@massa-ai/tools-api 1.32.0 → 1.33.0

package/dist/index.js

@@ -410,9 +410,9 @@ var init_massa_ai_config = __esm(() => {
     },
     embedding: {
       provider: "ollama",
-      model: "nomic-embed-text:latest",
+      model: "qwen3-embedding:4b",
       baseURL: "http://localhost:11434",
-      dimensions: 768
+      dimensions: 2560
     },
     compression: {
       defaultStrategy: "code_structure",
@@ -51290,6 +51290,18 @@ function _extractJsonObject(text2) {
 function timeoutSignal(timeoutMs) {
   return AbortSignal.timeout(timeoutMs);
 }
+function _isAbortOrTimeoutError(err) {
+  let cur = err;
+  for (let hops = 0;cur && hops < 5; hops++, cur = cur.cause) {
+    const name24 = typeof cur.name === "string" ? cur.name : "";
+    if (name24 === "TimeoutError" || name24 === "AbortError")
+      return true;
+    const message = typeof cur.message === "string" ? cur.message : "";
+    if (/timed out|operation was aborted/i.test(message))
+      return true;
+  }
+  return false;
+}
 async function llmComplete(prompt, opts = {}) {
   if (!isLlmEnabled()) {
     return { ok: false, error: "llm disabled" };
@@ -51387,7 +51399,7 @@ async function llmObject(prompt, schema, opts = {}) {
     });
     return { ok: false, error: "schema validation failed (fallback path)" };
   } catch (e) {
-    if (llm.disableThink) {
+    if (llm.disableThink && !_isAbortOrTimeoutError(e)) {
       const reasoning = _reasoningToText(result).length > 0 ? _reasoningToText(result) : _reasoningToText(e);
       if (reasoning.length > 0) {
         const parsed = _extractJsonObject(reasoning);
@@ -93979,7 +93991,7 @@ function createProvider(config3, providerId) {
   }
   return new AISDKEmbeddingProvider(config3, providerId);
 }
-var AISDKEmbeddingProvider;
+var OLLAMA_EMBED_NUM_CTX, AISDKEmbeddingProvider;
 var init_provider = __esm(() => {
   init_dist6();
   init_dist7();
@@ -93990,7 +94002,9 @@ var init_provider = __esm(() => {
   init_metrics2();
   init_rate_limiter2();
   init_dist();
+  init_config();
   init_local_transformers();
+  OLLAMA_EMBED_NUM_CTX = parsePositiveIntEnv(process.env.OLLAMA_EMBEDDING_NUM_CTX, 8192);
   AISDKEmbeddingProvider = class AISDKEmbeddingProvider {
     config;
     providerId;
@@ -94138,7 +94152,8 @@ var init_provider = __esm(() => {
               const inputText = this.sanitizeText(this.truncateText(text2));
               const response = await this.ollamaFetch("/api/embed", {
                 model: this.model,
-                input: inputText
+                input: inputText,
+                options: { num_ctx: OLLAMA_EMBED_NUM_CTX }
               });
               if (!response.ok) {
                 throw new Error(`Ollama API error: ${response.status} ${response.statusText}`);
@@ -94228,7 +94243,8 @@ var init_provider = __esm(() => {
           return await withTimeout(() => withRetry(async () => {
             const response = await this.ollamaFetch("/api/embed", {
               model: this.model,
-              input: texts.map((t2) => this.sanitizeText(this.truncateText(t2)))
+              input: texts.map((t2) => this.sanitizeText(this.truncateText(t2))),
+              options: { num_ctx: OLLAMA_EMBED_NUM_CTX }
             });
             if (!response.ok) {
               throw new Error(`Ollama batch API error: ${response.status} ${response.statusText}`);
@@ -105428,18 +105444,36 @@ function hasApiKey(providerName) {
   }
   return !!config3.apiKey;
 }
-var embeddingProviders;
+var fileEmbedding, selectedProvider, fileFor = (provider) => fileEmbedding?.provider === provider ? fileEmbedding : undefined, SELECTABLE_PROVIDERS, embeddingProviders;
 var init_config2 = __esm(() => {
   init_config();
+  init_dist();
+  fileEmbedding = loadConfigSafe().embedding;
+  selectedProvider = process.env.EMBEDDING_PROVIDER || fileEmbedding?.provider || "ollama";
+  SELECTABLE_PROVIDERS = new Set([
+    "ollama",
+    "mistral",
+    "google",
+    "openai",
+    "vercel",
+    "litellm",
+    "custom",
+    "transformers",
+    "local"
+  ]);
+  if (!SELECTABLE_PROVIDERS.has(selectedProvider)) {
+    logger.warn(`[EmbeddingConfig] selected provider "${selectedProvider}" has no runtime entry \u2014 falling back to priority order`, { source: process.env.EMBEDDING_PROVIDER ? "EMBEDDING_PROVIDER env" : "config.json embedding.provider" });
+  }
   embeddingProviders = {
     google: (() => {
-      const model = process.env.GOOGLE_EMBEDDING_MODEL || "gemini-embedding-001";
+      const file3 = fileFor("google");
+      const model = process.env.GOOGLE_EMBEDDING_MODEL || file3?.model || "gemini-embedding-001";
       return {
         provider: "google",
         model,
-        apiKey: process.env.GOOGLE_GENERATIVE_AI_API_KEY || process.env.GOOGLE_API_KEY,
-        dimensions: 3072,
-        priority: process.env.EMBEDDING_PROVIDER === "google" ? 1 : 10,
+        apiKey: process.env.GOOGLE_GENERATIVE_AI_API_KEY || process.env.GOOGLE_API_KEY || file3?.apiKey,
+        dimensions: file3?.dimensions ?? 3072,
+        priority: selectedProvider === "google" ? 1 : 10,
         timeout: 60000,
         maxRetries: 3,
         maxChars: getMaxChars("GOOGLE", model),
@@ -105454,7 +105488,7 @@ var init_config2 = __esm(() => {
         apiKey: process.env.AI_GATEWAY_API_KEY || process.env.VERCEL_AI_GATEWAY_API_KEY,
         baseURL: process.env.VERCEL_AI_GATEWAY_URL,
         dimensions: Number(process.env.VERCEL_EMBEDDING_DIMENSIONS || "4096"),
-        priority: process.env.EMBEDDING_PROVIDER === "vercel" ? 1 : 20,
+        priority: selectedProvider === "vercel" ? 1 : 20,
         timeout: 60000,
         maxRetries: 3,
         maxChars: getMaxChars("VERCEL", model),
@@ -105462,13 +105496,14 @@ var init_config2 = __esm(() => {
       };
     })(),
     ollama: (() => {
-      const model = process.env.OLLAMA_EMBEDDING_MODEL || "qwen3-embedding:8b";
+      const file3 = fileFor("ollama");
+      const model = process.env.OLLAMA_EMBEDDING_MODEL || file3?.model || "qwen3-embedding:4b";
       return {
         provider: "ollama",
         model,
-        baseURL: process.env.OLLAMA_BASE_URL || "http://localhost:11434",
-        dimensions: Number(process.env.OLLAMA_EMBEDDING_DIMENSIONS || "4096"),
-        priority: process.env.EMBEDDING_PROVIDER === "ollama" || !process.env.EMBEDDING_PROVIDER ? 1 : 50,
+        baseURL: process.env.OLLAMA_BASE_URL || file3?.baseURL || "http://localhost:11434",
+        dimensions: process.env.OLLAMA_EMBEDDING_DIMENSIONS ? Number(process.env.OLLAMA_EMBEDDING_DIMENSIONS) : file3?.dimensions ?? 2560,
+        priority: selectedProvider === "ollama" ? 1 : 50,
         timeout: 300000,
         maxRetries: 2,
         maxChars: getMaxChars("OLLAMA", model),
@@ -105476,13 +105511,14 @@ var init_config2 = __esm(() => {
       };
     })(),
     mistralText: (() => {
-      const model = process.env.MISTRAL_TEXT_EMBEDDING_MODEL || "mistral-embed";
+      const file3 = fileFor("mistral");
+      const model = process.env.MISTRAL_TEXT_EMBEDDING_MODEL || file3?.model || "mistral-embed";
       return {
         provider: "mistral",
         model,
-        apiKey: process.env.MISTRAL_API_KEY,
-        dimensions: 1024,
-        priority: process.env.EMBEDDING_PROVIDER === "mistral" ? 1 : 2,
+        apiKey: process.env.MISTRAL_API_KEY || file3?.apiKey,
+        dimensions: file3?.dimensions ?? 1024,
+        priority: selectedProvider === "mistral" ? 1 : 2,
         timeout: 60000,
         maxRetries: 3,
         maxChars: getMaxChars("MISTRAL", model),
@@ -105494,9 +105530,9 @@ var init_config2 = __esm(() => {
       return {
         provider: "mistral",
         model,
-        apiKey: process.env.MISTRAL_API_KEY,
+        apiKey: process.env.MISTRAL_API_KEY || fileFor("mistral")?.apiKey,
         dimensions: 1536,
-        priority: process.env.EMBEDDING_PROVIDER === "mistral" ? 1 : 3,
+        priority: selectedProvider === "mistral" ? 1 : 3,
         timeout: 60000,
         maxRetries: 3,
         maxChars: getMaxChars("MISTRAL", model),
@@ -105511,7 +105547,7 @@ var init_config2 = __esm(() => {
         apiKey: process.env.LITELLM_API_KEY,
         baseURL: process.env.LITELLM_BASE_URL,
         dimensions: Number(process.env.LITELLM_EMBEDDING_DIMENSIONS || "1024"),
-        priority: process.env.EMBEDDING_PROVIDER === "litellm" ? 1 : 15,
+        priority: selectedProvider === "litellm" ? 1 : 15,
         timeout: Number(process.env.LITELLM_EMBEDDING_TIMEOUT || "60000"),
         maxRetries: 3,
         maxChars: getMaxChars("LITELLM", model),
@@ -105526,7 +105562,7 @@ var init_config2 = __esm(() => {
         apiKey: process.env.CUSTOM_API_KEY,
         baseURL: process.env.CUSTOM_EMBEDDING_BASE_URL,
         dimensions: Number(process.env.CUSTOM_EMBEDDING_DIMENSIONS || "1536"),
-        priority: process.env.EMBEDDING_PROVIDER === "custom" ? 1 : 100,
+        priority: selectedProvider === "custom" ? 1 : 100,
         timeout: Number(process.env.CUSTOM_EMBEDDING_TIMEOUT || "60000"),
         maxRetries: 3,
         maxChars: getMaxChars("CUSTOM", model),
@@ -105539,7 +105575,7 @@ var init_config2 = __esm(() => {
         provider: "transformers",
         model,
         dimensions: Number(process.env.TRANSFORMERS_EMBEDDING_DIMENSIONS || "384"),
-        priority: process.env.EMBEDDING_PROVIDER === "transformers" ? 1 : 100,
+        priority: selectedProvider === "transformers" ? 1 : 100,
         timeout: 300000,
         maxRetries: 1,
         maxChars: getMaxChars("TRANSFORMERS", model),
@@ -105552,12 +105588,27 @@ var init_config2 = __esm(() => {
         provider: "transformers",
         model,
         dimensions: Number(process.env.TRANSFORMERS_EMBEDDING_DIMENSIONS || "384"),
-        priority: process.env.EMBEDDING_PROVIDER === "local" ? 1 : 100,
+        priority: selectedProvider === "local" ? 1 : 100,
         timeout: 300000,
         maxRetries: 1,
         maxChars: getMaxChars("TRANSFORMERS", model),
         rateLimits: getRateLimits("TRANSFORMERS")
       };
+    })(),
+    openai: (() => {
+      const file3 = fileFor("openai");
+      const model = process.env.OPENAI_EMBEDDING_MODEL || file3?.model || "text-embedding-3-small";
+      return {
+        provider: "openai",
+        model,
+        apiKey: process.env.OPENAI_API_KEY || file3?.apiKey,
+        dimensions: file3?.dimensions ?? 1536,
+        priority: selectedProvider === "openai" ? 1 : 10,
+        timeout: 60000,
+        maxRetries: 3,
+        maxChars: getMaxChars("OPENAI", model),
+        rateLimits: getRateLimits("OPENAI")
+      };
     })()
   };
 });
@@ -179113,7 +179164,7 @@ var systemRoutes = new Elysia({ prefix: "/api/v1/system" }).get("/info", async (
   return {
     ...ollamaStatus,
     models,
-    configuredModel: process.env.OLLAMA_EMBEDDING_MODEL || "qwen3-embedding:8b",
+    configuredModel: process.env.OLLAMA_EMBEDDING_MODEL || "qwen3-embedding:4b",
     baseUrl: process.env.OLLAMA_BASE_URL || "http://localhost:11434"
   };
 }, {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@massa-ai/tools-api",
-  "version": "1.32.0",
+  "version": "1.33.0",
   "author": "luizgmassa",
   "description": "massa-ai REST API server - Semantic code search, memory, and context compression",
   "type": "module",
@@ -21,8 +21,8 @@
     "test": "bun scripts/run-tests-isolated.ts"
   },
   "dependencies": {
-    "@massa-ai/core": "^1.32.0",
-    "@massa-ai/shared": "^1.32.0",
+    "@massa-ai/core": "^1.33.0",
+    "@massa-ai/shared": "^1.33.0",
     "elysia": "^1.2.25",
     "@elysiajs/swagger": "^1.2.0",
     "@elysiajs/cors": "^1.2.0",