@nuvin/nuvin-core 1.4.4 → 1.5.2

package/dist/VERSION

@@ -1,4 +1,4 @@
 {
-  "version": "1.4.4",
-  "commit": "a070d52"
+  "version": "1.5.2",
+  "commit": "f0799f5"
 }

package/dist/index.d.ts

@@ -283,6 +283,8 @@ type UsageData = {
     cost_details?: {
         upstream_inference_cost?: number;
     };
+    cache_creation_input_tokens?: number;
+    cache_read_input_tokens?: number;
 };
 type CompletionResult = {
     content: string;
@@ -1218,19 +1220,16 @@ interface TransportResponse {
 }
 interface HttpTransport {
     get(url: string, headers?: HttpHeaders, signal?: AbortSignal): Promise<TransportResponse>;
-    postJson(url: string, body: unknown, headers?: HttpHeaders, signal?: AbortSignal): Promise<TransportResponse>;
-    postStream(url: string, body: unknown, headers?: HttpHeaders, signal?: AbortSignal): Promise<Response>;
+    post(url: string, body: unknown, headers?: HttpHeaders, signal?: AbortSignal): Promise<Response>;
 }
 declare class FetchTransport implements HttpTransport {
     private opts;
     private logger;
     constructor(opts?: PersistOptions);
-    private getRequestSize;
     private getResponseSize;
     private headersToRecord;
+    post(url: string, body: unknown, headers?: HttpHeaders, signal?: AbortSignal): Promise<Response>;
     get(url: string, headers?: HttpHeaders, signal?: AbortSignal): Promise<TransportResponse>;
-    postJson(url: string, body: unknown, headers?: HttpHeaders, signal?: AbortSignal): Promise<TransportResponse>;
-    postStream(url: string, body: unknown, headers?: HttpHeaders, signal?: AbortSignal): Promise<Response>;
 }
 
 declare class GithubAuthTransport implements HttpTransport {
@@ -1250,9 +1249,8 @@ declare class GithubAuthTransport implements HttpTransport {
     private makeAuthHeaders;
     private determineInitiator;
     private generateRequestId;
+    post(url: string, body: unknown, headers?: HttpHeaders, signal?: AbortSignal): Promise<Response>;
     get(url: string, headers?: HttpHeaders, signal?: AbortSignal): Promise<TransportResponse>;
-    postJson(url: string, body: unknown, headers?: HttpHeaders, signal?: AbortSignal): Promise<TransportResponse>;
-    postStream(url: string, body: unknown, headers?: HttpHeaders, signal?: AbortSignal): Promise<Response>;
 }
 
 declare class LLMError extends Error {
@@ -1268,6 +1266,7 @@ declare abstract class BaseLLM implements LLMPort {
         enablePromptCaching?: boolean;
     });
     protected abstract createTransport(): HttpTransport;
+    protected transformUsage(rawUsage: unknown): UsageData | undefined;
     protected getTransport(): HttpTransport;
     protected applyCacheControl(params: CompletionParams): CompletionParams;
     generateCompletion(params: CompletionParams, signal?: AbortSignal): Promise<CompletionResult>;
@@ -1303,6 +1302,7 @@ declare class GithubLLM extends BaseLLM implements LLMPort {
     private readonly opts;
     constructor(opts?: GithubOptions);
     protected createTransport(): GithubAuthTransport;
+    protected transformUsage(rawUsage: unknown): UsageData | undefined;
     getModels(signal?: AbortSignal): Promise<ModelInfo[]>;
     private handleError;
     generateCompletion(params: CompletionParams, signal?: AbortSignal): Promise<CompletionResult>;
@@ -1344,6 +1344,7 @@ declare class AnthropicAISDKLLM {
     private transformMessages;
     private transformTools;
     private transformToolChoice;
+    private transformUsage;
     private handleError;
     generateCompletion(params: CompletionParams, signal?: AbortSignal): Promise<CompletionResult>;
     streamCompletion(params: CompletionParams, handlers?: {
@@ -1351,6 +1352,7 @@ declare class AnthropicAISDKLLM {
         onToolCallDelta?: (tc: ToolCall) => void;
         onStreamFinish?: (finishReason?: string, usage?: UsageData) => void;
     }, signal?: AbortSignal): Promise<CompletionResult>;
+    getModels(signal?: AbortSignal): Promise<ModelInfo[]>;
 }
 
 type ModelConfig = false | true | string | string[] | Array<{

package/dist/index.js

@@ -3048,7 +3048,7 @@ var AssignTool = class {
     properties: {
       description: {
         type: "string",
-        description: 'Explanation of why this task is being delegated (e.g., "Delegate code review to specialist agent")'
+        description: "A summary of the task to be performed by the delegated agent. Be specific about the desired outcome. From 5-10 words."
       },
       agent: {
         type: "string",
@@ -3056,10 +3056,10 @@ var AssignTool = class {
       },
       task: {
         type: "string",
-        description: "Task description (3-4 sentences explaining what to do)"
+        description: "Detailed description of the task to be performed by the agent."
       }
     },
-    required: ["agent", "task"]
+    required: ["agent", "task", "description"]
   };
   /**
    * Update the enabled agents configuration
@@ -3196,7 +3196,7 @@ var LLMResolver = class {
 };
 
 // src/agent-manager.ts
-var DEFAULT_TIMEOUT_MS = 3e5;
+var DEFAULT_TIMEOUT_MS = 3e6;
 var MAX_DELEGATION_DEPTH = 3;
 var AgentManager = class {
   constructor(delegatingConfig, delegatingTools, llmFactory, eventCallback, configResolver) {
@@ -3394,25 +3394,34 @@ var AgentManager = class {
    * Execute agent task with timeout
    */
   async executeWithTimeout(orchestrator, taskDescription, timeoutMs, signal) {
-    return new Promise((resolve6, reject) => {
-      const timer = setTimeout(() => {
-        reject(new Error(`Task execution timeout after ${timeoutMs}ms`));
-      }, timeoutMs);
-      const abortHandler = () => {
-        clearTimeout(timer);
-        reject(new Error("Sub-agent execution aborted by user"));
-      };
-      signal?.addEventListener("abort", abortHandler);
-      orchestrator.send(taskDescription, { conversationId: "default", signal }).then((response) => {
-        clearTimeout(timer);
-        signal?.removeEventListener("abort", abortHandler);
-        resolve6(response);
-      }).catch((error) => {
-        clearTimeout(timer);
-        signal?.removeEventListener("abort", abortHandler);
-        reject(error);
-      });
-    });
+    const timeoutController = new AbortController();
+    const combinedSignal = signal ? AbortSignal.any([signal, timeoutController.signal]) : timeoutController.signal;
+    let timer;
+    try {
+      return await Promise.race([
+        orchestrator.send(taskDescription, {
+          conversationId: "default",
+          signal: combinedSignal
+        }),
+        new Promise((_, reject) => {
+          if (signal?.aborted) {
+            reject(new Error("Sub-agent execution aborted by user"));
+            return;
+          }
+          signal?.addEventListener("abort", () => reject(new Error("Sub-agent execution aborted by user")), {
+            once: true
+          });
+        }),
+        new Promise((_, reject) => {
+          timer = setTimeout(() => {
+            timeoutController.abort();
+            reject(new Error(`Task execution timeout after ${timeoutMs}ms`));
+          }, timeoutMs);
+        })
+      ]);
+    } finally {
+      if (timer) clearTimeout(timer);
+    }
   }
   /**
    * Get active specialist agent count
@@ -3958,24 +3967,6 @@ function mergeChoices(choices) {
     ...extraFields
   };
 }
-function normalizeUsage(usage) {
-  if (!usage) return void 0;
-  const usageObj = usage;
-  const prompt_tokens = usage.prompt_tokens ?? (typeof usageObj.input_tokens === "number" ? usageObj.input_tokens : void 0);
-  const completion_tokens = usage.completion_tokens ?? (typeof usageObj.output_tokens === "number" ? usageObj.output_tokens : void 0);
-  const total_tokens = usage.total_tokens ?? (prompt_tokens != null && completion_tokens != null ? prompt_tokens + completion_tokens : void 0);
-  const cost = usage.cost ?? (typeof usageObj.estimated_cost === "number" ? usageObj.estimated_cost : void 0);
-  return {
-    prompt_tokens,
-    completion_tokens,
-    total_tokens,
-    ...usage.reasoning_tokens !== void 0 && { reasoning_tokens: usage.reasoning_tokens },
-    ...usage.prompt_tokens_details && { prompt_tokens_details: usage.prompt_tokens_details },
-    ...usage.completion_tokens_details && { completion_tokens_details: usage.completion_tokens_details },
-    ...cost !== void 0 && { cost },
-    ...usage.cost_details && { cost_details: usage.cost_details }
-  };
-}
 
 // src/llm-providers/base-llm.ts
 var LLMError = class _LLMError extends Error {
@@ -3997,6 +3988,28 @@ var BaseLLM = class {
     this.apiUrl = apiUrl;
     this.enablePromptCaching = options?.enablePromptCaching ?? false;
   }
+  transformUsage(rawUsage) {
+    if (!rawUsage) return void 0;
+    const usage = rawUsage;
+    const promptTokens = usage.prompt_tokens ?? usage.input_tokens;
+    const completionTokens = usage.completion_tokens ?? usage.output_tokens;
+    const totalTokens = usage.total_tokens ?? (promptTokens != null && completionTokens != null ? promptTokens + completionTokens : void 0);
+    const cost = usage.cost ?? usage.estimated_cost;
+    return {
+      prompt_tokens: promptTokens,
+      completion_tokens: completionTokens,
+      total_tokens: totalTokens,
+      ...usage.reasoning_tokens !== void 0 && { reasoning_tokens: usage.reasoning_tokens },
+      ...usage.prompt_tokens_details && { prompt_tokens_details: usage.prompt_tokens_details },
+      ...usage.completion_tokens_details && { completion_tokens_details: usage.completion_tokens_details },
+      ...cost !== void 0 && { cost },
+      ...usage.cost_details && { cost_details: usage.cost_details },
+      ...usage.cache_creation_input_tokens !== void 0 && {
+        cache_creation_input_tokens: usage.cache_creation_input_tokens
+      },
+      ...usage.cache_read_input_tokens !== void 0 && { cache_read_input_tokens: usage.cache_read_input_tokens }
+    };
+  }
   getTransport() {
     if (!this.transport) this.transport = this.createTransport();
     return this.transport;
@@ -4073,7 +4086,7 @@ var BaseLLM = class {
     if (enhancedParams.tools && enhancedParams.tools.length > 0) body.tools = enhancedParams.tools;
     if (enhancedParams.tool_choice && enhancedParams.tools && enhancedParams.tools.length > 0)
       body.tool_choice = enhancedParams.tool_choice;
-    const res = await this.getTransport().postJson("/chat/completions", body, void 0, signal);
+    const res = await this.getTransport().post("/chat/completions", body, void 0, signal);
     if (!res.ok) {
       const text = await res.text();
       const isRetryable = res.status === 429 || res.status >= 500;
@@ -4081,7 +4094,7 @@ var BaseLLM = class {
     }
     const data = await res.json();
     const merged = mergeChoices(data.choices);
-    const usage = normalizeUsage(data.usage);
+    const usage = this.transformUsage(data.usage);
     return { ...merged, ...usage ? { usage } : {} };
   }
   async streamCompletion(params, handlers = {}, signal) {
@@ -4099,12 +4112,7 @@ var BaseLLM = class {
     if (enhancedParams.tools && enhancedParams.tools.length > 0) body.tools = enhancedParams.tools;
     if (enhancedParams.tool_choice && enhancedParams.tools && enhancedParams.tools.length > 0)
       body.tool_choice = enhancedParams.tool_choice;
-    const res = await this.getTransport().postStream(
-      "/chat/completions",
-      body,
-      { Accept: "text/event-stream" },
-      signal
-    );
+    const res = await this.getTransport().post("/chat/completions", body, { Accept: "text/event-stream" }, signal);
     if (!res.ok) {
       const text = await res.text();
       const isRetryable = res.status === 429 || res.status >= 500;
@@ -4137,7 +4145,7 @@ var BaseLLM = class {
         }
         const usageData = evt.usage || choices[0]?.usage;
         if (usageData) {
-          usage = normalizeUsage(usageData);
+          usage = this.transformUsage(usageData);
           if (lastFinishReason && handlers.onStreamFinish) {
             handlers.onStreamFinish(lastFinishReason, usage);
           } else {
@@ -4232,9 +4240,6 @@ var BaseLLM = class {
   }
 };
 
-// src/transports/transport.ts
-import * as crypto2 from "crypto";
-
 // src/logger/network.ts
 var NetworkLogger = class {
   constructor(opts = {}) {
@@ -4381,10 +4386,6 @@ var FetchTransport = class {
     this.logger = new NetworkLogger(this.opts);
   }
   logger;
-  getRequestSize(body) {
-    if (!body) return 0;
-    return new TextEncoder().encode(JSON.stringify(body)).length;
-  }
   getResponseSize(headers) {
     const contentLength = headers.get("content-length");
     return contentLength ? parseInt(contentLength, 10) : 0;
@@ -4396,6 +4397,17 @@ var FetchTransport = class {
     });
     return record;
   }
+  async post(url, body, headers, signal) {
+    if (typeof fetch !== "function") {
+      throw new Error("Global fetch is not available. Use Node 18+ or a compatible runtime.");
+    }
+    return fetch(url, {
+      method: "POST",
+      headers: { "Content-Type": "application/json", ...headers || {} },
+      body: JSON.stringify(body ?? {}),
+      signal
+    });
+  }
   async get(url, headers, signal) {
     if (typeof fetch !== "function") {
       throw new Error("Global fetch is not available. Use Node 18+ or a compatible runtime.");
@@ -4453,123 +4465,10 @@ var FetchTransport = class {
       throw error;
     }
   }
-  async postJson(url, body, headers, signal) {
-    if (typeof fetch !== "function") {
-      throw new Error("Global fetch is not available. Use Node 18+ or a compatible runtime.");
-    }
-    const startTime = performance.now();
-    const reqBody = JSON.stringify(body ?? {});
-    try {
-      const res = await fetch(url, {
-        method: "POST",
-        headers: { "Content-Type": "application/json", ...headers || {} },
-        body: reqBody,
-        signal
-      });
-      const ttfb = performance.now() - startTime;
-      try {
-        const clone = res.clone();
-        const text = await clone.text();
-        let response;
-        try {
-          response = JSON.parse(text);
-        } catch {
-          response = text;
-        }
-        const duration = performance.now() - startTime;
-        await this.logger.logRequest({
-          url,
-          method: "POST",
-          headers: { ...headers || {} },
-          requestBody: body,
-          responseStatus: res.status,
-          responseHeaders: this.headersToRecord(res.headers),
-          ok: res.ok,
-          stream: false,
-          startTime,
-          duration,
-          ttfb,
-          requestSize: this.getRequestSize(body),
-          responseSize: this.getResponseSize(res.headers),
-          response
-        });
-      } catch {
-      }
-      return res;
-    } catch (error) {
-      const duration = performance.now() - startTime;
-      await this.logger.logRequest({
-        url,
-        method: "POST",
-        headers: { ...headers || {} },
-        requestBody: body,
-        responseStatus: 0,
-        ok: false,
-        stream: false,
-        startTime,
-        duration,
-        requestSize: this.getRequestSize(body),
-        error: error instanceof Error ? error.message : String(error)
-      });
-      throw error;
-    }
-  }
-  async postStream(url, body, headers, signal) {
-    if (typeof fetch !== "function") {
-      throw new Error("Global fetch is not available. Use Node 18+ or a compatible runtime.");
-    }
-    const startTime = performance.now();
-    const reqBody = JSON.stringify(body ?? {});
-    try {
-      const res = await fetch(url, {
-        method: "POST",
-        headers: { "Content-Type": "application/json", ...headers || {} },
-        body: reqBody,
-        signal
-      });
-      const ttfb = performance.now() - startTime;
-      const duration = performance.now() - startTime;
-      const logEntryData = {
-        id: crypto2.randomUUID ? crypto2.randomUUID() : Math.random().toString(36).slice(2),
-        timestamp: Date.now().toString(),
-        url,
-        method: "POST",
-        headers: { ...headers || {} },
-        requestBody: body,
-        responseStatus: res.status,
-        responseHeaders: this.headersToRecord(res.headers),
-        ok: res.ok,
-        stream: true,
-        startTime,
-        duration,
-        ttfb,
-        requestSize: this.getRequestSize(body),
-        responseSize: this.getResponseSize(res.headers)
-      };
-      void this.logger.logRequest(logEntryData);
-      return this.logger.createStreamTapper(res, logEntryData);
-    } catch (error) {
-      const duration = performance.now() - startTime;
-      await this.logger.logRequest({
-        url,
-        method: "POST",
-        headers: { ...headers || {} },
-        requestBody: body,
-        responseStatus: 0,
-        ok: false,
-        stream: true,
-        startTime,
-        duration,
-        requestSize: this.getRequestSize(body),
-        error: error instanceof Error ? error.message : String(error)
-      });
-      throw error;
-    }
-  }
 };
 
 // src/transports/github-transport.ts
-import * as crypto3 from "crypto";
+import * as crypto2 from "crypto";
 var isVisionMessage = (msg) => {
   if (!msg) return false;
   const content = msg.content;
@@ -4653,8 +4552,8 @@ var GithubAuthTransport = class {
     return "agent";
   }
   generateRequestId() {
-    if (crypto3?.randomUUID) {
-      return crypto3.randomUUID();
+    if (crypto2?.randomUUID) {
+      return crypto2.randomUUID();
     }
     return "xxxxxxxx-xxxx-4xxx-yxxx-xxxxxxxxxxxx".replace(/[xy]/g, (c) => {
       const r = Math.random() * 16 | 0;
@@ -4662,44 +4561,29 @@ var GithubAuthTransport = class {
       return v.toString(16);
     });
   }
-  async get(url, headers, signal) {
+  async post(url, body, headers, signal) {
     if (!this.apiKey && this.accessToken) {
       await this.exchangeToken(signal);
     }
     const fullUrl = this.buildFullUrl(url);
-    let res = await this.inner.get(fullUrl, this.makeAuthHeaders(headers), signal);
+    let res = await this.inner.post(fullUrl, body, this.makeAuthHeaders(headers, body), signal);
     if (res.status === 401 && this.accessToken) {
       await this.exchangeToken(signal);
       const retryUrl = this.buildFullUrl(url);
-      res = await this.inner.get(retryUrl, this.makeAuthHeaders(headers), signal);
+      res = await this.inner.post(retryUrl, body, this.makeAuthHeaders(headers, body), signal);
     }
     return res;
   }
-  async postJson(url, body, headers, signal) {
+  async get(url, headers, signal) {
     if (!this.apiKey && this.accessToken) {
       await this.exchangeToken(signal);
     }
     const fullUrl = this.buildFullUrl(url);
-    let res = await this.inner.postJson(fullUrl, body, this.makeAuthHeaders(headers, body), signal);
-    if (res.status === 401 && this.accessToken) {
-      await this.exchangeToken(signal);
-      const retryUrl = this.buildFullUrl(url);
-      res = await this.inner.postJson(retryUrl, body, this.makeAuthHeaders(headers, body), signal);
-    }
-    return res;
-  }
-  async postStream(url, body, headers, signal) {
-    let hdrs = this.makeAuthHeaders({ Accept: "text/event-stream", ...headers || {} }, body);
-    if ((!this.apiKey || !hdrs.Authorization) && this.accessToken) {
-      await this.exchangeToken(signal);
-      hdrs = this.makeAuthHeaders({ Accept: "text/event-stream", ...headers || {} }, body);
-    }
-    const fullUrl = this.buildFullUrl(url);
-    let res = await this.inner.postStream(fullUrl, body, hdrs, signal);
+    let res = await this.inner.get(fullUrl, this.makeAuthHeaders(headers), signal);
     if (res.status === 401 && this.accessToken) {
       await this.exchangeToken(signal);
       const retryUrl = this.buildFullUrl(url);
-      res = await this.inner.postStream(retryUrl, body, this.makeAuthHeaders(hdrs, body), signal);
+      res = await this.inner.get(retryUrl, this.makeAuthHeaders(headers), signal);
     }
     return res;
   }
@@ -4742,13 +4626,9 @@ var BaseBearerAuthTransport = class {
     const fullUrl = this.buildFullUrl(url);
     return this.inner.get(fullUrl, this.makeAuthHeaders(headers), signal);
   }
-  async postJson(url, body, headers, signal) {
+  async post(url, body, headers, signal) {
     const fullUrl = this.buildFullUrl(url);
-    return this.inner.postJson(fullUrl, body, this.makeAuthHeaders(headers), signal);
-  }
-  async postStream(url, body, headers, signal) {
-    const fullUrl = this.buildFullUrl(url);
-    return this.inner.postStream(fullUrl, body, this.makeAuthHeaders(headers), signal);
+    return this.inner.post(fullUrl, body, this.makeAuthHeaders(headers), signal);
   }
 };
 
@@ -4782,6 +4662,9 @@ function normalizeModelLimits(provider, model) {
         maxOutput: limits?.max_output_tokens
       };
     }
+    case "anthropic": {
+      return getFallbackLimits("anthropic", model.id);
+    }
     case "deepinfra": {
       const metadata = model.metadata;
       const contextLength = metadata?.context_length;
@@ -4846,10 +4729,18 @@ var FALLBACK_LIMITS = {
     "kimi-k2-turbo-preview": { contextWindow: 262144 }
   },
   anthropic: {
-    "claude-sonnet-4-5": { contextWindow: 2e5, maxOutput: 16e3 },
-    "claude-3-5-sonnet-20241022": { contextWindow: 2e5, maxOutput: 8192 },
+    "claude-opus-4-5-20251101": { contextWindow: 2e5, maxOutput: 16e3 },
+    "claude-haiku-4-5-20251001": { contextWindow: 2e5, maxOutput: 16e3 },
+    "claude-sonnet-4-5-20250929": { contextWindow: 2e5, maxOutput: 16e3 },
+    "claude-opus-4-1-20250805": { contextWindow: 2e5, maxOutput: 16e3 },
+    "claude-opus-4-20250514": { contextWindow: 2e5, maxOutput: 16e3 },
+    "claude-sonnet-4-20250514": { contextWindow: 2e5, maxOutput: 16e3 },
+    "claude-3-7-sonnet-20250219": { contextWindow: 2e5, maxOutput: 8192 },
+    "claude-3-5-haiku-20241022": { contextWindow: 2e5, maxOutput: 8192 },
+    "claude-3-haiku-20240307": { contextWindow: 2e5, maxOutput: 4096 },
     "claude-3-opus-20240229": { contextWindow: 2e5, maxOutput: 4096 },
-    "claude-3-haiku-20240307": { contextWindow: 2e5, maxOutput: 4096 }
+    "claude-sonnet-4-5": { contextWindow: 2e5, maxOutput: 16e3 },
+    "claude-3-5-sonnet-20241022": { contextWindow: 2e5, maxOutput: 8192 }
   }
 };
 function getFallbackLimits(provider, model) {
@@ -4859,7 +4750,10 @@ function getFallbackLimits(provider, model) {
 }
 function normalizeModelInfo(provider, model) {
   const id = model.id;
-  const name = model.name ?? id;
+  let name = model.name ?? id;
+  if (provider.toLowerCase() === "anthropic" && model.display_name) {
+    name = model.display_name;
+  }
   const limits = normalizeModelLimits(provider, model);
   return {
     id,
@@ -4881,9 +4775,7 @@ var GithubLLM = class extends BaseLLM {
       logLevel: "INFO",
       enableConsoleLog: false,
       maxFileSize: 5 * 1024 * 1024,
-      // 5MB before rotation
       captureResponseBody: true
-      // Disable for better performance
     });
     return new GithubAuthTransport(base, {
       baseUrl: this.opts.apiUrl,
@@ -4891,6 +4783,17 @@ var GithubLLM = class extends BaseLLM {
       accessToken: this.opts.accessToken
     });
   }
+  transformUsage(rawUsage) {
+    if (!rawUsage) return void 0;
+    const usage = rawUsage;
+    return {
+      prompt_tokens: usage.prompt_tokens,
+      completion_tokens: usage.completion_tokens,
+      total_tokens: usage.total_tokens,
+      prompt_tokens_details: usage.prompt_tokens_details,
+      completion_tokens_details: usage.completion_tokens_details
+    };
+  }
   async getModels(signal) {
     const res = await this.getTransport().get("/models", void 0, signal);
     if (!res.ok) {
@@ -5229,6 +5132,28 @@ var AnthropicAISDKLLM = class {
     }
     return "auto";
   }
+  transformUsage(rawUsage) {
+    if (!rawUsage) {
+      return { prompt_tokens: 0, completion_tokens: 0, total_tokens: 0 };
+    }
+    const inputTokens = rawUsage.inputTokens ?? rawUsage.input_tokens ?? 0;
+    const outputTokens = rawUsage.outputTokens ?? rawUsage.output_tokens ?? 0;
+    const cacheCreation = rawUsage.cacheCreationInputTokens ?? rawUsage.cache_creation_input_tokens ?? 0;
+    const cacheRead = rawUsage.cacheReadInputTokens ?? rawUsage.cache_read_input_tokens ?? 0;
+    const cachedTokens = rawUsage.cachedInputTokens ?? cacheCreation + cacheRead;
+    const promptTokens = inputTokens + cachedTokens;
+    const totalTokens = promptTokens + outputTokens;
+    return {
+      prompt_tokens: promptTokens,
+      completion_tokens: outputTokens,
+      total_tokens: totalTokens,
+      prompt_tokens_details: {
+        cached_tokens: cachedTokens
+      },
+      ...cacheCreation > 0 && { cache_creation_input_tokens: cacheCreation },
+      ...cacheRead > 0 && { cache_read_input_tokens: cacheRead }
+    };
+  }
   handleError(error) {
     if (error instanceof LLMError) {
       throw error;
@@ -5297,18 +5222,7 @@ var AnthropicAISDKLLM = class {
         }
       })) : void 0;
       const rawUsage = result.usage;
-      const usage = rawUsage ? {
-        prompt_tokens: rawUsage.inputTokens || 0,
-        completion_tokens: rawUsage.outputTokens || 0,
-        total_tokens: rawUsage.totalTokens || 0,
-        prompt_tokens_details: {
-          cached_tokens: rawUsage.cachedInputTokens || 0
-        }
-      } : {
-        prompt_tokens: 0,
-        completion_tokens: 0,
-        total_tokens: 0
-      };
+      const usage = this.transformUsage(rawUsage);
       return {
         content: result.text,
         ...tool_calls && { tool_calls },
@@ -5329,9 +5243,10 @@ var AnthropicAISDKLLM = class {
       messages,
       tools,
       toolChoice,
-      maxOutputTokens: params.maxTokens ?? 10240,
+      maxOutputTokens: params.maxTokens,
       temperature: params.temperature,
       abortSignal: signal,
+      maxRetries: 10,
       onError: (event) => {
         streamError = event.error;
       },
@@ -5357,18 +5272,7 @@ var AnthropicAISDKLLM = class {
         }
       })) : void 0;
       const rawUsage = await result.usage;
-      const usage = rawUsage ? {
-        prompt_tokens: rawUsage.inputTokens || 0,
-        completion_tokens: rawUsage.outputTokens || 0,
-        total_tokens: rawUsage.totalTokens || 0,
-        prompt_tokens_details: {
-          cached_tokens: rawUsage.cachedInputTokens || 0
-        }
-      } : {
-        prompt_tokens: 0,
-        completion_tokens: 0,
-        total_tokens: 0
-      };
+      const usage = this.transformUsage(rawUsage);
       const finishReason = await result.finishReason;
       await handlers.onStreamFinish?.(finishReason, usage);
       return {
@@ -5380,6 +5284,54 @@ var AnthropicAISDKLLM = class {
       this.handleError(error);
     }
   }
+  async getModels(signal) {
+    const baseURL = this.opts.baseURL || this.opts.apiUrl || "https://api.anthropic.com/v1";
+    const url = `${baseURL}/models`;
+    const headers = {
+      "anthropic-version": "2023-06-01",
+      "anthropic-beta": " oauth-2025-04-20,fine-grained-tool-streaming-2025-05-14"
+    };
+    if (this.opts.oauth) {
+      headers.authorization = `Bearer ${this.opts.oauth.access}`;
+    } else if (this.opts.apiKey) {
+      headers["x-api-key"] = this.opts.apiKey;
+    } else {
+      throw new LLMError("No API key or OAuth credentials provided", 401, false);
+    }
+    try {
+      const response = await fetch(url, {
+        method: "GET",
+        headers,
+        signal
+      });
+      if (!response.ok) {
+        if ((response.status === 401 || response.status === 403) && this.opts.oauth) {
+          await this.ensureValidToken();
+          headers.authorization = `Bearer ${this.opts.oauth.access}`;
+          const retryResponse = await fetch(url, {
+            method: "GET",
+            headers,
+            signal
+          });
+          if (!retryResponse.ok) {
+            const text2 = await retryResponse.text();
+            throw new LLMError(text2 || `Failed to fetch models: ${retryResponse.status}`, retryResponse.status);
+          }
+          const data2 = await retryResponse.json();
+          return data2.data.map((model) => normalizeModelInfo("anthropic", model));
+        }
+        const text = await response.text();
+        throw new LLMError(text || `Failed to fetch models: ${response.status}`, response.status);
+      }
+      const data = await response.json();
+      return data.data.map((model) => normalizeModelInfo("anthropic", model));
+    } catch (error) {
+      if (error instanceof LLMError) {
+        throw error;
+      }
+      throw new LLMError(error instanceof Error ? error.message : "Failed to fetch models", void 0, false, error);
+    }
+  }
 };
 
 // src/llm-providers/llm-provider-config.json

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@nuvin/nuvin-core",
-  "version": "1.4.4",
+  "version": "1.5.2",
   "description": "",
   "private": false,
   "main": "dist/index.js",
@@ -46,8 +46,8 @@
     "dev": "tsup --watch",
     "build": "node scripts/build.js",
     "clean": "rm -rf dist",
-    "test": "vitest run",
-    "test:watch": "vitest",
+    "test": "vitest --typecheck run",
+    "test:watch": "vitest --typecheck",
     "format": "pnpm exec biome format --changed --no-errors-on-unmatched --write ."
   }
 }