompt_tokens_details?.cached_tokens;
-					if (cached !== void 0 && cached > 0) usage.cachedTokens = cached;
-					const reasoning = chunk.usage.completion_tokens_details?.reasoning_tokens;
-					if (reasoning !== void 0 && reasoning > 0) usage.reasoningTokens = reasoning;
-				}
+				if (chunk.usage) usage = this.extractUsage(chunk.usage);
 			}
 			for (const acc of toolCallAccum.values()) {
 				if (!acc.name) continue;
@@ -224,6 +240,25 @@ var OpenAIModel = class {
 			usage
 		};
 	}
+	/**
+	* Spread-friendly `tools` fragment: the function tools, then the
+	* provider-defined entries whose `provider` equals this model's
+	* provider label (`"openai"`, or the vendor label an OpenAI-compatible
+	* wrapper sets, e.g. `"groq"`), sent verbatim. Chat Completions only
+	* accepts `function` and `custom` tool types from OpenAI itself; its
+	* hosted tools (`web_search`, `file_search`, …) need `api: "responses"`
+	* and OpenAI rejects them here with a 400.
+	*/
+	buildTools(options) {
+		const { tools, definitions } = selectProviderTools(options, {
+			accepts: [this.provider],
+			module: LOG_MODULE,
+			model: this.name,
+			logger: this.logger
+		});
+		const mapped = [...toOpenAITools(tools) ?? [], ...definitions];
+		return mapped.length > 0 ? { tools: mapped } : {};
+	}
 	/** Map the neutral cache hint only for the direct OpenAI provider. */
 	promptCacheBreakpoints(options) {
 		return this.provider === "openai" ? options?.cacheControl?.breakpoints ?? 0 : 0;
@@ -262,7 +297,9 @@ var OpenAIModel = class {
 	* (the hidden reasoning channel on o-series / gpt-5 models, already
 	* counted within `output`). Both are emitted only when the provider
 	* reports a positive value, so non-reasoning / uncached calls keep the
-	* lean `{ input, output, total }` shape.
+	* lean `{ input, output, total }` shape. `cacheWriteTokens` mirrors
+	* `prompt_tokens_details.cache_write_tokens` (billed writes on the
+	* models that charge for them). `cost` is attached when priced.
 	*/
 	extractUsage(raw) {
 		if (!raw) return {
@@ -271,12 +308,13 @@ var OpenAIModel = class {
 			total: 0
 		};
 		return buildUsage({
-			input: raw.prompt_tokens,
-			output: raw.completion_tokens,
-			total: raw.total_tokens,
+			input: raw.prompt_tokens ?? 0,
+			output: raw.completion_tokens ?? 0,
+			total: raw.total_tokens ?? 0,
 			cachedTokens: raw.prompt_tokens_details?.cached_tokens,
+			cacheWriteTokens: raw.prompt_tokens_details?.cache_write_tokens,
 			reasoningTokens: raw.completion_tokens_details?.reasoning_tokens
-		});
+		}, this.pricing);
 	}
 	/**
 	* Translate the neutral `ModelCallOptions.reasoning` hint into OpenAI's
@@ -352,6 +390,20 @@ var OpenAIModel = class {
 		}));
 	}
 };
+/**
+* Reasoning text an OpenAI-compatible vendor puts on a Chat Completions
+* stream delta. OpenAI itself streams none on this API. DeepSeek documents
+* `delta.reasoning_content` (API reference, chat completion chunk); Groq
+* documents the parsed reasoning under a `reasoning` field. Neither is in
+* the OpenAI SDK's delta type, so the delta is read as a plain record.
+*/
+function readReasoningDelta(delta) {
+	const record = delta;
+	for (const key of ["reasoning_content", "reasoning"]) {
+		const value = record[key];
+		if (typeof value === "string" && value !== "") return value;
+	}
+}
 
 //#endregion
 export { OpenAIModel };

package/esm/model.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"model.mjs","names":[],"sources":["../../../../../../../packages/ai-openai/src/model.ts"],"sourcesContent":["import {\n  InvalidRequestError,\n  ProviderError,\n  safeJsonParse,\n  type Message,\n  type ModelCallOptions,\n  type ModelCapabilities,\n  type ModelContract,\n  type ModelPricing,\n  type ModelResponse,\n  type ModelStreamChunk,\n  type ModelToolCallRequest,\n  type Usage,\n} from \"@warlock.js/ai\";\nimport { log, type Logger } from \"@warlock.js/logger\";\nimport type OpenAI from \"openai\";\nimport type { OpenAIModelConfig } from \"./config.type\";\nimport { inferReasoningCapability } from \"./known-reasoning-models\";\nimport { inferVisionCapability } from \"./known-vision-models\";\nimport {\n  buildPromptCacheParams,\n  buildUsage,\n  inferStructuredOutput,\n  mapFinishReason,\n  planStructuredOutput,\n  toOpenAIMessages,\n  toOpenAIReasoningEffort,\n  toOpenAITools,\n  wrapOpenAIError,\n} from \"./utils\";\n\nconst LOG_MODULE = \"ai.openai\";\n\n/**\n * OpenAI-backed implementation of `ModelContract`.\n *\n * **Role.** The provider-facing bridge between the vendor-neutral\n * `@warlock.js/ai` agent runtime and the official `openai` SDK. Agents,\n * workflows, and supervisors never talk to OpenAI directly — they hold a\n * `ModelContract`, and this class is what makes that contract concrete for\n * any OpenAI-compatible endpoint (OpenAI, Azure OpenAI, OpenRouter, local\n * gateways that speak the Chat Completions protocol).\n *\n * **Responsibility.**\n * - Owns: a long-lived `OpenAI` client + frozen `ModelConfig` (name,\n *   temperature, maxTokens) used as defaults for every call.\n * - Owns: translating vendor-neutral `Message[]` and\n *   `ToolContract[]` into OpenAI wire shapes on the way out, and\n *   translating OpenAI's response (content, finish reason, tool calls,\n *   usage) back into the neutral shapes on the way in.\n * - Does NOT own: dispatching tools, deciding whether to loop, tracking\n *   conversation history, or retrying on failure — those are agent\n *   concerns. The model is a stateless (per-call) protocol adapter.\n *\n * Because it holds a live client and shared defaults, it is modeled as a\n * class (see §4.2 of code-style.md — \"long-lived state across calls\").\n *\n * @example\n * import OpenAI from \"openai\";\n * const client = new OpenAI({ apiKey: process.env.OPENAI_API_KEY });\n * const model = new OpenAIModel(client, { name: \"gpt-4o\", temperature: 0.3 });\n *\n * const myAgent = agent({\n *   model,\n *   systemPrompt: \"You are a helpful assistant.\",\n *   tools: [searchTool],\n * });\n *\n * const result = await myAgent.execute(\"Summarize today's news.\");\n */\nexport class OpenAIModel implements ModelContract {\n  public readonly name: string;\n  public readonly provider: string;\n  public readonly capabilities: ModelCapabilities;\n  public readonly pricing?: ModelPricing;\n\n  private readonly client: OpenAI;\n  private readonly config: OpenAIModelConfig;\n  private readonly logger: Logger = log;\n\n  public constructor(client: OpenAI, config: OpenAIModelConfig, provider: string = \"openai\") {\n    // `api: \"responses\"` belongs to `OpenAIResponsesModel`, which only the\n    // `OpenAISDK.model()` factory constructs. This class speaks Chat\n    // Completions only; accepting the flag here would silently ignore it.\n    if (config.api === \"responses\") {\n      throw new InvalidRequestError(\n        'OpenAIModel speaks Chat Completions only and cannot serve api: \"responses\". Use OpenAISDK.model({ api: \"responses\" }) on the direct OpenAI provider.',\n        { context: { provider, model: config.name, api: config.api } },\n      );\n    }\n\n    this.client = client;\n    this.config = config;\n    this.name = config.name;\n    this.provider = provider;\n    this.pricing = config.pricing;\n    this.capabilities = {\n      structuredOutput: config.structuredOutput ?? inferStructuredOutput(config.responseFormat),\n      vision: config.vision ?? inferVisionCapability(config.name),\n      // o-series + gpt-5 / gpt-6 models surface a reasoning channel and accept\n      // the `reasoning_effort` param. Explicit config wins over the\n      // name-prefix inference.\n      reasoning: config.reasoning ?? inferReasoningCapability(config.name),\n      // Direct OpenAI requests support automatic cache hits and explicit\n      // content-part breakpoints. Wrapper endpoints have distinct provider\n      // labels and may reject these OpenAI-only request fields.\n      promptCaching: this.provider === \"openai\",\n      // PDF + audio INPUT are off by default — OpenAI accepts `file`\n      // (PDF) and `input_audio` parts only on specific models, so the\n      // flags are conservative/honest and opt-in via config rather than\n      // name-inferred. When set, the agent admits the attachments and\n      // `toOpenAIMessages` maps them to the real wire parts.\n      pdf: config.pdf ?? false,\n      audio: config.audio ?? false,\n    };\n  }\n\n  /**\n   * Single-shot completion. Sends the full message list to the Chat\n   * Completions endpoint, waits for the terminal response, and reshapes it\n   * into a vendor-neutral `ModelResponse`. Per-call `options` override the\n   * instance's `ModelConfig` defaults for this call only.\n   */\n  public async complete(messages: Message[], options?: ModelCallOptions): Promise<ModelResponse> {\n    // Per-call request/response logs are hot-path in production agents\n    // — keep them at `debug` so `info` stays reserved for lifecycle\n    // events (agent starting/completed, etc.). Operators who need to\n    // audit every LLM call can raise log-level at runtime.\n    this.logger.debug(LOG_MODULE, \"request\", \"Starting call to chat.completions\", {\n      model: this.name,\n      messageCount: messages.length,\n      streaming: false,\n      toolCount: options?.tools?.length ?? 0,\n    });\n\n    let response: OpenAI.Chat.Completions.ChatCompletion;\n\n    try {\n      response = await this.client.chat.completions.create(\n        {\n          model: this.name,\n          messages: toOpenAIMessages(messages, this.promptCacheBreakpoints(options)),\n          ...this.buildSamplingParams(options),\n          ...buildPromptCacheParams(this.provider, this.config),\n          tools: toOpenAITools(options?.tools),\n          ...this.buildResponseFormat(options?.responseSchema),\n          ...this.buildReasoningParams(options?.reasoning, Boolean(options?.tools?.length)),\n        },\n        options?.signal ? { signal: options.signal } : undefined,\n      );\n    } catch (thrown) {\n      const wrapped = wrapOpenAIError(thrown);\n\n      this.logger.error(LOG_MODULE, \"error\", wrapped.message, {\n        code: wrapped.code,\n        context: wrapped.context,\n      });\n\n      throw wrapped;\n    }\n\n    const [choice] = response.choices;\n\n    if (choice === undefined) {\n      // A 200 carrying an empty `choices` array. Every line below reads off\n      // `choice`, so without this the caller got a bare `TypeError: Cannot\n      // read properties of undefined` — no provider, no model, no hint that\n      // the response itself was the problem rather than our handling of it.\n      //\n      // `ProviderError` because that is what this is: the provider answered\n      // successfully and returned nothing usable. Callers already branch on\n      // `instanceof ProviderError` for any provider-side failure.\n      throw new ProviderError(\"OpenAI returned a response with no choices.\", {\n        context: { provider: \"openai\", model: this.name, choices: response.choices.length },\n      });\n    }\n\n    const finishReason = mapFinishReason(choice.finish_reason);\n    const usage = this.extractUsage(response.usage);\n\n    this.logger.debug(LOG_MODULE, \"response\", \"call to chat.completions succeeded\", {\n      finishReason,\n      usage,\n    });\n\n    return {\n      content: choice.message.content ?? \"\",\n      finishReason,\n      usage,\n      toolCalls: this.extractToolCalls(choice.message.tool_calls),\n    };\n  }\n\n  /**\n   * Incremental streaming completion. Yields neutral `ModelStreamChunk`s —\n   * `delta` for text tokens, `tool-call` when the model requests a tool,\n   * and a terminal `done` carrying the final finish reason + usage totals.\n   * Callers consume it with `for await`.\n   */\n  public async *stream(\n    messages: Message[],\n    options?: ModelCallOptions,\n  ): AsyncIterable<ModelStreamChunk> {\n    this.logger.debug(LOG_MODULE, \"request\", \"Starting streaming call to chat.completions\", {\n      model: this.name,\n      messageCount: messages.length,\n      streaming: true,\n      toolCount: options?.tools?.length ?? 0,\n    });\n\n    let stream: Awaited<ReturnType<typeof this.client.chat.completions.create>>;\n\n    try {\n      stream = await this.client.chat.completions.create(\n        {\n          model: this.name,\n          messages: toOpenAIMessages(messages, this.promptCacheBreakpoints(options)),\n          ...this.buildSamplingParams(options),\n          ...buildPromptCacheParams(this.provider, this.config),\n          tools: toOpenAITools(options?.tools),\n          stream: true,\n          stream_options: { include_usage: true },\n          ...this.buildResponseFormat(options?.responseSchema),\n          ...this.buildReasoningParams(options?.reasoning, Boolean(options?.tools?.length)),\n        },\n        options?.signal ? { signal: options.signal } : undefined,\n      );\n    } catch (thrown) {\n      const wrapped = wrapOpenAIError(thrown);\n\n      this.logger.error(LOG_MODULE, \"error\", wrapped.message, {\n        code: wrapped.code,\n        context: wrapped.context,\n      });\n\n      throw wrapped;\n    }\n\n    let rawFinishReason: string = \"stop\";\n    const usage: Usage = { input: 0, output: 0, total: 0 };\n    const toolCallAccum = new Map<number, { id: string; name: string; arguments: string }>();\n\n    try {\n      for await (const chunk of stream as AsyncIterable<OpenAI.Chat.Completions.ChatCompletionChunk>) {\n        const delta = chunk.choices[0]?.delta;\n        const finish = chunk.choices[0]?.finish_reason;\n\n        if (delta?.content) {\n          yield { type: \"delta\", content: delta.content };\n        }\n\n        if (delta?.tool_calls) {\n          for (const toolCall of delta.tool_calls) {\n            const idx = toolCall.index ?? 0;\n            if (!toolCallAccum.has(idx)) {\n              toolCallAccum.set(idx, { id: \"\", name: \"\", arguments: \"\" });\n            }\n            const acc = toolCallAccum.get(idx)!;\n            if (toolCall.id) acc.id = toolCall.id;\n            if (toolCall.function?.name) acc.name = toolCall.function.name;\n            if (toolCall.function?.arguments) acc.arguments += toolCall.function.arguments;\n          }\n        }\n\n        if (finish) {\n          rawFinishReason = finish;\n        }\n\n        if (chunk.usage) {\n          usage.input = chunk.usage.prompt_tokens ?? 0;\n          usage.output = chunk.usage.completion_tokens ?? 0;\n          usage.total = chunk.usage.total_tokens ?? 0;\n          const cached = chunk.usage.prompt_tokens_details?.cached_tokens;\n          if (cached !== undefined && cached > 0) {\n            usage.cachedTokens = cached;\n          }\n          const reasoning = chunk.usage.completion_tokens_details?.reasoning_tokens;\n          if (reasoning !== undefined && reasoning > 0) {\n            usage.reasoningTokens = reasoning;\n          }\n        }\n      }\n\n      for (const acc of toolCallAccum.values()) {\n        // Skip accumulators that never received a function name — those\n        // are partial fragments the model started but never identified\n        // (e.g. arguments-only deltas with no originating `id`/`name`).\n        // Yielding them produces nameless tool-calls the agent runtime\n        // can't dispatch and would mis-attribute as a registered tool.\n        if (!acc.name) continue;\n\n        yield {\n          type: \"tool-call\",\n          id: acc.id,\n          name: acc.name,\n          input: safeJsonParse<Record<string, unknown>>(acc.arguments, {}),\n        };\n      }\n    } catch (thrown) {\n      const wrapped = wrapOpenAIError(thrown);\n\n      this.logger.error(LOG_MODULE, \"error\", wrapped.message, {\n        code: wrapped.code,\n        context: wrapped.context,\n      });\n\n      throw wrapped;\n    }\n\n    const finishReason = mapFinishReason(rawFinishReason);\n\n    this.logger.debug(LOG_MODULE, \"response\", \"Streaming call to chat.completions succeeded\", {\n      finishReason,\n      usage,\n    });\n\n    yield { type: \"done\", finishReason, usage };\n  }\n\n  /** Map the neutral cache hint only for the direct OpenAI provider. */\n  private promptCacheBreakpoints(options: ModelCallOptions | undefined): number {\n    return this.provider === \"openai\" ? (options?.cacheControl?.breakpoints ?? 0) : 0;\n  }\n\n  /**\n   * Translate the neutral `responseSchema` option into OpenAI's\n   * `response_format` parameter. The mode decision (override handling,\n   * strict-compatibility downgrade to `json_object`) lives in\n   * `planStructuredOutput`, shared with the Responses adapter.\n   *\n   * Returns an empty spread when no schema was supplied, so the caller\n   * can unconditionally `...buildResponseFormat(...)` into the request.\n   */\n  private buildResponseFormat(responseSchema: Record<string, unknown> | undefined): {\n    response_format?: OpenAI.Chat.Completions.ChatCompletionCreateParams[\"response_format\"];\n  } {\n    const plan = planStructuredOutput(responseSchema, this.config.responseFormat);\n\n    if (plan.mode === \"json_schema\") {\n      return {\n        response_format: {\n          type: \"json_schema\",\n          json_schema: { name: \"response\", schema: plan.schema, strict: true },\n        },\n      };\n    }\n\n    if (plan.mode === \"json_object\") {\n      return { response_format: { type: \"json_object\" } };\n    }\n\n    return {};\n  }\n\n  /**\n   * Normalize OpenAI's `usage` block (which may be absent on some responses\n   * or partials) into the neutral `Usage` shape. Missing usage collapses to\n   * zeros rather than propagating `undefined`, so downstream aggregation\n   * math stays safe.\n   *\n   * `cachedTokens` mirrors `prompt_tokens_details.cached_tokens` (the\n   * subset of the prompt served from OpenAI's automatic prompt cache);\n   * `reasoningTokens` mirrors `completion_tokens_details.reasoning_tokens`\n   * (the hidden reasoning channel on o-series / gpt-5 models, already\n   * counted within `output`). Both are emitted only when the provider\n   * reports a positive value, so non-reasoning / uncached calls keep the\n   * lean `{ input, output, total }` shape.\n   */\n  private extractUsage(raw: OpenAI.Completions.CompletionUsage | undefined): Usage {\n    if (!raw) {\n      return { input: 0, output: 0, total: 0 };\n    }\n\n    return buildUsage({\n      input: raw.prompt_tokens,\n      output: raw.completion_tokens,\n      total: raw.total_tokens,\n      cachedTokens: raw.prompt_tokens_details?.cached_tokens,\n      reasoningTokens: raw.completion_tokens_details?.reasoning_tokens,\n    });\n  }\n\n  /**\n   * Translate the neutral `ModelCallOptions.reasoning` hint into OpenAI's\n   * `reasoning_effort` request param. Only `effort` maps — OpenAI's Chat\n   * Completions API exposes a discrete effort knob, not a token budget,\n   * so `reasoning.maxTokens` (the Anthropic extended-thinking cap) has no\n   * wire equivalent here and is silently ignored.\n   *\n   * The neutral `ReasoningEffort` (`\"low\" | \"medium\" | \"high\" | \"none\"`)\n   * is a subset of OpenAI's accepted values, so it forwards verbatim —\n   * `\"none\"` included. `\"none\"` is load-bearing: gpt-5 / o-series models\n   * **reject function tools** on Chat Completions while reasoning is\n   * active, and the endpoint accepts tools only when `reasoning_effort`\n   * is `\"none\"` (the alternative is the Responses API). This is why the\n   * param is EMITTED for `\"none\"` rather than omitted — omitting it\n   * leaves the model reasoning server-side by default, so tools would\n   * still be rejected.\n   *\n   * No-ops when the model is not reasoning-capable\n   * (`capabilities.reasoning` is false — e.g. `gpt-4o`, which 400s on any\n   * `reasoning_effort`).\n   *\n   * When the caller gave no explicit `effort`: defaults to `\"none\"` IF\n   * `hasTools` is true, else omits the param (provider default — the\n   * pre-existing behavior for a tool-less call). The default exists\n   * because, on a reasoning-capable model, there is no working alternative\n   * to `\"none\"` when tools are attached — omitting `reasoning_effort`\n   * leaves reasoning on server-side, which Chat Completions rejects tools\n   * for (empty replies on some model generations, a hard 400 — \"Function\n   * tools with reasoning_effort are not supported ... in\n   * /v1/chat/completions\" — on newer ones). An explicit `effort` from the\n   * caller always wins over this default, in either direction.\n   *\n   * Returns an empty spread when nothing applies, so the caller can\n   * unconditionally `...buildReasoningParams(...)` into the request.\n   */\n  /**\n   * Output-length and sampling params, in current Chat Completions shapes only.\n   *\n   * - `max_completion_tokens`, never the legacy `max_tokens`: current\n   *   models (the gpt-5 family and later) answer `400 unsupported_parameter`\n   *   to `max_tokens`, and `max_completion_tokens` is accepted everywhere.\n   * - No `temperature` for a reasoning-capable model: those accept only the\n   *   default value and reject any other with a 400.\n   */\n  private buildSamplingParams(options: ModelCallOptions | undefined): {\n    max_completion_tokens?: number;\n    temperature?: number;\n  } {\n    const maxTokens = options?.maxTokens ?? this.config.maxTokens;\n    const temperature = options?.temperature ?? this.config.temperature;\n\n    return {\n      max_completion_tokens: maxTokens,\n      ...(!this.capabilities.reasoning && { temperature }),\n    };\n  }\n\n  private buildReasoningParams(\n    reasoning: ModelCallOptions[\"reasoning\"],\n    hasTools: boolean,\n  ): {\n    reasoning_effort?: OpenAI.Chat.Completions.ChatCompletionCreateParams[\"reasoning_effort\"];\n  } {\n    if (!this.capabilities.reasoning) {\n      return {};\n    }\n\n    if (!reasoning?.effort) {\n      return hasTools ? { reasoning_effort: \"none\" } : {};\n    }\n\n    return { reasoning_effort: toOpenAIReasoningEffort(reasoning.effort) };\n  }\n\n  /**\n   * Reshape OpenAI's `tool_calls` array into the neutral\n   * `ModelToolCallRequest[]`. The raw `arguments` field is a JSON string\n   * per OpenAI's protocol — we parse it defensively via `safeJsonParse` so\n   * malformed or empty arguments yield an empty object instead of crashing\n   * the trip. Returns `undefined` when no tools were requested so callers\n   * can branch on presence.\n   */\n  private extractToolCalls(\n    rawToolCalls: OpenAI.Chat.Completions.ChatCompletionMessageToolCall[] | undefined,\n  ): ModelToolCallRequest[] | undefined {\n    if (!rawToolCalls || rawToolCalls.length === 0) {\n      return undefined;\n    }\n\n    return rawToolCalls.map((toolCall) => ({\n      id: toolCall.id,\n      name: (toolCall as any).function.name,\n      input: safeJsonParse<Record<string, unknown>>((toolCall as any).function.arguments, {}),\n    }));\n  }\n}\n"],"mappings":";;;;;;;;;;;;;;;AA+BA,MAAM,aAAa;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAuCnB,IAAa,cAAb,MAAkD;CAUhD,AAAO,YAAY,QAAgB,QAA2B,WAAmB,UAAU;gBAFzD;EAMhC,IAAI,OAAO,QAAQ,aACjB,MAAM,IAAI,oBACR,4JACA,EAAE,SAAS;GAAE;GAAU,OAAO,OAAO;GAAM,KAAK,OAAO;EAAI,EAAE,CAC/D;EAGF,KAAK,SAAS;EACd,KAAK,SAAS;EACd,KAAK,OAAO,OAAO;EACnB,KAAK,WAAW;EAChB,KAAK,UAAU,OAAO;EACtB,KAAK,eAAe;GAClB,kBAAkB,OAAO,oBAAoB,sBAAsB,OAAO,cAAc;GACxF,QAAQ,OAAO,UAAU,sBAAsB,OAAO,IAAI;GAI1D,WAAW,OAAO,aAAa,yBAAyB,OAAO,IAAI;GAInE,eAAe,KAAK,aAAa;GAMjC,KAAK,OAAO,OAAO;GACnB,OAAO,OAAO,SAAS;EACzB;CACF;;;;;;;CAQA,MAAa,SAAS,UAAqB,SAAoD;EAK7F,KAAK,OAAO,MAAM,YAAY,WAAW,qCAAqC;GAC5E,OAAO,KAAK;GACZ,cAAc,SAAS;GACvB,WAAW;GACX,WAAW,SAAS,OAAO,UAAU;EACvC,CAAC;EAED,IAAI;EAEJ,IAAI;GACF,WAAW,MAAM,KAAK,OAAO,KAAK,YAAY,OAC5C;IACE,OAAO,KAAK;IACZ,UAAU,iBAAiB,UAAU,KAAK,uBAAuB,OAAO,CAAC;IACzE,GAAG,KAAK,oBAAoB,OAAO;IACnC,GAAG,uBAAuB,KAAK,UAAU,KAAK,MAAM;IACpD,OAAO,cAAc,SAAS,KAAK;IACnC,GAAG,KAAK,oBAAoB,SAAS,cAAc;IACnD,GAAG,KAAK,qBAAqB,SAAS,WAAW,QAAQ,SAAS,OAAO,MAAM,CAAC;GAClF,GACA,SAAS,SAAS,EAAE,QAAQ,QAAQ,OAAO,IAAI,MACjD;EACF,SAAS,QAAQ;GACf,MAAM,UAAU,gBAAgB,MAAM;GAEtC,KAAK,OAAO,MAAM,YAAY,SAAS,QAAQ,SAAS;IACtD,MAAM,QAAQ;IACd,SAAS,QAAQ;GACnB,CAAC;GAED,MAAM;EACR;EAEA,MAAM,CAAC,UAAU,SAAS;EAE1B,IAAI,WAAW,QASb,MAAM,IAAI,cAAc,+CAA+C,EACrE,SAAS;GAAE,UAAU;GAAU,OAAO,KAAK;GAAM,SAAS,SAAS,QAAQ;EAAO,EACpF,CAAC;EAGH,MAAM,eAAe,gBAAgB,OAAO,aAAa;EACzD,MAAM,QAAQ,KAAK,aAAa,SAAS,KAAK;EAE9C,KAAK,OAAO,MAAM,YAAY,YAAY,sCAAsC;GAC9E;GACA;EACF,CAAC;EAED,OAAO;GACL,SAAS,OAAO,QAAQ,WAAW;GACnC;GACA;GACA,WAAW,KAAK,iBAAiB,OAAO,QAAQ,UAAU;EAC5D;CACF;;;;;;;CAQA,OAAc,OACZ,UACA,SACiC;EACjC,KAAK,OAAO,MAAM,YAAY,WAAW,+CAA+C;GACtF,OAAO,KAAK;GACZ,cAAc,SAAS;GACvB,WAAW;GACX,WAAW,SAAS,OAAO,UAAU;EACvC,CAAC;EAED,IAAI;EAEJ,IAAI;GACF,SAAS,MAAM,KAAK,OAAO,KAAK,YAAY,OAC1C;IACE,OAAO,KAAK;IACZ,UAAU,iBAAiB,UAAU,KAAK,uBAAuB,OAAO,CAAC;IACzE,GAAG,KAAK,oBAAoB,OAAO;IACnC,GAAG,uBAAuB,KAAK,UAAU,KAAK,MAAM;IACpD,OAAO,cAAc,SAAS,KAAK;IACnC,QAAQ;IACR,gBAAgB,EAAE,eAAe,KAAK;IACtC,GAAG,KAAK,oBAAoB,SAAS,cAAc;IACnD,GAAG,KAAK,qBAAqB,SAAS,WAAW,QAAQ,SAAS,OAAO,MAAM,CAAC;GAClF,GACA,SAAS,SAAS,EAAE,QAAQ,QAAQ,OAAO,IAAI,MACjD;EACF,SAAS,QAAQ;GACf,MAAM,UAAU,gBAAgB,MAAM;GAEtC,KAAK,OAAO,MAAM,YAAY,SAAS,QAAQ,SAAS;IACtD,MAAM,QAAQ;IACd,SAAS,QAAQ;GACnB,CAAC;GAED,MAAM;EACR;EAEA,IAAI,kBAA0B;EAC9B,MAAM,QAAe;GAAE,OAAO;GAAG,QAAQ;GAAG,OAAO;EAAE;EACrD,MAAM,gCAAgB,IAAI,IAA6D;EAEvF,IAAI;GACF,WAAW,MAAM,SAAS,QAAsE;IAC9F,MAAM,QAAQ,MAAM,QAAQ,EAAE,EAAE;IAChC,MAAM,SAAS,MAAM,QAAQ,EAAE,EAAE;IAEjC,IAAI,OAAO,SACT,MAAM;KAAE,MAAM;KAAS,SAAS,MAAM;IAAQ;IAGhD,IAAI,OAAO,YACT,KAAK,MAAM,YAAY,MAAM,YAAY;KACvC,MAAM,MAAM,SAAS,SAAS;KAC9B,IAAI,CAAC,cAAc,IAAI,GAAG,GACxB,cAAc,IAAI,KAAK;MAAE,IAAI;MAAI,MAAM;MAAI,WAAW;KAAG,CAAC;KAE5D,MAAM,MAAM,cAAc,IAAI,GAAG;KACjC,IAAI,SAAS,IAAI,IAAI,KAAK,SAAS;KACnC,IAAI,SAAS,UAAU,MAAM,IAAI,OAAO,SAAS,SAAS;KAC1D,IAAI,SAAS,UAAU,WAAW,IAAI,aAAa,SAAS,SAAS;IACvE;IAGF,IAAI,QACF,kBAAkB;IAGpB,IAAI,MAAM,OAAO;KACf,MAAM,QAAQ,MAAM,MAAM,iBAAiB;KAC3C,MAAM,SAAS,MAAM,MAAM,qBAAqB;KAChD,MAAM,QAAQ,MAAM,MAAM,gBAAgB;KAC1C,MAAM,SAAS,MAAM,MAAM,uBAAuB;KAClD,IAAI,WAAW,UAAa,SAAS,GACnC,MAAM,eAAe;KAEvB,MAAM,YAAY,MAAM,MAAM,2BAA2B;KACzD,IAAI,cAAc,UAAa,YAAY,GACzC,MAAM,kBAAkB;IAE5B;GACF;GAEA,KAAK,MAAM,OAAO,cAAc,OAAO,GAAG;IAMxC,IAAI,CAAC,IAAI,MAAM;IAEf,MAAM;KACJ,MAAM;KACN,IAAI,IAAI;KACR,MAAM,IAAI;KACV,OAAO,cAAuC,IAAI,WAAW,CAAC,CAAC;IACjE;GACF;EACF,SAAS,QAAQ;GACf,MAAM,UAAU,gBAAgB,MAAM;GAEtC,KAAK,OAAO,MAAM,YAAY,SAAS,QAAQ,SAAS;IACtD,MAAM,QAAQ;IACd,SAAS,QAAQ;GACnB,CAAC;GAED,MAAM;EACR;EAEA,MAAM,eAAe,gBAAgB,eAAe;EAEpD,KAAK,OAAO,MAAM,YAAY,YAAY,gDAAgD;GACxF;GACA;EACF,CAAC;EAED,MAAM;GAAE,MAAM;GAAQ;GAAc;EAAM;CAC5C;;CAGA,AAAQ,uBAAuB,SAA+C;EAC5E,OAAO,KAAK,aAAa,WAAY,SAAS,cAAc,eAAe,IAAK;CAClF;;;;;;;;;;CAWA,AAAQ,oBAAoB,gBAE1B;EACA,MAAM,OAAO,qBAAqB,gBAAgB,KAAK,OAAO,cAAc;EAE5E,IAAI,KAAK,SAAS,eAChB,OAAO,EACL,iBAAiB;GACf,MAAM;GACN,aAAa;IAAE,MAAM;IAAY,QAAQ,KAAK;IAAQ,QAAQ;GAAK;EACrE,EACF;EAGF,IAAI,KAAK,SAAS,eAChB,OAAO,EAAE,iBAAiB,EAAE,MAAM,cAAc,EAAE;EAGpD,OAAO,CAAC;CACV;;;;;;;;;;;;;;;CAgBA,AAAQ,aAAa,KAA4D;EAC/E,IAAI,CAAC,KACH,OAAO;GAAE,OAAO;GAAG,QAAQ;GAAG,OAAO;EAAE;EAGzC,OAAO,WAAW;GAChB,OAAO,IAAI;GACX,QAAQ,IAAI;GACZ,OAAO,IAAI;GACX,cAAc,IAAI,uBAAuB;GACzC,iBAAiB,IAAI,2BAA2B;EAClD,CAAC;CACH;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;CA8CA,AAAQ,oBAAoB,SAG1B;EACA,MAAM,YAAY,SAAS,aAAa,KAAK,OAAO;EACpD,MAAM,cAAc,SAAS,eAAe,KAAK,OAAO;EAExD,OAAO;GACL,uBAAuB;GACvB,GAAI,CAAC,KAAK,aAAa,aAAa,EAAE,YAAY;EACpD;CACF;CAEA,AAAQ,qBACN,WACA,UAGA;EACA,IAAI,CAAC,KAAK,aAAa,WACrB,OAAO,CAAC;EAGV,IAAI,CAAC,WAAW,QACd,OAAO,WAAW,EAAE,kBAAkB,OAAO,IAAI,CAAC;EAGpD,OAAO,EAAE,kBAAkB,wBAAwB,UAAU,MAAM,EAAE;CACvE;;;;;;;;;CAUA,AAAQ,iBACN,cACoC;EACpC,IAAI,CAAC,gBAAgB,aAAa,WAAW,GAC3C;EAGF,OAAO,aAAa,KAAK,cAAc;GACrC,IAAI,SAAS;GACb,MAAO,SAAiB,SAAS;GACjC,OAAO,cAAwC,SAAiB,SAAS,WAAW,CAAC,CAAC;EACxF,EAAE;CACJ;AACF"}
+{"version":3,"file":"model.mjs","names":[],"sources":["../../../../../../../packages/ai-openai/src/model.ts"],"sourcesContent":["import {\n  InvalidRequestError,\n  ProviderError,\n  resolveStreamChunks,\n  safeJsonParse,\n  selectProviderTools,\n  type Message,\n  type ModelCallOptions,\n  type ModelCapabilities,\n  type ModelContract,\n  type ModelPricing,\n  type ModelResponse,\n  type ModelStreamChunk,\n  type ModelToolCallRequest,\n  type Usage,\n} from \"@warlock.js/ai\";\nimport { log, type Logger } from \"@warlock.js/logger\";\nimport type OpenAI from \"openai\";\nimport { mergeProviderBody } from \"./provider-options\";\nimport type { OpenAIModelConfig } from \"./config.type\";\nimport { inferReasoningCapability } from \"./known-reasoning-models\";\nimport { inferVisionCapability } from \"./known-vision-models\";\nimport {\n  buildPromptCacheParams,\n  buildUsage,\n  inferStructuredOutput,\n  mapFinishReason,\n  planStructuredOutput,\n  toOpenAIMessages,\n  toOpenAIReasoningEffort,\n  toOpenAITools,\n  wrapOpenAIError,\n} from \"./utils\";\n\nconst LOG_MODULE = \"ai.openai\";\n\n/**\n * OpenAI-backed implementation of `ModelContract`.\n *\n * **Role.** The provider-facing bridge between the vendor-neutral\n * `@warlock.js/ai` agent runtime and the official `openai` SDK. Agents,\n * workflows, and supervisors never talk to OpenAI directly — they hold a\n * `ModelContract`, and this class is what makes that contract concrete for\n * any OpenAI-compatible endpoint (OpenAI, Azure OpenAI, OpenRouter, local\n * gateways that speak the Chat Completions protocol).\n *\n * **Responsibility.**\n * - Owns: a long-lived `OpenAI` client + frozen `ModelConfig` (name,\n *   temperature, maxTokens) used as defaults for every call.\n * - Owns: translating vendor-neutral `Message[]` and\n *   `ToolContract[]` into OpenAI wire shapes on the way out, and\n *   translating OpenAI's response (content, finish reason, tool calls,\n *   usage) back into the neutral shapes on the way in.\n * - Does NOT own: dispatching tools, deciding whether to loop, tracking\n *   conversation history, or retrying on failure — those are agent\n *   concerns. The model is a stateless (per-call) protocol adapter.\n *\n * Because it holds a live client and shared defaults, it is modeled as a\n * class (see §4.2 of code-style.md — \"long-lived state across calls\").\n *\n * @example\n * import OpenAI from \"openai\";\n * const client = new OpenAI({ apiKey: process.env.OPENAI_API_KEY });\n * const model = new OpenAIModel(client, { name: \"gpt-4o\", temperature: 0.3 });\n *\n * const myAgent = agent({\n *   model,\n *   systemPrompt: \"You are a helpful assistant.\",\n *   tools: [searchTool],\n * });\n *\n * const result = await myAgent.execute(\"Summarize today's news.\");\n */\nexport class OpenAIModel implements ModelContract {\n  public readonly name: string;\n  public readonly provider: string;\n  public readonly capabilities: ModelCapabilities;\n  public readonly pricing?: ModelPricing;\n\n  private readonly client: OpenAI;\n  private readonly config: OpenAIModelConfig;\n  private readonly logger: Logger = log;\n\n  public constructor(client: OpenAI, config: OpenAIModelConfig, provider: string = \"openai\") {\n    // `api: \"responses\"` belongs to `OpenAIResponsesModel`, which only the\n    // `OpenAISDK.model()` factory constructs. This class speaks Chat\n    // Completions only; accepting the flag here would silently ignore it.\n    if (config.api === \"responses\") {\n      throw new InvalidRequestError(\n        'OpenAIModel speaks Chat Completions only and cannot serve api: \"responses\". Use OpenAISDK.model({ api: \"responses\" }) on the direct OpenAI provider.',\n        { context: { provider, model: config.name, api: config.api } },\n      );\n    }\n\n    this.client = client;\n    this.config = config;\n    this.name = config.name;\n    this.provider = provider;\n    this.pricing = config.pricing;\n    this.capabilities = {\n      structuredOutput: config.structuredOutput ?? inferStructuredOutput(config.responseFormat),\n      vision: config.vision ?? inferVisionCapability(config.name),\n      // o-series + gpt-5 / gpt-6 models surface a reasoning channel and accept\n      // the `reasoning_effort` param. Explicit config wins over the\n      // name-prefix inference.\n      reasoning: config.reasoning ?? inferReasoningCapability(config.name),\n      // Direct OpenAI requests support automatic cache hits and explicit\n      // content-part breakpoints. Wrapper endpoints have distinct provider\n      // labels and may reject these OpenAI-only request fields.\n      promptCaching: this.provider === \"openai\",\n      // PDF + audio INPUT are off by default — OpenAI accepts `file`\n      // (PDF) and `input_audio` parts only on specific models, so the\n      // flags are conservative/honest and opt-in via config rather than\n      // name-inferred. When set, the agent admits the attachments and\n      // `toOpenAIMessages` maps them to the real wire parts.\n      pdf: config.pdf ?? false,\n      audio: config.audio ?? false,\n    };\n  }\n\n  /**\n   * Single-shot completion. Sends the full message list to the Chat\n   * Completions endpoint, waits for the terminal response, and reshapes it\n   * into a vendor-neutral `ModelResponse`. Per-call `options` override the\n   * instance's `ModelConfig` defaults for this call only.\n   */\n  public async complete(messages: Message[], options?: ModelCallOptions): Promise<ModelResponse> {\n    // Per-call request/response logs are hot-path in production agents\n    // — keep them at `debug` so `info` stays reserved for lifecycle\n    // events (agent starting/completed, etc.). Operators who need to\n    // audit every LLM call can raise log-level at runtime.\n    this.logger.debug(LOG_MODULE, \"request\", \"Starting call to chat.completions\", {\n      model: this.name,\n      messageCount: messages.length,\n      streaming: false,\n      toolCount: options?.tools?.length ?? 0,\n    });\n\n    let response: OpenAI.Chat.Completions.ChatCompletion;\n\n    try {\n      response = await this.client.chat.completions.create(\n        mergeProviderBody<OpenAI.Chat.Completions.ChatCompletionCreateParamsNonStreaming>(\n          {\n            model: this.name,\n            messages: toOpenAIMessages(messages, this.promptCacheBreakpoints(options)),\n            ...this.buildSamplingParams(options),\n            ...buildPromptCacheParams(this.provider, this.config),\n            ...this.buildTools(options),\n            ...this.buildResponseFormat(options?.responseSchema),\n            ...this.buildReasoningParams(options?.reasoning, Boolean(options?.tools?.length)),\n          },\n          options,\n        ),\n        options?.signal ? { signal: options.signal } : undefined,\n      );\n    } catch (thrown) {\n      const wrapped = wrapOpenAIError(thrown);\n\n      this.logger.error(LOG_MODULE, \"error\", wrapped.message, {\n        code: wrapped.code,\n        context: wrapped.context,\n      });\n\n      throw wrapped;\n    }\n\n    const [choice] = response.choices;\n\n    if (choice === undefined) {\n      // A 200 carrying an empty `choices` array. Every line below reads off\n      // `choice`, so without this the caller got a bare `TypeError: Cannot\n      // read properties of undefined` — no provider, no model, no hint that\n      // the response itself was the problem rather than our handling of it.\n      //\n      // `ProviderError` because that is what this is: the provider answered\n      // successfully and returned nothing usable. Callers already branch on\n      // `instanceof ProviderError` for any provider-side failure.\n      throw new ProviderError(\"OpenAI returned a response with no choices.\", {\n        context: { provider: \"openai\", model: this.name, choices: response.choices.length },\n      });\n    }\n\n    const finishReason = mapFinishReason(choice.finish_reason);\n    const usage = this.extractUsage(response.usage);\n\n    this.logger.debug(LOG_MODULE, \"response\", \"call to chat.completions succeeded\", {\n      finishReason,\n      usage,\n    });\n\n    return {\n      content: choice.message.content ?? \"\",\n      finishReason,\n      usage,\n      toolCalls: this.extractToolCalls(choice.message.tool_calls),\n    };\n  }\n\n  /**\n   * Incremental streaming completion. Yields neutral `ModelStreamChunk`s —\n   * `delta` for text tokens, `tool-call` when the model requests a tool,\n   * and a terminal `done` carrying the final finish reason + usage totals.\n   * Callers consume it with `for await`.\n   *\n   * With `options.streamChunks` it also yields `tool-args` for each\n   * `tool_calls[].function.arguments` fragment and `reasoning` for a\n   * vendor's `reasoning_content` / `reasoning` delta field. Usage arrives\n   * only on the final chunk here, so no `usage` chunk is emitted.\n   */\n  public async *stream(\n    messages: Message[],\n    options?: ModelCallOptions,\n  ): AsyncIterable<ModelStreamChunk> {\n    this.logger.debug(LOG_MODULE, \"request\", \"Starting streaming call to chat.completions\", {\n      model: this.name,\n      messageCount: messages.length,\n      streaming: true,\n      toolCount: options?.tools?.length ?? 0,\n    });\n\n    let stream: Awaited<ReturnType<typeof this.client.chat.completions.create>>;\n\n    try {\n      stream = await this.client.chat.completions.create(\n        mergeProviderBody<OpenAI.Chat.Completions.ChatCompletionCreateParamsStreaming>(\n          {\n            model: this.name,\n            messages: toOpenAIMessages(messages, this.promptCacheBreakpoints(options)),\n            ...this.buildSamplingParams(options),\n            ...buildPromptCacheParams(this.provider, this.config),\n            ...this.buildTools(options),\n            stream: true,\n            stream_options: { include_usage: true },\n            ...this.buildResponseFormat(options?.responseSchema),\n            ...this.buildReasoningParams(options?.reasoning, Boolean(options?.tools?.length)),\n          },\n          options,\n        ),\n        options?.signal ? { signal: options.signal } : undefined,\n      );\n    } catch (thrown) {\n      const wrapped = wrapOpenAIError(thrown);\n\n      this.logger.error(LOG_MODULE, \"error\", wrapped.message, {\n        code: wrapped.code,\n        context: wrapped.context,\n      });\n\n      throw wrapped;\n    }\n\n    let rawFinishReason: string = \"stop\";\n    let usage: Usage = { input: 0, output: 0, total: 0 };\n    const want = resolveStreamChunks(options);\n    const toolCallAccum = new Map<\n      number,\n      { id: string; name: string; arguments: string; streamed: number }\n    >();\n\n    try {\n      for await (const chunk of stream as AsyncIterable<OpenAI.Chat.Completions.ChatCompletionChunk>) {\n        const delta = chunk.choices[0]?.delta;\n        const finish = chunk.choices[0]?.finish_reason;\n\n        if (want.reasoning && delta) {\n          const reasoning = readReasoningDelta(delta);\n\n          if (reasoning) {\n            yield { type: \"reasoning\", delta: reasoning };\n          }\n        }\n\n        if (delta?.content) {\n          yield { type: \"delta\", content: delta.content };\n        }\n\n        if (delta?.tool_calls) {\n          for (const toolCall of delta.tool_calls) {\n            const idx = toolCall.index ?? 0;\n            if (!toolCallAccum.has(idx)) {\n              toolCallAccum.set(idx, { id: \"\", name: \"\", arguments: \"\", streamed: 0 });\n            }\n            const acc = toolCallAccum.get(idx)!;\n            if (toolCall.id) acc.id = toolCall.id;\n            if (toolCall.function?.name) acc.name = toolCall.function.name;\n            if (toolCall.function?.arguments) acc.arguments += toolCall.function.arguments;\n\n            // The first fragment of a call carries its id and name; later\n            // ones only `index` + argument text. Anything that arrived\n            // before the name is flushed once the call is identified.\n            if (want.toolArgs && acc.name && acc.arguments.length > acc.streamed) {\n              yield {\n                type: \"tool-args\",\n                callId: acc.id,\n                name: acc.name,\n                delta: acc.arguments.slice(acc.streamed),\n              };\n              acc.streamed = acc.arguments.length;\n            }\n          }\n        }\n\n        if (finish) {\n          rawFinishReason = finish;\n        }\n\n        if (chunk.usage) {\n          usage = this.extractUsage(chunk.usage);\n        }\n      }\n\n      for (const acc of toolCallAccum.values()) {\n        // Skip accumulators that never received a function name — those\n        // are partial fragments the model started but never identified\n        // (e.g. arguments-only deltas with no originating `id`/`name`).\n        // Yielding them produces nameless tool-calls the agent runtime\n        // can't dispatch and would mis-attribute as a registered tool.\n        if (!acc.name) continue;\n\n        yield {\n          type: \"tool-call\",\n          id: acc.id,\n          name: acc.name,\n          input: safeJsonParse<Record<string, unknown>>(acc.arguments, {}),\n        };\n      }\n    } catch (thrown) {\n      const wrapped = wrapOpenAIError(thrown);\n\n      this.logger.error(LOG_MODULE, \"error\", wrapped.message, {\n        code: wrapped.code,\n        context: wrapped.context,\n      });\n\n      throw wrapped;\n    }\n\n    const finishReason = mapFinishReason(rawFinishReason);\n\n    this.logger.debug(LOG_MODULE, \"response\", \"Streaming call to chat.completions succeeded\", {\n      finishReason,\n      usage,\n    });\n\n    yield { type: \"done\", finishReason, usage };\n  }\n\n  /**\n   * Spread-friendly `tools` fragment: the function tools, then the\n   * provider-defined entries whose `provider` equals this model's\n   * provider label (`\"openai\"`, or the vendor label an OpenAI-compatible\n   * wrapper sets, e.g. `\"groq\"`), sent verbatim. Chat Completions only\n   * accepts `function` and `custom` tool types from OpenAI itself; its\n   * hosted tools (`web_search`, `file_search`, …) need `api: \"responses\"`\n   * and OpenAI rejects them here with a 400.\n   */\n  private buildTools(options: ModelCallOptions | undefined): {\n    tools?: OpenAI.Chat.Completions.ChatCompletionTool[];\n  } {\n    const { tools, definitions } = selectProviderTools(options, {\n      accepts: [this.provider],\n      module: LOG_MODULE,\n      model: this.name,\n      logger: this.logger,\n    });\n    const mapped = [\n      ...(toOpenAITools(tools) ?? []),\n      ...(definitions as unknown as OpenAI.Chat.Completions.ChatCompletionTool[]),\n    ];\n\n    return mapped.length > 0 ? { tools: mapped } : {};\n  }\n\n  /** Map the neutral cache hint only for the direct OpenAI provider. */\n  private promptCacheBreakpoints(options: ModelCallOptions | undefined): number {\n    return this.provider === \"openai\" ? (options?.cacheControl?.breakpoints ?? 0) : 0;\n  }\n\n  /**\n   * Translate the neutral `responseSchema` option into OpenAI's\n   * `response_format` parameter. The mode decision (override handling,\n   * strict-compatibility downgrade to `json_object`) lives in\n   * `planStructuredOutput`, shared with the Responses adapter.\n   *\n   * Returns an empty spread when no schema was supplied, so the caller\n   * can unconditionally `...buildResponseFormat(...)` into the request.\n   */\n  private buildResponseFormat(responseSchema: Record<string, unknown> | undefined): {\n    response_format?: OpenAI.Chat.Completions.ChatCompletionCreateParams[\"response_format\"];\n  } {\n    const plan = planStructuredOutput(responseSchema, this.config.responseFormat);\n\n    if (plan.mode === \"json_schema\") {\n      return {\n        response_format: {\n          type: \"json_schema\",\n          json_schema: { name: \"response\", schema: plan.schema, strict: true },\n        },\n      };\n    }\n\n    if (plan.mode === \"json_object\") {\n      return { response_format: { type: \"json_object\" } };\n    }\n\n    return {};\n  }\n\n  /**\n   * Normalize OpenAI's `usage` block (which may be absent on some responses\n   * or partials) into the neutral `Usage` shape. Missing usage collapses to\n   * zeros rather than propagating `undefined`, so downstream aggregation\n   * math stays safe.\n   *\n   * `cachedTokens` mirrors `prompt_tokens_details.cached_tokens` (the\n   * subset of the prompt served from OpenAI's automatic prompt cache);\n   * `reasoningTokens` mirrors `completion_tokens_details.reasoning_tokens`\n   * (the hidden reasoning channel on o-series / gpt-5 models, already\n   * counted within `output`). Both are emitted only when the provider\n   * reports a positive value, so non-reasoning / uncached calls keep the\n   * lean `{ input, output, total }` shape. `cacheWriteTokens` mirrors\n   * `prompt_tokens_details.cache_write_tokens` (billed writes on the\n   * models that charge for them). `cost` is attached when priced.\n   */\n  private extractUsage(raw: OpenAI.Completions.CompletionUsage | undefined): Usage {\n    if (!raw) {\n      return { input: 0, output: 0, total: 0 };\n    }\n\n    return buildUsage(\n      {\n        input: raw.prompt_tokens ?? 0,\n        output: raw.completion_tokens ?? 0,\n        total: raw.total_tokens ?? 0,\n        cachedTokens: raw.prompt_tokens_details?.cached_tokens,\n        cacheWriteTokens: raw.prompt_tokens_details?.cache_write_tokens,\n        reasoningTokens: raw.completion_tokens_details?.reasoning_tokens,\n      },\n      this.pricing,\n    );\n  }\n\n  /**\n   * Translate the neutral `ModelCallOptions.reasoning` hint into OpenAI's\n   * `reasoning_effort` request param. Only `effort` maps — OpenAI's Chat\n   * Completions API exposes a discrete effort knob, not a token budget,\n   * so `reasoning.maxTokens` (the Anthropic extended-thinking cap) has no\n   * wire equivalent here and is silently ignored.\n   *\n   * The neutral `ReasoningEffort` (`\"low\" | \"medium\" | \"high\" | \"none\"`)\n   * is a subset of OpenAI's accepted values, so it forwards verbatim —\n   * `\"none\"` included. `\"none\"` is load-bearing: gpt-5 / o-series models\n   * **reject function tools** on Chat Completions while reasoning is\n   * active, and the endpoint accepts tools only when `reasoning_effort`\n   * is `\"none\"` (the alternative is the Responses API). This is why the\n   * param is EMITTED for `\"none\"` rather than omitted — omitting it\n   * leaves the model reasoning server-side by default, so tools would\n   * still be rejected.\n   *\n   * No-ops when the model is not reasoning-capable\n   * (`capabilities.reasoning` is false — e.g. `gpt-4o`, which 400s on any\n   * `reasoning_effort`).\n   *\n   * When the caller gave no explicit `effort`: defaults to `\"none\"` IF\n   * `hasTools` is true, else omits the param (provider default — the\n   * pre-existing behavior for a tool-less call). The default exists\n   * because, on a reasoning-capable model, there is no working alternative\n   * to `\"none\"` when tools are attached — omitting `reasoning_effort`\n   * leaves reasoning on server-side, which Chat Completions rejects tools\n   * for (empty replies on some model generations, a hard 400 — \"Function\n   * tools with reasoning_effort are not supported ... in\n   * /v1/chat/completions\" — on newer ones). An explicit `effort` from the\n   * caller always wins over this default, in either direction.\n   *\n   * Returns an empty spread when nothing applies, so the caller can\n   * unconditionally `...buildReasoningParams(...)` into the request.\n   */\n  /**\n   * Output-length and sampling params, in current Chat Completions shapes only.\n   *\n   * - `max_completion_tokens`, never the legacy `max_tokens`: current\n   *   models (the gpt-5 family and later) answer `400 unsupported_parameter`\n   *   to `max_tokens`, and `max_completion_tokens` is accepted everywhere.\n   * - No `temperature` for a reasoning-capable model: those accept only the\n   *   default value and reject any other with a 400.\n   */\n  private buildSamplingParams(options: ModelCallOptions | undefined): {\n    max_completion_tokens?: number;\n    temperature?: number;\n  } {\n    const maxTokens = options?.maxTokens ?? this.config.maxTokens;\n    const temperature = options?.temperature ?? this.config.temperature;\n\n    return {\n      max_completion_tokens: maxTokens,\n      ...(!this.capabilities.reasoning && { temperature }),\n    };\n  }\n\n  private buildReasoningParams(\n    reasoning: ModelCallOptions[\"reasoning\"],\n    hasTools: boolean,\n  ): {\n    reasoning_effort?: OpenAI.Chat.Completions.ChatCompletionCreateParams[\"reasoning_effort\"];\n  } {\n    if (!this.capabilities.reasoning) {\n      return {};\n    }\n\n    if (!reasoning?.effort) {\n      return hasTools ? { reasoning_effort: \"none\" } : {};\n    }\n\n    return { reasoning_effort: toOpenAIReasoningEffort(reasoning.effort) };\n  }\n\n  /**\n   * Reshape OpenAI's `tool_calls` array into the neutral\n   * `ModelToolCallRequest[]`. The raw `arguments` field is a JSON string\n   * per OpenAI's protocol — we parse it defensively via `safeJsonParse` so\n   * malformed or empty arguments yield an empty object instead of crashing\n   * the trip. Returns `undefined` when no tools were requested so callers\n   * can branch on presence.\n   */\n  private extractToolCalls(\n    rawToolCalls: OpenAI.Chat.Completions.ChatCompletionMessageToolCall[] | undefined,\n  ): ModelToolCallRequest[] | undefined {\n    if (!rawToolCalls || rawToolCalls.length === 0) {\n      return undefined;\n    }\n\n    return rawToolCalls.map((toolCall) => ({\n      id: toolCall.id,\n      name: (toolCall as any).function.name,\n      input: safeJsonParse<Record<string, unknown>>((toolCall as any).function.arguments, {}),\n    }));\n  }\n}\n\n/**\n * Reasoning text an OpenAI-compatible vendor puts on a Chat Completions\n * stream delta. OpenAI itself streams none on this API. DeepSeek documents\n * `delta.reasoning_content` (API reference, chat completion chunk); Groq\n * documents the parsed reasoning under a `reasoning` field. Neither is in\n * the OpenAI SDK's delta type, so the delta is read as a plain record.\n */\nfunction readReasoningDelta(delta: object): string | undefined {\n  const record = delta as Record<string, unknown>;\n\n  for (const key of [\"reasoning_content\", \"reasoning\"]) {\n    const value = record[key];\n\n    if (typeof value === \"string\" && value !== \"\") {\n      return value;\n    }\n  }\n\n  return undefined;\n}\n"],"mappings":";;;;;;;;;;;;;;;;AAkCA,MAAM,aAAa;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAuCnB,IAAa,cAAb,MAAkD;CAUhD,AAAO,YAAY,QAAgB,QAA2B,WAAmB,UAAU;gBAFzD;EAMhC,IAAI,OAAO,QAAQ,aACjB,MAAM,IAAI,oBACR,4JACA,EAAE,SAAS;GAAE;GAAU,OAAO,OAAO;GAAM,KAAK,OAAO;EAAI,EAAE,CAC/D;EAGF,KAAK,SAAS;EACd,KAAK,SAAS;EACd,KAAK,OAAO,OAAO;EACnB,KAAK,WAAW;EAChB,KAAK,UAAU,OAAO;EACtB,KAAK,eAAe;GAClB,kBAAkB,OAAO,oBAAoB,sBAAsB,OAAO,cAAc;GACxF,QAAQ,OAAO,UAAU,sBAAsB,OAAO,IAAI;GAI1D,WAAW,OAAO,aAAa,yBAAyB,OAAO,IAAI;GAInE,eAAe,KAAK,aAAa;GAMjC,KAAK,OAAO,OAAO;GACnB,OAAO,OAAO,SAAS;EACzB;CACF;;;;;;;CAQA,MAAa,SAAS,UAAqB,SAAoD;EAK7F,KAAK,OAAO,MAAM,YAAY,WAAW,qCAAqC;GAC5E,OAAO,KAAK;GACZ,cAAc,SAAS;GACvB,WAAW;GACX,WAAW,SAAS,OAAO,UAAU;EACvC,CAAC;EAED,IAAI;EAEJ,IAAI;GACF,WAAW,MAAM,KAAK,OAAO,KAAK,YAAY,OAC5C,kBACE;IACE,OAAO,KAAK;IACZ,UAAU,iBAAiB,UAAU,KAAK,uBAAuB,OAAO,CAAC;IACzE,GAAG,KAAK,oBAAoB,OAAO;IACnC,GAAG,uBAAuB,KAAK,UAAU,KAAK,MAAM;IACpD,GAAG,KAAK,WAAW,OAAO;IAC1B,GAAG,KAAK,oBAAoB,SAAS,cAAc;IACnD,GAAG,KAAK,qBAAqB,SAAS,WAAW,QAAQ,SAAS,OAAO,MAAM,CAAC;GAClF,GACA,OACF,GACA,SAAS,SAAS,EAAE,QAAQ,QAAQ,OAAO,IAAI,MACjD;EACF,SAAS,QAAQ;GACf,MAAM,UAAU,gBAAgB,MAAM;GAEtC,KAAK,OAAO,MAAM,YAAY,SAAS,QAAQ,SAAS;IACtD,MAAM,QAAQ;IACd,SAAS,QAAQ;GACnB,CAAC;GAED,MAAM;EACR;EAEA,MAAM,CAAC,UAAU,SAAS;EAE1B,IAAI,WAAW,QASb,MAAM,IAAI,cAAc,+CAA+C,EACrE,SAAS;GAAE,UAAU;GAAU,OAAO,KAAK;GAAM,SAAS,SAAS,QAAQ;EAAO,EACpF,CAAC;EAGH,MAAM,eAAe,gBAAgB,OAAO,aAAa;EACzD,MAAM,QAAQ,KAAK,aAAa,SAAS,KAAK;EAE9C,KAAK,OAAO,MAAM,YAAY,YAAY,sCAAsC;GAC9E;GACA;EACF,CAAC;EAED,OAAO;GACL,SAAS,OAAO,QAAQ,WAAW;GACnC;GACA;GACA,WAAW,KAAK,iBAAiB,OAAO,QAAQ,UAAU;EAC5D;CACF;;;;;;;;;;;;CAaA,OAAc,OACZ,UACA,SACiC;EACjC,KAAK,OAAO,MAAM,YAAY,WAAW,+CAA+C;GACtF,OAAO,KAAK;GACZ,cAAc,SAAS;GACvB,WAAW;GACX,WAAW,SAAS,OAAO,UAAU;EACvC,CAAC;EAED,IAAI;EAEJ,IAAI;GACF,SAAS,MAAM,KAAK,OAAO,KAAK,YAAY,OAC1C,kBACE;IACE,OAAO,KAAK;IACZ,UAAU,iBAAiB,UAAU,KAAK,uBAAuB,OAAO,CAAC;IACzE,GAAG,KAAK,oBAAoB,OAAO;IACnC,GAAG,uBAAuB,KAAK,UAAU,KAAK,MAAM;IACpD,GAAG,KAAK,WAAW,OAAO;IAC1B,QAAQ;IACR,gBAAgB,EAAE,eAAe,KAAK;IACtC,GAAG,KAAK,oBAAoB,SAAS,cAAc;IACnD,GAAG,KAAK,qBAAqB,SAAS,WAAW,QAAQ,SAAS,OAAO,MAAM,CAAC;GAClF,GACA,OACF,GACA,SAAS,SAAS,EAAE,QAAQ,QAAQ,OAAO,IAAI,MACjD;EACF,SAAS,QAAQ;GACf,MAAM,UAAU,gBAAgB,MAAM;GAEtC,KAAK,OAAO,MAAM,YAAY,SAAS,QAAQ,SAAS;IACtD,MAAM,QAAQ;IACd,SAAS,QAAQ;GACnB,CAAC;GAED,MAAM;EACR;EAEA,IAAI,kBAA0B;EAC9B,IAAI,QAAe;GAAE,OAAO;GAAG,QAAQ;GAAG,OAAO;EAAE;EACnD,MAAM,OAAO,oBAAoB,OAAO;EACxC,MAAM,gCAAgB,IAAI,IAGxB;EAEF,IAAI;GACF,WAAW,MAAM,SAAS,QAAsE;IAC9F,MAAM,QAAQ,MAAM,QAAQ,EAAE,EAAE;IAChC,MAAM,SAAS,MAAM,QAAQ,EAAE,EAAE;IAEjC,IAAI,KAAK,aAAa,OAAO;KAC3B,MAAM,YAAY,mBAAmB,KAAK;KAE1C,IAAI,WACF,MAAM;MAAE,MAAM;MAAa,OAAO;KAAU;IAEhD;IAEA,IAAI,OAAO,SACT,MAAM;KAAE,MAAM;KAAS,SAAS,MAAM;IAAQ;IAGhD,IAAI,OAAO,YACT,KAAK,MAAM,YAAY,MAAM,YAAY;KACvC,MAAM,MAAM,SAAS,SAAS;KAC9B,IAAI,CAAC,cAAc,IAAI,GAAG,GACxB,cAAc,IAAI,KAAK;MAAE,IAAI;MAAI,MAAM;MAAI,WAAW;MAAI,UAAU;KAAE,CAAC;KAEzE,MAAM,MAAM,cAAc,IAAI,GAAG;KACjC,IAAI,SAAS,IAAI,IAAI,KAAK,SAAS;KACnC,IAAI,SAAS,UAAU,MAAM,IAAI,OAAO,SAAS,SAAS;KAC1D,IAAI,SAAS,UAAU,WAAW,IAAI,aAAa,SAAS,SAAS;KAKrE,IAAI,KAAK,YAAY,IAAI,QAAQ,IAAI,UAAU,SAAS,IAAI,UAAU;MACpE,MAAM;OACJ,MAAM;OACN,QAAQ,IAAI;OACZ,MAAM,IAAI;OACV,OAAO,IAAI,UAAU,MAAM,IAAI,QAAQ;MACzC;MACA,IAAI,WAAW,IAAI,UAAU;KAC/B;IACF;IAGF,IAAI,QACF,kBAAkB;IAGpB,IAAI,MAAM,OACR,QAAQ,KAAK,aAAa,MAAM,KAAK;GAEzC;GAEA,KAAK,MAAM,OAAO,cAAc,OAAO,GAAG;IAMxC,IAAI,CAAC,IAAI,MAAM;IAEf,MAAM;KACJ,MAAM;KACN,IAAI,IAAI;KACR,MAAM,IAAI;KACV,OAAO,cAAuC,IAAI,WAAW,CAAC,CAAC;IACjE;GACF;EACF,SAAS,QAAQ;GACf,MAAM,UAAU,gBAAgB,MAAM;GAEtC,KAAK,OAAO,MAAM,YAAY,SAAS,QAAQ,SAAS;IACtD,MAAM,QAAQ;IACd,SAAS,QAAQ;GACnB,CAAC;GAED,MAAM;EACR;EAEA,MAAM,eAAe,gBAAgB,eAAe;EAEpD,KAAK,OAAO,MAAM,YAAY,YAAY,gDAAgD;GACxF;GACA;EACF,CAAC;EAED,MAAM;GAAE,MAAM;GAAQ;GAAc;EAAM;CAC5C;;;;;;;;;;CAWA,AAAQ,WAAW,SAEjB;EACA,MAAM,EAAE,OAAO,gBAAgB,oBAAoB,SAAS;GAC1D,SAAS,CAAC,KAAK,QAAQ;GACvB,QAAQ;GACR,OAAO,KAAK;GACZ,QAAQ,KAAK;EACf,CAAC;EACD,MAAM,SAAS,CACb,GAAI,cAAc,KAAK,KAAK,CAAC,GAC7B,GAAI,WACN;EAEA,OAAO,OAAO,SAAS,IAAI,EAAE,OAAO,OAAO,IAAI,CAAC;CAClD;;CAGA,AAAQ,uBAAuB,SAA+C;EAC5E,OAAO,KAAK,aAAa,WAAY,SAAS,cAAc,eAAe,IAAK;CAClF;;;;;;;;;;CAWA,AAAQ,oBAAoB,gBAE1B;EACA,MAAM,OAAO,qBAAqB,gBAAgB,KAAK,OAAO,cAAc;EAE5E,IAAI,KAAK,SAAS,eAChB,OAAO,EACL,iBAAiB;GACf,MAAM;GACN,aAAa;IAAE,MAAM;IAAY,QAAQ,KAAK;IAAQ,QAAQ;GAAK;EACrE,EACF;EAGF,IAAI,KAAK,SAAS,eAChB,OAAO,EAAE,iBAAiB,EAAE,MAAM,cAAc,EAAE;EAGpD,OAAO,CAAC;CACV;;;;;;;;;;;;;;;;;CAkBA,AAAQ,aAAa,KAA4D;EAC/E,IAAI,CAAC,KACH,OAAO;GAAE,OAAO;GAAG,QAAQ;GAAG,OAAO;EAAE;EAGzC,OAAO,WACL;GACE,OAAO,IAAI,iBAAiB;GAC5B,QAAQ,IAAI,qBAAqB;GACjC,OAAO,IAAI,gBAAgB;GAC3B,cAAc,IAAI,uBAAuB;GACzC,kBAAkB,IAAI,uBAAuB;GAC7C,iBAAiB,IAAI,2BAA2B;EAClD,GACA,KAAK,OACP;CACF;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;CA8CA,AAAQ,oBAAoB,SAG1B;EACA,MAAM,YAAY,SAAS,aAAa,KAAK,OAAO;EACpD,MAAM,cAAc,SAAS,eAAe,KAAK,OAAO;EAExD,OAAO;GACL,uBAAuB;GACvB,GAAI,CAAC,KAAK,aAAa,aAAa,EAAE,YAAY;EACpD;CACF;CAEA,AAAQ,qBACN,WACA,UAGA;EACA,IAAI,CAAC,KAAK,aAAa,WACrB,OAAO,CAAC;EAGV,IAAI,CAAC,WAAW,QACd,OAAO,WAAW,EAAE,kBAAkB,OAAO,IAAI,CAAC;EAGpD,OAAO,EAAE,kBAAkB,wBAAwB,UAAU,MAAM,EAAE;CACvE;;;;;;;;;CAUA,AAAQ,iBACN,cACoC;EACpC,IAAI,CAAC,gBAAgB,aAAa,WAAW,GAC3C;EAGF,OAAO,aAAa,KAAK,cAAc;GACrC,IAAI,SAAS;GACb,MAAO,SAAiB,SAAS;GACjC,OAAO,cAAwC,SAAiB,SAAS,WAAW,CAAC,CAAC;EACxF,EAAE;CACJ;AACF;;;;;;;;AASA,SAAS,mBAAmB,OAAmC;CAC7D,MAAM,SAAS;CAEf,KAAK,MAAM,OAAO,CAAC,qBAAqB,WAAW,GAAG;EACpD,MAAM,QAAQ,OAAO;EAErB,IAAI,OAAO,UAAU,YAAY,UAAU,IACzC,OAAO;CAEX;AAGF"}

package/esm/provider-options.d.mts

@@ -0,0 +1,8 @@
+//#region ../../packages/ai-openai/src/provider-options.d.ts
+/** Adapter-only fields merged after the neutral request parameters. */
+type OpenAICallOptions = {
+  body?: Record<string, unknown>;
+};
+//#endregion
+export { OpenAICallOptions };
+//# sourceMappingURL=provider-options.d.mts.map

package/esm/provider-options.d.mts.map

@@ -0,0 +1 @@
+{"version":3,"file":"provider-options.d.mts","names":[],"sources":["../../../../../../../packages/ai-openai/src/provider-options.ts"],"mappings":";;KAGY,iBAAA;EAAsB,IAAA,GAAO,MAAM;AAAA"}