@semiont/inference 0.5.8 → 0.5.10

package/README.md

@@ -111,7 +111,7 @@ const items = JSON.parse(json); // guaranteed to be an array
 
 Each implementation honors the contract with its provider's mechanism:
 - **Ollama**: grammar-constrained sampling — the request's `format` field carries a minimal array schema.
-- **Anthropic**: assistant-turn prefill (`[`); the prefix is re-attached to the returned text so callers always see a complete JSON document.
+- **Anthropic**: forced structured tool-use — a single tool is offered and forced via `tool_choice`, so the model answers by filling the tool's input, which the API serializes as escaped JSON. The array is carried under an `items` property (tool inputs must be objects) and unwrapped to a top-level array on return.
 
 Current callers all expect arrays (entity extraction, motivation detection). If an object-emitting caller appears, the option grows a `root: 'array' | 'object'` field — see the notes in [src/interface.ts](src/interface.ts).
 

package/dist/index.d.ts

@@ -14,10 +14,11 @@ interface InferenceOptions {
      * Constrain output to a parseable JSON array. Every implementation
      * MUST satisfy this contract using whatever mechanism its provider
      * supports — Ollama uses grammar-constrained sampling
-     * (`format: "json"`); Anthropic uses assistant-turn prefill (`[`)
-     * and re-attaches the prefix on return. Callers can rely on the
-     * returned `text` being a top-level JSON array regardless of
-     * provider.
+     * (`format: "json"`); Anthropic uses forced structured tool-use (a
+     * single tool, forced via `tool_choice`, whose array result the API
+     * serializes as escaped JSON) and unwraps the array on return.
+     * Callers can rely on the returned `text` being a top-level JSON
+     * array regardless of provider.
      *
      * Current callers all expect arrays (entity extraction, motivation
      * detection). If an object-emitting caller appears, this option

package/dist/index.js

@@ -1,6 +1,19 @@
 // src/implementations/anthropic.ts
 import Anthropic from "@anthropic-ai/sdk";
 import { recordInferenceUsage } from "@semiont/observability";
+var JSON_ARRAY_TOOL = {
+  name: "emit_json_array",
+  description: 'Return your entire answer by calling this tool. Put the JSON array of results under the "items" property, and emit no prose.',
+  input_schema: {
+    type: "object",
+    properties: {
+      // Element shape is unconstrained here — the prompt carries the per-element
+      // schema; the tool only enforces that the top-level result is an array.
+      items: { type: "array", items: {} }
+    },
+    required: ["items"]
+  }
+};
 var AnthropicInferenceClient = class {
   type = "anthropic";
   modelId;
@@ -20,7 +33,6 @@ var AnthropicInferenceClient = class {
   }
   async generateTextWithMetadata(prompt, maxTokens, temperature, options) {
     const jsonMode = options?.format === "json";
-    const prefill = jsonMode ? "[" : void 0;
     this.logger?.debug("Generating text with inference client", {
       model: this.modelId,
       promptLength: prompt.length,
@@ -28,12 +40,6 @@ var AnthropicInferenceClient = class {
       temperature,
       format: options?.format
     });
-    const messages = [
-      { role: "user", content: prompt }
-    ];
-    if (prefill) {
-      messages.push({ role: "assistant", content: prefill });
-    }
     const start = performance.now();
     let response;
     try {
@@ -41,7 +47,10 @@ var AnthropicInferenceClient = class {
         model: this.modelId,
         max_tokens: maxTokens,
         temperature,
-        messages
+        messages: [{ role: "user", content: prompt }],
+        // JSON mode → force the structured-output tool. No prefill assistant
+        // turn: the constraint now lives in the tool call, not in free text.
+        ...jsonMode ? { tools: [JSON_ARRAY_TOOL], tool_choice: { type: "tool", name: JSON_ARRAY_TOOL.name } } : {}
       });
     } catch (err) {
       recordInferenceUsage({
@@ -57,21 +66,45 @@ var AnthropicInferenceClient = class {
       contentBlocks: response.content.length,
       stopReason: response.stop_reason
     });
-    const textContent = response.content.find((c) => c.type === "text");
-    if (!textContent || textContent.type !== "text") {
-      recordInferenceUsage({
-        provider: this.type,
-        model: this.modelId,
-        durationMs: performance.now() - start,
-        outcome: "error",
-        inputTokens: response.usage?.input_tokens,
-        outputTokens: response.usage?.output_tokens
-      });
-      this.logger?.error("No text content in inference response", {
-        model: this.modelId,
-        contentTypes: response.content.map((c) => c.type)
-      });
-      throw new Error("No text content in inference response");
+    let text;
+    if (jsonMode) {
+      const toolUse = response.content.find((c) => c.type === "tool_use");
+      if (!toolUse || toolUse.type !== "tool_use") {
+        recordInferenceUsage({
+          provider: this.type,
+          model: this.modelId,
+          durationMs: performance.now() - start,
+          outcome: "error",
+          inputTokens: response.usage?.input_tokens,
+          outputTokens: response.usage?.output_tokens
+        });
+        this.logger?.error("No tool_use content in inference response", {
+          model: this.modelId,
+          contentTypes: response.content.map((c) => c.type)
+        });
+        throw new Error("No tool_use content in inference response");
+      }
+      const input = toolUse.input;
+      const items = Array.isArray(input.items) ? input.items : [];
+      text = JSON.stringify(items);
+    } else {
+      const textContent = response.content.find((c) => c.type === "text");
+      if (!textContent || textContent.type !== "text") {
+        recordInferenceUsage({
+          provider: this.type,
+          model: this.modelId,
+          durationMs: performance.now() - start,
+          outcome: "error",
+          inputTokens: response.usage?.input_tokens,
+          outputTokens: response.usage?.output_tokens
+        });
+        this.logger?.error("No text content in inference response", {
+          model: this.modelId,
+          contentTypes: response.content.map((c) => c.type)
+        });
+        throw new Error("No text content in inference response");
+      }
+      text = textContent.text;
     }
     recordInferenceUsage({
       provider: this.type,
@@ -81,7 +114,6 @@ var AnthropicInferenceClient = class {
       inputTokens: response.usage?.input_tokens,
       outputTokens: response.usage?.output_tokens
     });
-    const text = prefill ? prefill + textContent.text : textContent.text;
     this.logger?.info("Text generation completed", {
       model: this.modelId,
       textLength: text.length,

package/dist/index.js.map

@@ -1 +1 @@
-{"version":3,"sources":["../src/implementations/anthropic.ts","../src/implementations/ollama.ts","../src/factory.ts","../src/implementations/mock.ts"],"sourcesContent":["// Anthropic Claude implementation of InferenceClient interface\n\nimport Anthropic from '@anthropic-ai/sdk';\nimport type { Logger } from '@semiont/core';\nimport { recordInferenceUsage } from '@semiont/observability';\nimport { InferenceClient, InferenceOptions, InferenceResponse } from '../interface.js';\n\nexport class AnthropicInferenceClient implements InferenceClient {\n  readonly type = 'anthropic' as const;\n  readonly modelId: string;\n  private client: Anthropic;\n  private logger?: Logger;\n\n  constructor(apiKey: string, model: string, baseURL?: string, logger?: Logger) {\n    this.client = new Anthropic({\n      apiKey,\n      baseURL: baseURL || 'https://api.anthropic.com',\n    });\n    this.modelId = model;\n    this.logger = logger;\n  }\n\n  async generateText(prompt: string, maxTokens: number, temperature: number, options?: InferenceOptions): Promise<string> {\n    const response = await this.generateTextWithMetadata(prompt, maxTokens, temperature, options);\n    return response.text;\n  }\n\n  async generateTextWithMetadata(prompt: string, maxTokens: number, temperature: number, options?: InferenceOptions): Promise<InferenceResponse> {\n    // Anthropic has no grammar-constrained sampling layer like Ollama's\n    // `format: \"json\"`. The closest equivalent is *prefill*: add an\n    // assistant turn whose content is the opening bracket of the\n    // expected structure. Claude continues from there, syntactically\n    // committed to producing valid JSON. The prefill characters don't\n    // appear in the response, so we prepend them ourselves on return.\n    //\n    // We assume `[` (array) — every current caller of the inference\n    // layer that asks for JSON expects an array (entity extraction,\n    // motivation detection). If an object-emitting caller appears, the\n    // option needs to grow a `root: 'array' | 'object'` field; for now\n    // a single shape keeps the contract small.\n    const jsonMode = options?.format === 'json';\n    const prefill = jsonMode ? '[' : undefined;\n\n    this.logger?.debug('Generating text with inference client', {\n      model: this.modelId,\n      promptLength: prompt.length,\n      maxTokens,\n      temperature,\n      format: options?.format,\n    });\n\n    const messages: Array<{ role: 'user' | 'assistant'; content: string }> = [\n      { role: 'user', content: prompt },\n    ];\n    if (prefill) {\n      messages.push({ role: 'assistant', content: prefill });\n    }\n\n    const start = performance.now();\n    let response: Awaited<ReturnType<typeof this.client.messages.create>>;\n    try {\n      response = await this.client.messages.create({\n        model: this.modelId,\n        max_tokens: maxTokens,\n        temperature,\n        messages,\n      });\n    } catch (err) {\n      recordInferenceUsage({\n        provider: this.type,\n        model: this.modelId,\n        durationMs: performance.now() - start,\n        outcome: 'error',\n      });\n      throw err;\n    }\n\n    this.logger?.debug('Inference response received', {\n      model: this.modelId,\n      contentBlocks: response.content.length,\n      stopReason: response.stop_reason\n    });\n\n    const textContent = response.content.find(c => c.type === 'text');\n\n    if (!textContent || textContent.type !== 'text') {\n      recordInferenceUsage({\n        provider: this.type,\n        model: this.modelId,\n        durationMs: performance.now() - start,\n        outcome: 'error',\n        inputTokens: response.usage?.input_tokens,\n        outputTokens: response.usage?.output_tokens,\n      });\n      this.logger?.error('No text content in inference response', {\n        model: this.modelId,\n        contentTypes: response.content.map(c => c.type)\n      });\n      throw new Error('No text content in inference response');\n    }\n\n    recordInferenceUsage({\n      provider: this.type,\n      model: this.modelId,\n      durationMs: performance.now() - start,\n      outcome: 'success',\n      inputTokens: response.usage?.input_tokens,\n      outputTokens: response.usage?.output_tokens,\n    });\n\n    // Re-attach the prefill prefix so the returned text is what the caller\n    // *would have seen* if Anthropic had a native JSON mode — a complete,\n    // parseable JSON document. Without this, the consumer would get the\n    // tail (`{...}, {...}]`) and parse-fail trying to read it as a top-\n    // level array.\n    const text = prefill ? prefill + textContent.text : textContent.text;\n\n    this.logger?.info('Text generation completed', {\n      model: this.modelId,\n      textLength: text.length,\n      stopReason: response.stop_reason\n    });\n\n    return {\n      text,\n      stopReason: response.stop_reason || 'unknown'\n    };\n  }\n}\n","// Ollama implementation of InferenceClient interface\n// Uses native Ollama HTTP API (no SDK dependency)\n\nimport type { Logger } from '@semiont/core';\nimport { recordInferenceUsage } from '@semiont/observability';\nimport { InferenceClient, InferenceOptions, InferenceResponse } from '../interface.js';\n\ninterface OllamaGenerateResponse {\n  response: string;\n  done: boolean;\n  done_reason?: string;\n  /** Number of prompt tokens evaluated. Available on most Ollama versions. */\n  prompt_eval_count?: number;\n  /** Number of tokens generated. */\n  eval_count?: number;\n}\n\nexport class OllamaInferenceClient implements InferenceClient {\n  readonly type = 'ollama' as const;\n  readonly modelId: string;\n  private baseURL: string;\n  private logger?: Logger;\n\n  constructor(model: string, baseURL?: string, logger?: Logger) {\n    this.baseURL = (baseURL || 'http://localhost:11434').replace(/\\/+$/, '');\n    this.modelId = model;\n    this.logger = logger;\n  }\n\n  async generateText(prompt: string, maxTokens: number, temperature: number, options?: InferenceOptions): Promise<string> {\n    const response = await this.generateTextWithMetadata(prompt, maxTokens, temperature, options);\n    return response.text;\n  }\n\n  async generateTextWithMetadata(prompt: string, maxTokens: number, temperature: number, options?: InferenceOptions): Promise<InferenceResponse> {\n    this.logger?.debug('Generating text with Ollama', {\n      model: this.modelId,\n      promptLength: prompt.length,\n      maxTokens,\n      temperature,\n      format: options?.format,\n    });\n\n    const url = `${this.baseURL}/api/generate`;\n    const start = performance.now();\n\n    // Ollama's `format` parameter accepts either the literal string\n    // `\"json\"` (any valid JSON, including objects, numbers, etc.) or a\n    // JSON schema (constrains the top-level shape). The contract on the\n    // inference side is \"parseable JSON array,\" so we pass a minimal\n    // array schema rather than the bare `\"json\"` string — without it,\n    // the model can satisfy \"valid JSON\" by emitting `{\"entities\": [...]}`\n    // and break every consumer that expects to call `.map` on the\n    // top-level value. The schema's `items: {}` keeps element shape\n    // unconstrained — the prompt still carries the per-element schema;\n    // we only enforce the outer array.\n    const body: Record<string, unknown> = {\n      model: this.modelId,\n      prompt,\n      stream: false,\n      think: false,\n      options: {\n        num_predict: maxTokens,\n        temperature,\n      },\n    };\n    if (options?.format === 'json') {\n      body['format'] = { type: 'array', items: {} };\n    }\n\n    let res: Response;\n    try {\n      res = await fetch(url, {\n        method: 'POST',\n        headers: { 'Content-Type': 'application/json' },\n        body: JSON.stringify(body),\n      });\n    } catch (err) {\n      recordInferenceUsage({\n        provider: this.type,\n        model: this.modelId,\n        durationMs: performance.now() - start,\n        outcome: 'error',\n      });\n      throw err;\n    }\n\n    if (!res.ok) {\n      recordInferenceUsage({\n        provider: this.type,\n        model: this.modelId,\n        durationMs: performance.now() - start,\n        outcome: 'error',\n      });\n      const body = await res.text();\n      this.logger?.error('Ollama API error', {\n        model: this.modelId,\n        status: res.status,\n        body,\n      });\n      throw new Error(`Ollama API error (${res.status}): ${body}`);\n    }\n\n    const data = await res.json() as OllamaGenerateResponse;\n\n    if (!data.response) {\n      recordInferenceUsage({\n        provider: this.type,\n        model: this.modelId,\n        durationMs: performance.now() - start,\n        outcome: 'error',\n        inputTokens: data.prompt_eval_count,\n        outputTokens: data.eval_count,\n      });\n      this.logger?.error('Empty response from Ollama', { model: this.modelId });\n      throw new Error('Empty response from Ollama');\n    }\n\n    recordInferenceUsage({\n      provider: this.type,\n      model: this.modelId,\n      durationMs: performance.now() - start,\n      outcome: 'success',\n      inputTokens: data.prompt_eval_count,\n      outputTokens: data.eval_count,\n    });\n\n    const stopReason = mapStopReason(data.done_reason);\n\n    this.logger?.info('Text generation completed', {\n      model: this.modelId,\n      textLength: data.response.length,\n      stopReason,\n    });\n\n    return {\n      text: data.response,\n      stopReason,\n    };\n  }\n}\n\nfunction mapStopReason(doneReason: string | undefined): string {\n  switch (doneReason) {\n    case 'stop': return 'end_turn';\n    case 'length': return 'max_tokens';\n    default: return doneReason || 'unknown';\n  }\n}\n","// Factory for creating inference client instances based on configuration\n\nimport type { Logger } from '@semiont/core';\nimport { InferenceClient } from './interface.js';\nimport { AnthropicInferenceClient } from './implementations/anthropic.js';\nimport { OllamaInferenceClient } from './implementations/ollama.js';\n\nexport type InferenceClientType = 'anthropic' | 'ollama';\n\nexport interface InferenceClientConfig {\n  type: InferenceClientType;\n  apiKey?: string;\n  model: string;\n  endpoint?: string;\n  baseURL?: string;\n}\n\nexport function createInferenceClient(config: InferenceClientConfig, logger?: Logger): InferenceClient {\n  switch (config.type) {\n    case 'anthropic': {\n      if (!config.apiKey || config.apiKey.trim() === '') {\n        throw new Error('apiKey is required for Anthropic inference client');\n      }\n      return new AnthropicInferenceClient(\n        config.apiKey,\n        config.model,\n        config.endpoint || config.baseURL,\n        logger\n      );\n    }\n\n    case 'ollama': {\n      return new OllamaInferenceClient(\n        config.model,\n        config.endpoint || config.baseURL,\n        logger\n      );\n    }\n\n    default:\n      throw new Error(`Unsupported inference client type: ${config.type}`);\n  }\n}\n","// Mock implementation of InferenceClient for testing\n\nimport { InferenceClient, InferenceOptions, InferenceResponse } from '../interface.js';\n\nexport class MockInferenceClient implements InferenceClient {\n  readonly type = 'mock' as const;\n  readonly modelId = 'mock-model' as const;\n  private responses: string[] = [];\n  private responseIndex: number = 0;\n  private stopReasons: string[] = [];\n  public calls: Array<{ prompt: string; maxTokens: number; temperature: number; options?: InferenceOptions }> = [];\n\n  constructor(responses: string[] = ['Mock response'], stopReasons?: string[]) {\n    this.responses = responses;\n    this.stopReasons = stopReasons || responses.map(() => 'end_turn');\n  }\n\n  async generateText(prompt: string, maxTokens: number, temperature: number, options?: InferenceOptions): Promise<string> {\n    const response = await this.generateTextWithMetadata(prompt, maxTokens, temperature, options);\n    return response.text;\n  }\n\n  async generateTextWithMetadata(prompt: string, maxTokens: number, temperature: number, options?: InferenceOptions): Promise<InferenceResponse> {\n    this.calls.push({ prompt, maxTokens, temperature, ...(options ? { options } : {}) });\n\n    const text = this.responses[this.responseIndex];\n    const stopReason = this.stopReasons[this.responseIndex] || 'end_turn';\n\n    if (this.responseIndex < this.responses.length - 1) {\n      this.responseIndex++;\n    }\n\n    return { text, stopReason };\n  }\n\n  // Test helper methods\n  reset(): void {\n    this.calls = [];\n    this.responseIndex = 0;\n  }\n\n  setResponses(responses: string[], stopReasons?: string[]): void {\n    this.responses = responses;\n    this.stopReasons = stopReasons || responses.map(() => 'end_turn');\n    this.responseIndex = 0;\n  }\n}\n"],"mappings":";AAEA,OAAO,eAAe;AAEtB,SAAS,4BAA4B;AAG9B,IAAM,2BAAN,MAA0D;AAAA,EACtD,OAAO;AAAA,EACP;AAAA,EACD;AAAA,EACA;AAAA,EAER,YAAY,QAAgB,OAAe,SAAkB,QAAiB;AAC5E,SAAK,SAAS,IAAI,UAAU;AAAA,MAC1B;AAAA,MACA,SAAS,WAAW;AAAA,IACtB,CAAC;AACD,SAAK,UAAU;AACf,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,MAAM,aAAa,QAAgB,WAAmB,aAAqB,SAA6C;AACtH,UAAM,WAAW,MAAM,KAAK,yBAAyB,QAAQ,WAAW,aAAa,OAAO;AAC5F,WAAO,SAAS;AAAA,EAClB;AAAA,EAEA,MAAM,yBAAyB,QAAgB,WAAmB,aAAqB,SAAwD;AAa7I,UAAM,WAAW,SAAS,WAAW;AACrC,UAAM,UAAU,WAAW,MAAM;AAEjC,SAAK,QAAQ,MAAM,yCAAyC;AAAA,MAC1D,OAAO,KAAK;AAAA,MACZ,cAAc,OAAO;AAAA,MACrB;AAAA,MACA;AAAA,MACA,QAAQ,SAAS;AAAA,IACnB,CAAC;AAED,UAAM,WAAmE;AAAA,MACvE,EAAE,MAAM,QAAQ,SAAS,OAAO;AAAA,IAClC;AACA,QAAI,SAAS;AACX,eAAS,KAAK,EAAE,MAAM,aAAa,SAAS,QAAQ,CAAC;AAAA,IACvD;AAEA,UAAM,QAAQ,YAAY,IAAI;AAC9B,QAAI;AACJ,QAAI;AACF,iBAAW,MAAM,KAAK,OAAO,SAAS,OAAO;AAAA,QAC3C,OAAO,KAAK;AAAA,QACZ,YAAY;AAAA,QACZ;AAAA,QACA;AAAA,MACF,CAAC;AAAA,IACH,SAAS,KAAK;AACZ,2BAAqB;AAAA,QACnB,UAAU,KAAK;AAAA,QACf,OAAO,KAAK;AAAA,QACZ,YAAY,YAAY,IAAI,IAAI;AAAA,QAChC,SAAS;AAAA,MACX,CAAC;AACD,YAAM;AAAA,IACR;AAEA,SAAK,QAAQ,MAAM,+BAA+B;AAAA,MAChD,OAAO,KAAK;AAAA,MACZ,eAAe,SAAS,QAAQ;AAAA,MAChC,YAAY,SAAS;AAAA,IACvB,CAAC;AAED,UAAM,cAAc,SAAS,QAAQ,KAAK,OAAK,EAAE,SAAS,MAAM;AAEhE,QAAI,CAAC,eAAe,YAAY,SAAS,QAAQ;AAC/C,2BAAqB;AAAA,QACnB,UAAU,KAAK;AAAA,QACf,OAAO,KAAK;AAAA,QACZ,YAAY,YAAY,IAAI,IAAI;AAAA,QAChC,SAAS;AAAA,QACT,aAAa,SAAS,OAAO;AAAA,QAC7B,cAAc,SAAS,OAAO;AAAA,MAChC,CAAC;AACD,WAAK,QAAQ,MAAM,yCAAyC;AAAA,QAC1D,OAAO,KAAK;AAAA,QACZ,cAAc,SAAS,QAAQ,IAAI,OAAK,EAAE,IAAI;AAAA,MAChD,CAAC;AACD,YAAM,IAAI,MAAM,uCAAuC;AAAA,IACzD;AAEA,yBAAqB;AAAA,MACnB,UAAU,KAAK;AAAA,MACf,OAAO,KAAK;AAAA,MACZ,YAAY,YAAY,IAAI,IAAI;AAAA,MAChC,SAAS;AAAA,MACT,aAAa,SAAS,OAAO;AAAA,MAC7B,cAAc,SAAS,OAAO;AAAA,IAChC,CAAC;AAOD,UAAM,OAAO,UAAU,UAAU,YAAY,OAAO,YAAY;AAEhE,SAAK,QAAQ,KAAK,6BAA6B;AAAA,MAC7C,OAAO,KAAK;AAAA,MACZ,YAAY,KAAK;AAAA,MACjB,YAAY,SAAS;AAAA,IACvB,CAAC;AAED,WAAO;AAAA,MACL;AAAA,MACA,YAAY,SAAS,eAAe;AAAA,IACtC;AAAA,EACF;AACF;;;AC5HA,SAAS,wBAAAA,6BAA4B;AAa9B,IAAM,wBAAN,MAAuD;AAAA,EACnD,OAAO;AAAA,EACP;AAAA,EACD;AAAA,EACA;AAAA,EAER,YAAY,OAAe,SAAkB,QAAiB;AAC5D,SAAK,WAAW,WAAW,0BAA0B,QAAQ,QAAQ,EAAE;AACvE,SAAK,UAAU;AACf,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,MAAM,aAAa,QAAgB,WAAmB,aAAqB,SAA6C;AACtH,UAAM,WAAW,MAAM,KAAK,yBAAyB,QAAQ,WAAW,aAAa,OAAO;AAC5F,WAAO,SAAS;AAAA,EAClB;AAAA,EAEA,MAAM,yBAAyB,QAAgB,WAAmB,aAAqB,SAAwD;AAC7I,SAAK,QAAQ,MAAM,+BAA+B;AAAA,MAChD,OAAO,KAAK;AAAA,MACZ,cAAc,OAAO;AAAA,MACrB;AAAA,MACA;AAAA,MACA,QAAQ,SAAS;AAAA,IACnB,CAAC;AAED,UAAM,MAAM,GAAG,KAAK,OAAO;AAC3B,UAAM,QAAQ,YAAY,IAAI;AAY9B,UAAM,OAAgC;AAAA,MACpC,OAAO,KAAK;AAAA,MACZ;AAAA,MACA,QAAQ;AAAA,MACR,OAAO;AAAA,MACP,SAAS;AAAA,QACP,aAAa;AAAA,QACb;AAAA,MACF;AAAA,IACF;AACA,QAAI,SAAS,WAAW,QAAQ;AAC9B,WAAK,QAAQ,IAAI,EAAE,MAAM,SAAS,OAAO,CAAC,EAAE;AAAA,IAC9C;AAEA,QAAI;AACJ,QAAI;AACF,YAAM,MAAM,MAAM,KAAK;AAAA,QACrB,QAAQ;AAAA,QACR,SAAS,EAAE,gBAAgB,mBAAmB;AAAA,QAC9C,MAAM,KAAK,UAAU,IAAI;AAAA,MAC3B,CAAC;AAAA,IACH,SAAS,KAAK;AACZ,MAAAA,sBAAqB;AAAA,QACnB,UAAU,KAAK;AAAA,QACf,OAAO,KAAK;AAAA,QACZ,YAAY,YAAY,IAAI,IAAI;AAAA,QAChC,SAAS;AAAA,MACX,CAAC;AACD,YAAM;AAAA,IACR;AAEA,QAAI,CAAC,IAAI,IAAI;AACX,MAAAA,sBAAqB;AAAA,QACnB,UAAU,KAAK;AAAA,QACf,OAAO,KAAK;AAAA,QACZ,YAAY,YAAY,IAAI,IAAI;AAAA,QAChC,SAAS;AAAA,MACX,CAAC;AACD,YAAMC,QAAO,MAAM,IAAI,KAAK;AAC5B,WAAK,QAAQ,MAAM,oBAAoB;AAAA,QACrC,OAAO,KAAK;AAAA,QACZ,QAAQ,IAAI;AAAA,QACZ,MAAAA;AAAA,MACF,CAAC;AACD,YAAM,IAAI,MAAM,qBAAqB,IAAI,MAAM,MAAMA,KAAI,EAAE;AAAA,IAC7D;AAEA,UAAM,OAAO,MAAM,IAAI,KAAK;AAE5B,QAAI,CAAC,KAAK,UAAU;AAClB,MAAAD,sBAAqB;AAAA,QACnB,UAAU,KAAK;AAAA,QACf,OAAO,KAAK;AAAA,QACZ,YAAY,YAAY,IAAI,IAAI;AAAA,QAChC,SAAS;AAAA,QACT,aAAa,KAAK;AAAA,QAClB,cAAc,KAAK;AAAA,MACrB,CAAC;AACD,WAAK,QAAQ,MAAM,8BAA8B,EAAE,OAAO,KAAK,QAAQ,CAAC;AACxE,YAAM,IAAI,MAAM,4BAA4B;AAAA,IAC9C;AAEA,IAAAA,sBAAqB;AAAA,MACnB,UAAU,KAAK;AAAA,MACf,OAAO,KAAK;AAAA,MACZ,YAAY,YAAY,IAAI,IAAI;AAAA,MAChC,SAAS;AAAA,MACT,aAAa,KAAK;AAAA,MAClB,cAAc,KAAK;AAAA,IACrB,CAAC;AAED,UAAM,aAAa,cAAc,KAAK,WAAW;AAEjD,SAAK,QAAQ,KAAK,6BAA6B;AAAA,MAC7C,OAAO,KAAK;AAAA,MACZ,YAAY,KAAK,SAAS;AAAA,MAC1B;AAAA,IACF,CAAC;AAED,WAAO;AAAA,MACL,MAAM,KAAK;AAAA,MACX;AAAA,IACF;AAAA,EACF;AACF;AAEA,SAAS,cAAc,YAAwC;AAC7D,UAAQ,YAAY;AAAA,IAClB,KAAK;AAAQ,aAAO;AAAA,IACpB,KAAK;AAAU,aAAO;AAAA,IACtB;AAAS,aAAO,cAAc;AAAA,EAChC;AACF;;;ACnIO,SAAS,sBAAsB,QAA+B,QAAkC;AACrG,UAAQ,OAAO,MAAM;AAAA,IACnB,KAAK,aAAa;AAChB,UAAI,CAAC,OAAO,UAAU,OAAO,OAAO,KAAK,MAAM,IAAI;AACjD,cAAM,IAAI,MAAM,mDAAmD;AAAA,MACrE;AACA,aAAO,IAAI;AAAA,QACT,OAAO;AAAA,QACP,OAAO;AAAA,QACP,OAAO,YAAY,OAAO;AAAA,QAC1B;AAAA,MACF;AAAA,IACF;AAAA,IAEA,KAAK,UAAU;AACb,aAAO,IAAI;AAAA,QACT,OAAO;AAAA,QACP,OAAO,YAAY,OAAO;AAAA,QAC1B;AAAA,MACF;AAAA,IACF;AAAA,IAEA;AACE,YAAM,IAAI,MAAM,sCAAsC,OAAO,IAAI,EAAE;AAAA,EACvE;AACF;;;ACtCO,IAAM,sBAAN,MAAqD;AAAA,EACjD,OAAO;AAAA,EACP,UAAU;AAAA,EACX,YAAsB,CAAC;AAAA,EACvB,gBAAwB;AAAA,EACxB,cAAwB,CAAC;AAAA,EAC1B,QAAuG,CAAC;AAAA,EAE/G,YAAY,YAAsB,CAAC,eAAe,GAAG,aAAwB;AAC3E,SAAK,YAAY;AACjB,SAAK,cAAc,eAAe,UAAU,IAAI,MAAM,UAAU;AAAA,EAClE;AAAA,EAEA,MAAM,aAAa,QAAgB,WAAmB,aAAqB,SAA6C;AACtH,UAAM,WAAW,MAAM,KAAK,yBAAyB,QAAQ,WAAW,aAAa,OAAO;AAC5F,WAAO,SAAS;AAAA,EAClB;AAAA,EAEA,MAAM,yBAAyB,QAAgB,WAAmB,aAAqB,SAAwD;AAC7I,SAAK,MAAM,KAAK,EAAE,QAAQ,WAAW,aAAa,GAAI,UAAU,EAAE,QAAQ,IAAI,CAAC,EAAG,CAAC;AAEnF,UAAM,OAAO,KAAK,UAAU,KAAK,aAAa;AAC9C,UAAM,aAAa,KAAK,YAAY,KAAK,aAAa,KAAK;AAE3D,QAAI,KAAK,gBAAgB,KAAK,UAAU,SAAS,GAAG;AAClD,WAAK;AAAA,IACP;AAEA,WAAO,EAAE,MAAM,WAAW;AAAA,EAC5B;AAAA;AAAA,EAGA,QAAc;AACZ,SAAK,QAAQ,CAAC;AACd,SAAK,gBAAgB;AAAA,EACvB;AAAA,EAEA,aAAa,WAAqB,aAA8B;AAC9D,SAAK,YAAY;AACjB,SAAK,cAAc,eAAe,UAAU,IAAI,MAAM,UAAU;AAChE,SAAK,gBAAgB;AAAA,EACvB;AACF;","names":["recordInferenceUsage","body"]}
+{"version":3,"sources":["../src/implementations/anthropic.ts","../src/implementations/ollama.ts","../src/factory.ts","../src/implementations/mock.ts"],"sourcesContent":["// Anthropic Claude implementation of InferenceClient interface\n\nimport Anthropic from '@anthropic-ai/sdk';\nimport type { Logger } from '@semiont/core';\nimport { recordInferenceUsage } from '@semiont/observability';\nimport { InferenceClient, InferenceOptions, InferenceResponse } from '../interface.js';\n\n// Forced-tool channel for JSON mode. Anthropic has no grammar-constrained\n// sampling like Ollama's `format`; the equivalent hard guarantee is a *tool\n// call*. We offer exactly one tool and force it via `tool_choice`, so the model\n// must answer by filling the tool's input — which the API serializes as\n// properly-escaped JSON. That kills both free-text failure modes at the source:\n// trailing prose after the `]` (variant 1) and an unescaped `\"` inside a string\n// (variant 2), neither of which a prefill could prevent.\n//\n// A tool's input must be an *object*, so the array is carried under `items`\n// and unwrapped on return (see generateTextWithMetadata) — the caller still\n// receives a top-level JSON array in `text`, exactly as on Ollama.\nconst JSON_ARRAY_TOOL: Anthropic.Tool = {\n  name: 'emit_json_array',\n  description:\n    'Return your entire answer by calling this tool. Put the JSON array of results under the \"items\" property, and emit no prose.',\n  input_schema: {\n    type: 'object',\n    properties: {\n      // Element shape is unconstrained here — the prompt carries the per-element\n      // schema; the tool only enforces that the top-level result is an array.\n      items: { type: 'array', items: {} },\n    },\n    required: ['items'],\n  },\n};\n\nexport class AnthropicInferenceClient implements InferenceClient {\n  readonly type = 'anthropic' as const;\n  readonly modelId: string;\n  private client: Anthropic;\n  private logger?: Logger;\n\n  constructor(apiKey: string, model: string, baseURL?: string, logger?: Logger) {\n    this.client = new Anthropic({\n      apiKey,\n      baseURL: baseURL || 'https://api.anthropic.com',\n    });\n    this.modelId = model;\n    this.logger = logger;\n  }\n\n  async generateText(prompt: string, maxTokens: number, temperature: number, options?: InferenceOptions): Promise<string> {\n    const response = await this.generateTextWithMetadata(prompt, maxTokens, temperature, options);\n    return response.text;\n  }\n\n  async generateTextWithMetadata(prompt: string, maxTokens: number, temperature: number, options?: InferenceOptions): Promise<InferenceResponse> {\n    const jsonMode = options?.format === 'json';\n\n    this.logger?.debug('Generating text with inference client', {\n      model: this.modelId,\n      promptLength: prompt.length,\n      maxTokens,\n      temperature,\n      format: options?.format,\n    });\n\n    const start = performance.now();\n    let response: Awaited<ReturnType<typeof this.client.messages.create>>;\n    try {\n      response = await this.client.messages.create({\n        model: this.modelId,\n        max_tokens: maxTokens,\n        temperature,\n        messages: [{ role: 'user', content: prompt }],\n        // JSON mode → force the structured-output tool. No prefill assistant\n        // turn: the constraint now lives in the tool call, not in free text.\n        ...(jsonMode\n          ? { tools: [JSON_ARRAY_TOOL], tool_choice: { type: 'tool' as const, name: JSON_ARRAY_TOOL.name } }\n          : {}),\n      });\n    } catch (err) {\n      recordInferenceUsage({\n        provider: this.type,\n        model: this.modelId,\n        durationMs: performance.now() - start,\n        outcome: 'error',\n      });\n      throw err;\n    }\n\n    this.logger?.debug('Inference response received', {\n      model: this.modelId,\n      contentBlocks: response.content.length,\n      stopReason: response.stop_reason\n    });\n\n    let text: string;\n    if (jsonMode) {\n      // The answer arrives as a tool_use block, not text. Unwrap the `items`\n      // array and re-serialize it so `text` is a complete, parseable top-level\n      // JSON array — the cross-provider contract every consumer reads.\n      const toolUse = response.content.find(c => c.type === 'tool_use');\n      if (!toolUse || toolUse.type !== 'tool_use') {\n        recordInferenceUsage({\n          provider: this.type,\n          model: this.modelId,\n          durationMs: performance.now() - start,\n          outcome: 'error',\n          inputTokens: response.usage?.input_tokens,\n          outputTokens: response.usage?.output_tokens,\n        });\n        this.logger?.error('No tool_use content in inference response', {\n          model: this.modelId,\n          contentTypes: response.content.map(c => c.type)\n        });\n        throw new Error('No tool_use content in inference response');\n      }\n      // `input` is typed `unknown` by the SDK. A truncated (`max_tokens`)\n      // response may carry partial or absent `items` — fall back to the partial\n      // array, or `[]` if absent; the consumer flags truncation via stopReason.\n      const input = toolUse.input as { items?: unknown };\n      const items = Array.isArray(input.items) ? input.items : [];\n      text = JSON.stringify(items);\n    } else {\n      const textContent = response.content.find(c => c.type === 'text');\n      if (!textContent || textContent.type !== 'text') {\n        recordInferenceUsage({\n          provider: this.type,\n          model: this.modelId,\n          durationMs: performance.now() - start,\n          outcome: 'error',\n          inputTokens: response.usage?.input_tokens,\n          outputTokens: response.usage?.output_tokens,\n        });\n        this.logger?.error('No text content in inference response', {\n          model: this.modelId,\n          contentTypes: response.content.map(c => c.type)\n        });\n        throw new Error('No text content in inference response');\n      }\n      text = textContent.text;\n    }\n\n    recordInferenceUsage({\n      provider: this.type,\n      model: this.modelId,\n      durationMs: performance.now() - start,\n      outcome: 'success',\n      inputTokens: response.usage?.input_tokens,\n      outputTokens: response.usage?.output_tokens,\n    });\n\n    this.logger?.info('Text generation completed', {\n      model: this.modelId,\n      textLength: text.length,\n      stopReason: response.stop_reason\n    });\n\n    return {\n      text,\n      stopReason: response.stop_reason || 'unknown'\n    };\n  }\n}\n","// Ollama implementation of InferenceClient interface\n// Uses native Ollama HTTP API (no SDK dependency)\n\nimport type { Logger } from '@semiont/core';\nimport { recordInferenceUsage } from '@semiont/observability';\nimport { InferenceClient, InferenceOptions, InferenceResponse } from '../interface.js';\n\ninterface OllamaGenerateResponse {\n  response: string;\n  done: boolean;\n  done_reason?: string;\n  /** Number of prompt tokens evaluated. Available on most Ollama versions. */\n  prompt_eval_count?: number;\n  /** Number of tokens generated. */\n  eval_count?: number;\n}\n\nexport class OllamaInferenceClient implements InferenceClient {\n  readonly type = 'ollama' as const;\n  readonly modelId: string;\n  private baseURL: string;\n  private logger?: Logger;\n\n  constructor(model: string, baseURL?: string, logger?: Logger) {\n    this.baseURL = (baseURL || 'http://localhost:11434').replace(/\\/+$/, '');\n    this.modelId = model;\n    this.logger = logger;\n  }\n\n  async generateText(prompt: string, maxTokens: number, temperature: number, options?: InferenceOptions): Promise<string> {\n    const response = await this.generateTextWithMetadata(prompt, maxTokens, temperature, options);\n    return response.text;\n  }\n\n  async generateTextWithMetadata(prompt: string, maxTokens: number, temperature: number, options?: InferenceOptions): Promise<InferenceResponse> {\n    this.logger?.debug('Generating text with Ollama', {\n      model: this.modelId,\n      promptLength: prompt.length,\n      maxTokens,\n      temperature,\n      format: options?.format,\n    });\n\n    const url = `${this.baseURL}/api/generate`;\n    const start = performance.now();\n\n    // Ollama's `format` parameter accepts either the literal string\n    // `\"json\"` (any valid JSON, including objects, numbers, etc.) or a\n    // JSON schema (constrains the top-level shape). The contract on the\n    // inference side is \"parseable JSON array,\" so we pass a minimal\n    // array schema rather than the bare `\"json\"` string — without it,\n    // the model can satisfy \"valid JSON\" by emitting `{\"entities\": [...]}`\n    // and break every consumer that expects to call `.map` on the\n    // top-level value. The schema's `items: {}` keeps element shape\n    // unconstrained — the prompt still carries the per-element schema;\n    // we only enforce the outer array.\n    const body: Record<string, unknown> = {\n      model: this.modelId,\n      prompt,\n      stream: false,\n      think: false,\n      options: {\n        num_predict: maxTokens,\n        temperature,\n      },\n    };\n    if (options?.format === 'json') {\n      body['format'] = { type: 'array', items: {} };\n    }\n\n    let res: Response;\n    try {\n      res = await fetch(url, {\n        method: 'POST',\n        headers: { 'Content-Type': 'application/json' },\n        body: JSON.stringify(body),\n      });\n    } catch (err) {\n      recordInferenceUsage({\n        provider: this.type,\n        model: this.modelId,\n        durationMs: performance.now() - start,\n        outcome: 'error',\n      });\n      throw err;\n    }\n\n    if (!res.ok) {\n      recordInferenceUsage({\n        provider: this.type,\n        model: this.modelId,\n        durationMs: performance.now() - start,\n        outcome: 'error',\n      });\n      const body = await res.text();\n      this.logger?.error('Ollama API error', {\n        model: this.modelId,\n        status: res.status,\n        body,\n      });\n      throw new Error(`Ollama API error (${res.status}): ${body}`);\n    }\n\n    const data = await res.json() as OllamaGenerateResponse;\n\n    if (!data.response) {\n      recordInferenceUsage({\n        provider: this.type,\n        model: this.modelId,\n        durationMs: performance.now() - start,\n        outcome: 'error',\n        inputTokens: data.prompt_eval_count,\n        outputTokens: data.eval_count,\n      });\n      this.logger?.error('Empty response from Ollama', { model: this.modelId });\n      throw new Error('Empty response from Ollama');\n    }\n\n    recordInferenceUsage({\n      provider: this.type,\n      model: this.modelId,\n      durationMs: performance.now() - start,\n      outcome: 'success',\n      inputTokens: data.prompt_eval_count,\n      outputTokens: data.eval_count,\n    });\n\n    const stopReason = mapStopReason(data.done_reason);\n\n    this.logger?.info('Text generation completed', {\n      model: this.modelId,\n      textLength: data.response.length,\n      stopReason,\n    });\n\n    return {\n      text: data.response,\n      stopReason,\n    };\n  }\n}\n\nfunction mapStopReason(doneReason: string | undefined): string {\n  switch (doneReason) {\n    case 'stop': return 'end_turn';\n    case 'length': return 'max_tokens';\n    default: return doneReason || 'unknown';\n  }\n}\n","// Factory for creating inference client instances based on configuration\n\nimport type { Logger } from '@semiont/core';\nimport { InferenceClient } from './interface.js';\nimport { AnthropicInferenceClient } from './implementations/anthropic.js';\nimport { OllamaInferenceClient } from './implementations/ollama.js';\n\nexport type InferenceClientType = 'anthropic' | 'ollama';\n\nexport interface InferenceClientConfig {\n  type: InferenceClientType;\n  apiKey?: string;\n  model: string;\n  endpoint?: string;\n  baseURL?: string;\n}\n\nexport function createInferenceClient(config: InferenceClientConfig, logger?: Logger): InferenceClient {\n  switch (config.type) {\n    case 'anthropic': {\n      if (!config.apiKey || config.apiKey.trim() === '') {\n        throw new Error('apiKey is required for Anthropic inference client');\n      }\n      return new AnthropicInferenceClient(\n        config.apiKey,\n        config.model,\n        config.endpoint || config.baseURL,\n        logger\n      );\n    }\n\n    case 'ollama': {\n      return new OllamaInferenceClient(\n        config.model,\n        config.endpoint || config.baseURL,\n        logger\n      );\n    }\n\n    default:\n      throw new Error(`Unsupported inference client type: ${config.type}`);\n  }\n}\n","// Mock implementation of InferenceClient for testing\n\nimport { InferenceClient, InferenceOptions, InferenceResponse } from '../interface.js';\n\nexport class MockInferenceClient implements InferenceClient {\n  readonly type = 'mock' as const;\n  readonly modelId = 'mock-model' as const;\n  private responses: string[] = [];\n  private responseIndex: number = 0;\n  private stopReasons: string[] = [];\n  public calls: Array<{ prompt: string; maxTokens: number; temperature: number; options?: InferenceOptions }> = [];\n\n  constructor(responses: string[] = ['Mock response'], stopReasons?: string[]) {\n    this.responses = responses;\n    this.stopReasons = stopReasons || responses.map(() => 'end_turn');\n  }\n\n  async generateText(prompt: string, maxTokens: number, temperature: number, options?: InferenceOptions): Promise<string> {\n    const response = await this.generateTextWithMetadata(prompt, maxTokens, temperature, options);\n    return response.text;\n  }\n\n  async generateTextWithMetadata(prompt: string, maxTokens: number, temperature: number, options?: InferenceOptions): Promise<InferenceResponse> {\n    this.calls.push({ prompt, maxTokens, temperature, ...(options ? { options } : {}) });\n\n    const text = this.responses[this.responseIndex];\n    const stopReason = this.stopReasons[this.responseIndex] || 'end_turn';\n\n    if (this.responseIndex < this.responses.length - 1) {\n      this.responseIndex++;\n    }\n\n    return { text, stopReason };\n  }\n\n  // Test helper methods\n  reset(): void {\n    this.calls = [];\n    this.responseIndex = 0;\n  }\n\n  setResponses(responses: string[], stopReasons?: string[]): void {\n    this.responses = responses;\n    this.stopReasons = stopReasons || responses.map(() => 'end_turn');\n    this.responseIndex = 0;\n  }\n}\n"],"mappings":";AAEA,OAAO,eAAe;AAEtB,SAAS,4BAA4B;AAcrC,IAAM,kBAAkC;AAAA,EACtC,MAAM;AAAA,EACN,aACE;AAAA,EACF,cAAc;AAAA,IACZ,MAAM;AAAA,IACN,YAAY;AAAA;AAAA;AAAA,MAGV,OAAO,EAAE,MAAM,SAAS,OAAO,CAAC,EAAE;AAAA,IACpC;AAAA,IACA,UAAU,CAAC,OAAO;AAAA,EACpB;AACF;AAEO,IAAM,2BAAN,MAA0D;AAAA,EACtD,OAAO;AAAA,EACP;AAAA,EACD;AAAA,EACA;AAAA,EAER,YAAY,QAAgB,OAAe,SAAkB,QAAiB;AAC5E,SAAK,SAAS,IAAI,UAAU;AAAA,MAC1B;AAAA,MACA,SAAS,WAAW;AAAA,IACtB,CAAC;AACD,SAAK,UAAU;AACf,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,MAAM,aAAa,QAAgB,WAAmB,aAAqB,SAA6C;AACtH,UAAM,WAAW,MAAM,KAAK,yBAAyB,QAAQ,WAAW,aAAa,OAAO;AAC5F,WAAO,SAAS;AAAA,EAClB;AAAA,EAEA,MAAM,yBAAyB,QAAgB,WAAmB,aAAqB,SAAwD;AAC7I,UAAM,WAAW,SAAS,WAAW;AAErC,SAAK,QAAQ,MAAM,yCAAyC;AAAA,MAC1D,OAAO,KAAK;AAAA,MACZ,cAAc,OAAO;AAAA,MACrB;AAAA,MACA;AAAA,MACA,QAAQ,SAAS;AAAA,IACnB,CAAC;AAED,UAAM,QAAQ,YAAY,IAAI;AAC9B,QAAI;AACJ,QAAI;AACF,iBAAW,MAAM,KAAK,OAAO,SAAS,OAAO;AAAA,QAC3C,OAAO,KAAK;AAAA,QACZ,YAAY;AAAA,QACZ;AAAA,QACA,UAAU,CAAC,EAAE,MAAM,QAAQ,SAAS,OAAO,CAAC;AAAA;AAAA;AAAA,QAG5C,GAAI,WACA,EAAE,OAAO,CAAC,eAAe,GAAG,aAAa,EAAE,MAAM,QAAiB,MAAM,gBAAgB,KAAK,EAAE,IAC/F,CAAC;AAAA,MACP,CAAC;AAAA,IACH,SAAS,KAAK;AACZ,2BAAqB;AAAA,QACnB,UAAU,KAAK;AAAA,QACf,OAAO,KAAK;AAAA,QACZ,YAAY,YAAY,IAAI,IAAI;AAAA,QAChC,SAAS;AAAA,MACX,CAAC;AACD,YAAM;AAAA,IACR;AAEA,SAAK,QAAQ,MAAM,+BAA+B;AAAA,MAChD,OAAO,KAAK;AAAA,MACZ,eAAe,SAAS,QAAQ;AAAA,MAChC,YAAY,SAAS;AAAA,IACvB,CAAC;AAED,QAAI;AACJ,QAAI,UAAU;AAIZ,YAAM,UAAU,SAAS,QAAQ,KAAK,OAAK,EAAE,SAAS,UAAU;AAChE,UAAI,CAAC,WAAW,QAAQ,SAAS,YAAY;AAC3C,6BAAqB;AAAA,UACnB,UAAU,KAAK;AAAA,UACf,OAAO,KAAK;AAAA,UACZ,YAAY,YAAY,IAAI,IAAI;AAAA,UAChC,SAAS;AAAA,UACT,aAAa,SAAS,OAAO;AAAA,UAC7B,cAAc,SAAS,OAAO;AAAA,QAChC,CAAC;AACD,aAAK,QAAQ,MAAM,6CAA6C;AAAA,UAC9D,OAAO,KAAK;AAAA,UACZ,cAAc,SAAS,QAAQ,IAAI,OAAK,EAAE,IAAI;AAAA,QAChD,CAAC;AACD,cAAM,IAAI,MAAM,2CAA2C;AAAA,MAC7D;AAIA,YAAM,QAAQ,QAAQ;AACtB,YAAM,QAAQ,MAAM,QAAQ,MAAM,KAAK,IAAI,MAAM,QAAQ,CAAC;AAC1D,aAAO,KAAK,UAAU,KAAK;AAAA,IAC7B,OAAO;AACL,YAAM,cAAc,SAAS,QAAQ,KAAK,OAAK,EAAE,SAAS,MAAM;AAChE,UAAI,CAAC,eAAe,YAAY,SAAS,QAAQ;AAC/C,6BAAqB;AAAA,UACnB,UAAU,KAAK;AAAA,UACf,OAAO,KAAK;AAAA,UACZ,YAAY,YAAY,IAAI,IAAI;AAAA,UAChC,SAAS;AAAA,UACT,aAAa,SAAS,OAAO;AAAA,UAC7B,cAAc,SAAS,OAAO;AAAA,QAChC,CAAC;AACD,aAAK,QAAQ,MAAM,yCAAyC;AAAA,UAC1D,OAAO,KAAK;AAAA,UACZ,cAAc,SAAS,QAAQ,IAAI,OAAK,EAAE,IAAI;AAAA,QAChD,CAAC;AACD,cAAM,IAAI,MAAM,uCAAuC;AAAA,MACzD;AACA,aAAO,YAAY;AAAA,IACrB;AAEA,yBAAqB;AAAA,MACnB,UAAU,KAAK;AAAA,MACf,OAAO,KAAK;AAAA,MACZ,YAAY,YAAY,IAAI,IAAI;AAAA,MAChC,SAAS;AAAA,MACT,aAAa,SAAS,OAAO;AAAA,MAC7B,cAAc,SAAS,OAAO;AAAA,IAChC,CAAC;AAED,SAAK,QAAQ,KAAK,6BAA6B;AAAA,MAC7C,OAAO,KAAK;AAAA,MACZ,YAAY,KAAK;AAAA,MACjB,YAAY,SAAS;AAAA,IACvB,CAAC;AAED,WAAO;AAAA,MACL;AAAA,MACA,YAAY,SAAS,eAAe;AAAA,IACtC;AAAA,EACF;AACF;;;AC7JA,SAAS,wBAAAA,6BAA4B;AAa9B,IAAM,wBAAN,MAAuD;AAAA,EACnD,OAAO;AAAA,EACP;AAAA,EACD;AAAA,EACA;AAAA,EAER,YAAY,OAAe,SAAkB,QAAiB;AAC5D,SAAK,WAAW,WAAW,0BAA0B,QAAQ,QAAQ,EAAE;AACvE,SAAK,UAAU;AACf,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,MAAM,aAAa,QAAgB,WAAmB,aAAqB,SAA6C;AACtH,UAAM,WAAW,MAAM,KAAK,yBAAyB,QAAQ,WAAW,aAAa,OAAO;AAC5F,WAAO,SAAS;AAAA,EAClB;AAAA,EAEA,MAAM,yBAAyB,QAAgB,WAAmB,aAAqB,SAAwD;AAC7I,SAAK,QAAQ,MAAM,+BAA+B;AAAA,MAChD,OAAO,KAAK;AAAA,MACZ,cAAc,OAAO;AAAA,MACrB;AAAA,MACA;AAAA,MACA,QAAQ,SAAS;AAAA,IACnB,CAAC;AAED,UAAM,MAAM,GAAG,KAAK,OAAO;AAC3B,UAAM,QAAQ,YAAY,IAAI;AAY9B,UAAM,OAAgC;AAAA,MACpC,OAAO,KAAK;AAAA,MACZ;AAAA,MACA,QAAQ;AAAA,MACR,OAAO;AAAA,MACP,SAAS;AAAA,QACP,aAAa;AAAA,QACb;AAAA,MACF;AAAA,IACF;AACA,QAAI,SAAS,WAAW,QAAQ;AAC9B,WAAK,QAAQ,IAAI,EAAE,MAAM,SAAS,OAAO,CAAC,EAAE;AAAA,IAC9C;AAEA,QAAI;AACJ,QAAI;AACF,YAAM,MAAM,MAAM,KAAK;AAAA,QACrB,QAAQ;AAAA,QACR,SAAS,EAAE,gBAAgB,mBAAmB;AAAA,QAC9C,MAAM,KAAK,UAAU,IAAI;AAAA,MAC3B,CAAC;AAAA,IACH,SAAS,KAAK;AACZ,MAAAA,sBAAqB;AAAA,QACnB,UAAU,KAAK;AAAA,QACf,OAAO,KAAK;AAAA,QACZ,YAAY,YAAY,IAAI,IAAI;AAAA,QAChC,SAAS;AAAA,MACX,CAAC;AACD,YAAM;AAAA,IACR;AAEA,QAAI,CAAC,IAAI,IAAI;AACX,MAAAA,sBAAqB;AAAA,QACnB,UAAU,KAAK;AAAA,QACf,OAAO,KAAK;AAAA,QACZ,YAAY,YAAY,IAAI,IAAI;AAAA,QAChC,SAAS;AAAA,MACX,CAAC;AACD,YAAMC,QAAO,MAAM,IAAI,KAAK;AAC5B,WAAK,QAAQ,MAAM,oBAAoB;AAAA,QACrC,OAAO,KAAK;AAAA,QACZ,QAAQ,IAAI;AAAA,QACZ,MAAAA;AAAA,MACF,CAAC;AACD,YAAM,IAAI,MAAM,qBAAqB,IAAI,MAAM,MAAMA,KAAI,EAAE;AAAA,IAC7D;AAEA,UAAM,OAAO,MAAM,IAAI,KAAK;AAE5B,QAAI,CAAC,KAAK,UAAU;AAClB,MAAAD,sBAAqB;AAAA,QACnB,UAAU,KAAK;AAAA,QACf,OAAO,KAAK;AAAA,QACZ,YAAY,YAAY,IAAI,IAAI;AAAA,QAChC,SAAS;AAAA,QACT,aAAa,KAAK;AAAA,QAClB,cAAc,KAAK;AAAA,MACrB,CAAC;AACD,WAAK,QAAQ,MAAM,8BAA8B,EAAE,OAAO,KAAK,QAAQ,CAAC;AACxE,YAAM,IAAI,MAAM,4BAA4B;AAAA,IAC9C;AAEA,IAAAA,sBAAqB;AAAA,MACnB,UAAU,KAAK;AAAA,MACf,OAAO,KAAK;AAAA,MACZ,YAAY,YAAY,IAAI,IAAI;AAAA,MAChC,SAAS;AAAA,MACT,aAAa,KAAK;AAAA,MAClB,cAAc,KAAK;AAAA,IACrB,CAAC;AAED,UAAM,aAAa,cAAc,KAAK,WAAW;AAEjD,SAAK,QAAQ,KAAK,6BAA6B;AAAA,MAC7C,OAAO,KAAK;AAAA,MACZ,YAAY,KAAK,SAAS;AAAA,MAC1B;AAAA,IACF,CAAC;AAED,WAAO;AAAA,MACL,MAAM,KAAK;AAAA,MACX;AAAA,IACF;AAAA,EACF;AACF;AAEA,SAAS,cAAc,YAAwC;AAC7D,UAAQ,YAAY;AAAA,IAClB,KAAK;AAAQ,aAAO;AAAA,IACpB,KAAK;AAAU,aAAO;AAAA,IACtB;AAAS,aAAO,cAAc;AAAA,EAChC;AACF;;;ACnIO,SAAS,sBAAsB,QAA+B,QAAkC;AACrG,UAAQ,OAAO,MAAM;AAAA,IACnB,KAAK,aAAa;AAChB,UAAI,CAAC,OAAO,UAAU,OAAO,OAAO,KAAK,MAAM,IAAI;AACjD,cAAM,IAAI,MAAM,mDAAmD;AAAA,MACrE;AACA,aAAO,IAAI;AAAA,QACT,OAAO;AAAA,QACP,OAAO;AAAA,QACP,OAAO,YAAY,OAAO;AAAA,QAC1B;AAAA,MACF;AAAA,IACF;AAAA,IAEA,KAAK,UAAU;AACb,aAAO,IAAI;AAAA,QACT,OAAO;AAAA,QACP,OAAO,YAAY,OAAO;AAAA,QAC1B;AAAA,MACF;AAAA,IACF;AAAA,IAEA;AACE,YAAM,IAAI,MAAM,sCAAsC,OAAO,IAAI,EAAE;AAAA,EACvE;AACF;;;ACtCO,IAAM,sBAAN,MAAqD;AAAA,EACjD,OAAO;AAAA,EACP,UAAU;AAAA,EACX,YAAsB,CAAC;AAAA,EACvB,gBAAwB;AAAA,EACxB,cAAwB,CAAC;AAAA,EAC1B,QAAuG,CAAC;AAAA,EAE/G,YAAY,YAAsB,CAAC,eAAe,GAAG,aAAwB;AAC3E,SAAK,YAAY;AACjB,SAAK,cAAc,eAAe,UAAU,IAAI,MAAM,UAAU;AAAA,EAClE;AAAA,EAEA,MAAM,aAAa,QAAgB,WAAmB,aAAqB,SAA6C;AACtH,UAAM,WAAW,MAAM,KAAK,yBAAyB,QAAQ,WAAW,aAAa,OAAO;AAC5F,WAAO,SAAS;AAAA,EAClB;AAAA,EAEA,MAAM,yBAAyB,QAAgB,WAAmB,aAAqB,SAAwD;AAC7I,SAAK,MAAM,KAAK,EAAE,QAAQ,WAAW,aAAa,GAAI,UAAU,EAAE,QAAQ,IAAI,CAAC,EAAG,CAAC;AAEnF,UAAM,OAAO,KAAK,UAAU,KAAK,aAAa;AAC9C,UAAM,aAAa,KAAK,YAAY,KAAK,aAAa,KAAK;AAE3D,QAAI,KAAK,gBAAgB,KAAK,UAAU,SAAS,GAAG;AAClD,WAAK;AAAA,IACP;AAEA,WAAO,EAAE,MAAM,WAAW;AAAA,EAC5B;AAAA;AAAA,EAGA,QAAc;AACZ,SAAK,QAAQ,CAAC;AACd,SAAK,gBAAgB;AAAA,EACvB;AAAA,EAEA,aAAa,WAAqB,aAA8B;AAC9D,SAAK,YAAY;AACjB,SAAK,cAAc,eAAe,UAAU,IAAI,MAAM,UAAU;AAChE,SAAK,gBAAgB;AAAA,EACvB;AACF;","names":["recordInferenceUsage","body"]}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@semiont/inference",
-  "version": "0.5.8",
+  "version": "0.5.10",
   "engines": {
     "node": ">=24.0.0"
   },
@@ -27,9 +27,9 @@
     "test:coverage": "vitest run --coverage"
   },
   "dependencies": {
-    "@anthropic-ai/sdk": "^0.104.1",
-    "@semiont/core": "0.5.8",
-    "@semiont/observability": "0.5.8"
+    "@anthropic-ai/sdk": "^0.105.0",
+    "@semiont/core": "0.5.10",
+    "@semiont/observability": "0.5.10"
   },
   "devDependencies": {
     "@vitest/coverage-v8": "^4.1.8",