@retrivora-ai/rag-engine 1.9.7 → 1.9.8

package/dist/server.mjs

@@ -59,9 +59,9 @@ var __forAwait = (obj, it, method) => (it = obj[__knownSymbol("asyncIterator")])
 function isRecord(value) {
   return typeof value === "object" && value !== null;
 }
-function resolvePath(obj, path) {
-  if (!path || !obj) return void 0;
-  return path.replace(/\[(\w+)\]/g, ".$1").replace(/^\./, "").split(".").reduce((current, segment) => {
+function resolvePath(obj, path2) {
+  if (!path2 || !obj) return void 0;
+  return path2.replace(/\[(\w+)\]/g, ".$1").replace(/^\./, "").split(".").reduce((current, segment) => {
     if (!isRecord(current) && !Array.isArray(current)) {
       return void 0;
     }
@@ -171,7 +171,7 @@ var init_PineconeProvider = __esm({
       static getHealthChecker() {
         return {
           async check(config) {
-            var _a, _b;
+            var _a2, _b;
             const opts = config.options || {};
             const indexName = config.indexName;
             const timestamp = Date.now();
@@ -179,7 +179,7 @@ var init_PineconeProvider = __esm({
               const { Pinecone: Pinecone2 } = await import("@pinecone-database/pinecone");
               const client = new Pinecone2({ apiKey: opts.apiKey });
               const indexes = await client.listIndexes();
-              const indexNames = (_b = (_a = indexes.indexes) == null ? void 0 : _a.map((i) => i.name)) != null ? _b : [];
+              const indexNames = (_b = (_a2 = indexes.indexes) == null ? void 0 : _a2.map((i) => i.name)) != null ? _b : [];
               if (!indexNames.includes(indexName)) {
                 return {
                   healthy: false,
@@ -206,10 +206,10 @@ var init_PineconeProvider = __esm({
         };
       }
       async initialize() {
-        var _a, _b;
+        var _a2, _b;
         this.client = new Pinecone({ apiKey: this.apiKey });
         const indexes = await this.client.listIndexes();
-        const names = (_b = (_a = indexes.indexes) == null ? void 0 : _a.map((i) => i.name)) != null ? _b : [];
+        const names = (_b = (_a2 = indexes.indexes) == null ? void 0 : _a2.map((i) => i.name)) != null ? _b : [];
         if (!names.includes(this.indexName)) {
           throw new Error(
             `[PineconeProvider] Index "${this.indexName}" not found. Available: ${names.join(", ")}`
@@ -221,12 +221,12 @@ var init_PineconeProvider = __esm({
         return namespace ? idx.namespace(namespace) : idx.namespace("");
       }
       async upsert(doc, namespace) {
-        var _a;
+        var _a2;
         await this.index(namespace).upsert({
           records: [{
             id: String(doc.id),
             values: doc.vector,
-            metadata: __spreadValues({ content: doc.content }, (_a = doc.metadata) != null ? _a : {})
+            metadata: __spreadValues({ content: doc.content }, (_a2 = doc.metadata) != null ? _a2 : {})
           }]
         });
       }
@@ -234,29 +234,29 @@ var init_PineconeProvider = __esm({
         const BATCH = 100;
         for (let i = 0; i < docs.length; i += BATCH) {
           const records = docs.slice(i, i + BATCH).map((d) => {
-            var _a;
+            var _a2;
             return {
               id: String(d.id),
               values: d.vector,
-              metadata: __spreadValues({ content: d.content }, (_a = d.metadata) != null ? _a : {})
+              metadata: __spreadValues({ content: d.content }, (_a2 = d.metadata) != null ? _a2 : {})
             };
           });
           await this.index(namespace).upsert({ records });
         }
       }
       async query(vector, topK, namespace, filter) {
-        var _a;
+        var _a2;
         const pineconeFilter = this.sanitizeFilter(filter);
         const result = await this.index(namespace).query(__spreadValues({
           vector,
           topK,
           includeMetadata: true
         }, Object.keys(pineconeFilter).length > 0 ? { filter: pineconeFilter } : {}));
-        return ((_a = result.matches) != null ? _a : []).map((m) => {
-          var _a2, _b;
+        return ((_a2 = result.matches) != null ? _a2 : []).map((m) => {
+          var _a3, _b;
           return {
             id: m.id,
-            score: (_a2 = m.score) != null ? _a2 : 0,
+            score: (_a3 = m.score) != null ? _a3 : 0,
             content: ((_b = m.metadata) == null ? void 0 : _b.content) || "",
             metadata: m.metadata
           };
@@ -295,13 +295,13 @@ var init_PostgreSQLProvider = __esm({
     init_BaseVectorProvider();
     PostgreSQLProvider = class extends BaseVectorProvider {
       constructor(config) {
-        var _a;
+        var _a2;
         super(config);
         this.tableName = this.indexName.replace(/[^a-z0-9_]/gi, "_");
         const opts = config.options;
         if (!opts.connectionString) throw new Error("[PostgreSQLProvider] options.connectionString is required");
         this.connectionString = opts.connectionString;
-        this.dimensions = (_a = opts.dimensions) != null ? _a : 1536;
+        this.dimensions = (_a2 = opts.dimensions) != null ? _a2 : 1536;
       }
       static getValidator() {
         return {
@@ -382,7 +382,7 @@ var init_PostgreSQLProvider = __esm({
         }
       }
       async upsert(doc, namespace = "") {
-        var _a;
+        var _a2;
         const vectorLiteral = `[${doc.vector.join(",")}]`;
         await this.pool.query(
           `INSERT INTO ${this.tableName} (id, namespace, content, metadata, embedding)
@@ -392,7 +392,7 @@ var init_PostgreSQLProvider = __esm({
              content   = EXCLUDED.content,
              metadata  = EXCLUDED.metadata,
              embedding = EXCLUDED.embedding`,
-          [doc.id, namespace, doc.content, JSON.stringify((_a = doc.metadata) != null ? _a : {}), vectorLiteral]
+          [doc.id, namespace, doc.content, JSON.stringify((_a2 = doc.metadata) != null ? _a2 : {}), vectorLiteral]
         );
       }
       async batchUpsert(docs, namespace = "") {
@@ -405,9 +405,9 @@ var init_PostgreSQLProvider = __esm({
             const batch = docs.slice(i, i + BATCH_SIZE);
             const values = [];
             const valuePlaceholders = batch.map((doc, idx) => {
-              var _a;
+              var _a2;
               const offset = idx * 5;
-              values.push(doc.id, namespace, doc.content, JSON.stringify((_a = doc.metadata) != null ? _a : {}), `[${doc.vector.join(",")}]`);
+              values.push(doc.id, namespace, doc.content, JSON.stringify((_a2 = doc.metadata) != null ? _a2 : {}), `[${doc.vector.join(",")}]`);
               return `($${offset + 1}, $${offset + 2}, $${offset + 3}, $${offset + 4}, $${offset + 5}::vector)`;
             }).join(", ");
             const query = `
@@ -490,8 +490,8 @@ var init_PostgreSQLProvider = __esm({
 
 // src/utils/synonyms.ts
 function resolveMetadataValue(meta, uiKey) {
-  var _a;
-  const synonyms = (_a = FIELD_SYNONYMS[uiKey]) != null ? _a : [];
+  var _a2;
+  const synonyms = (_a2 = FIELD_SYNONYMS[uiKey]) != null ? _a2 : [];
   const keys = Object.keys(meta);
   const systemKeys = ["namespace", "filename", "filesize", "filetype", "docid", "uploadedat", "dimension", "chunkindex"];
   let match = keys.find((k) => k.toLowerCase() === uiKey.toLowerCase());
@@ -578,14 +578,14 @@ var init_MultiTablePostgresProvider = __esm({
     init_synonyms();
     MultiTablePostgresProvider = class extends BaseVectorProvider {
       constructor(config) {
-        var _a, _b, _c;
+        var _a2, _b, _c;
         super(config);
         const opts = config.options || {};
         if (!opts.connectionString) {
           throw new Error("[MultiTablePostgresProvider] options.connectionString is required");
         }
         this.connectionString = opts.connectionString;
-        this.dimensions = (_a = opts.dimensions) != null ? _a : 768;
+        this.dimensions = (_a2 = opts.dimensions) != null ? _a2 : 768;
         this.tables = [];
         const rawSearchFields = (_c = (_b = opts.searchFields) != null ? _b : process.env.VECTOR_DB_SEARCH_FIELDS) != null ? _c : ["name", "product_name", "productname", "title"];
         this.searchFields = typeof rawSearchFields === "string" ? rawSearchFields.split(",").map((f) => f.trim()).filter(Boolean) : rawSearchFields;
@@ -643,11 +643,11 @@ var init_MultiTablePostgresProvider = __esm({
        * Batch upsert documents by dynamically provisioning tables and columns based on CSV headers.
        */
       async batchUpsert(docs, namespace = "") {
-        var _a;
+        var _a2;
         if (docs.length === 0) return;
         const docsByFile = {};
         for (const doc of docs) {
-          const fileName = ((_a = doc.metadata) == null ? void 0 : _a.fileName) || this.uploadTable;
+          const fileName = ((_a2 = doc.metadata) == null ? void 0 : _a2.fileName) || this.uploadTable;
           if (!docsByFile[fileName]) docsByFile[fileName] = [];
           docsByFile[fileName].push(doc);
         }
@@ -700,11 +700,11 @@ var init_MultiTablePostgresProvider = __esm({
               const headerColumns = csvHeaders.map((h) => `"${h}"`).join(", ");
               const allColumns = `id, ${headerColumns ? headerColumns + ", " : ""}namespace, content, metadata, embedding`;
               const valuePlaceholders = batch.map((doc, idx) => {
-                var _a2, _b;
+                var _a3, _b;
                 const offset = idx * (5 + csvHeaders.length);
                 values.push(doc.id);
                 for (const h of csvHeaders) {
-                  values.push(String(((_a2 = doc.metadata) == null ? void 0 : _a2[h]) || ""));
+                  values.push(String(((_a3 = doc.metadata) == null ? void 0 : _a3[h]) || ""));
                 }
                 values.push(namespace, doc.content, JSON.stringify((_b = doc.metadata) != null ? _b : {}), `[${doc.vector.join(",")}]`);
                 const docPlaceholders = [];
@@ -743,7 +743,7 @@ var init_MultiTablePostgresProvider = __esm({
        * Query all configured tables and merge results, sorted by cosine similarity score.
        */
       async query(vector, topK, _namespace, _filter) {
-        var _a;
+        var _a2;
         if (!this.pool) {
           throw new Error("[MultiTablePostgresProvider] Provider not initialized. Call initialize() first.");
         }
@@ -774,11 +774,11 @@ var init_MultiTablePostgresProvider = __esm({
             const filterParams = [];
             if (metadataFilters && Object.keys(metadataFilters).length > 0) {
               const conditions = Object.entries(metadataFilters).map(([key, val]) => {
-                var _a3;
+                var _a4;
                 filterParams.push(String(val));
                 const baseOffset = queryText && dynamicKeywordQuery ? 2 : 1;
                 const paramIdx = baseOffset + filterParams.length;
-                const synonyms = (_a3 = FIELD_SYNONYMS[key]) != null ? _a3 : [];
+                const synonyms = (_a4 = FIELD_SYNONYMS[key]) != null ? _a4 : [];
                 const keysToCheck = [key, ...synonyms];
                 const coalesceExprs = keysToCheck.flatMap((k) => [
                   `metadata->>'${k}'`,
@@ -847,7 +847,7 @@ var init_MultiTablePostgresProvider = __esm({
             }
             const tableResults = [];
             for (const row of result.rows) {
-              const _a2 = row, { hybrid_score, id } = _a2, rest = __objRest(_a2, ["hybrid_score", "id"]);
+              const _a3 = row, { hybrid_score, id } = _a3, rest = __objRest(_a3, ["hybrid_score", "id"]);
               delete rest.embedding;
               delete rest.vector_score;
               delete rest.keyword_score;
@@ -876,10 +876,10 @@ var init_MultiTablePostgresProvider = __esm({
         }
         allResults.sort((a, b) => b.score - a.score);
         if (allResults.length === 0) return [];
-        const bestMatchTable = (_a = allResults[0].metadata) == null ? void 0 : _a.source_table;
+        const bestMatchTable = (_a2 = allResults[0].metadata) == null ? void 0 : _a2.source_table;
         const finalSorted = allResults.filter((res) => {
-          var _a2;
-          return ((_a2 = res.metadata) == null ? void 0 : _a2.source_table) === bestMatchTable;
+          var _a3;
+          return ((_a3 = res.metadata) == null ? void 0 : _a3.source_table) === bestMatchTable;
         });
         console.log(`[MultiTablePostgresProvider] --- Search Complete ---`);
         console.log(`[MultiTablePostgresProvider] Final top match from "${bestMatchTable}" with score ${finalSorted[0].score.toFixed(4)}`);
@@ -1284,14 +1284,14 @@ var init_QdrantProvider = __esm({
        * Samples points from the collection to discover available payload fields.
        */
       async discoverSchema() {
-        var _a;
+        var _a2;
         try {
           const { data } = await this.http.post(`/collections/${this.indexName}/points/scroll`, {
             limit: 20,
             with_payload: true,
             with_vector: false
           });
-          const points = ((_a = data.result) == null ? void 0 : _a.points) || [];
+          const points = ((_a2 = data.result) == null ? void 0 : _a2.points) || [];
           const keys = /* @__PURE__ */ new Set();
           for (const point of points) {
             const payload = point.payload || {};
@@ -1317,12 +1317,12 @@ var init_QdrantProvider = __esm({
        * Ensures the collection exists. Creates it if missing.
        */
       async ensureCollection() {
-        var _a;
+        var _a2;
         try {
           await this.http.get(`/collections/${this.indexName}`);
           console.log(`[QdrantProvider] \u2705 Collection "${this.indexName}" already exists.`);
         } catch (err) {
-          if (axios4.isAxiosError(err) && ((_a = err.response) == null ? void 0 : _a.status) === 404) {
+          if (axios4.isAxiosError(err) && ((_a2 = err.response) == null ? void 0 : _a2.status) === 404) {
             const opts = this.config.options;
             const dimensionsForCreate = opts.dimensions || 1536;
             console.log(`[QdrantProvider] \u23F3 Creating collection "${this.indexName}" (dimensions: ${dimensionsForCreate})...`);
@@ -1342,7 +1342,7 @@ var init_QdrantProvider = __esm({
        * Ensures that a payload field has an index.
        */
       async ensureIndex(fieldName, schema = "keyword") {
-        var _a;
+        var _a2;
         let fullPath = fieldName;
         if (!this.isFlatPayload && !fieldName.includes(".") && fieldName !== "namespace" && fieldName !== this.contentField) {
           fullPath = `${this.metadataField}.${fieldName}`;
@@ -1355,7 +1355,7 @@ var init_QdrantProvider = __esm({
           console.log(`[QdrantProvider] \u2705 Ensured ${schema} index for "${fullPath}"`);
         } catch (err) {
           let status;
-          if (axios4.isAxiosError(err)) status = (_a = err.response) == null ? void 0 : _a.status;
+          if (axios4.isAxiosError(err)) status = (_a2 = err.response) == null ? void 0 : _a2.status;
           if (status === 409) return;
           console.warn(`[QdrantProvider] \u26A0\uFE0F Could not ensure index for "${fullPath}":`, err instanceof Error ? err.message : String(err));
         }
@@ -1384,7 +1384,7 @@ var init_QdrantProvider = __esm({
         await this.http.put(`/collections/${this.indexName}/points`, payload);
       }
       async query(vector, topK, namespace, _filter) {
-        var _a;
+        var _a2;
         const must = [];
         if (namespace) {
           must.push({ key: "namespace", match: { value: namespace } });
@@ -1406,7 +1406,7 @@ var init_QdrantProvider = __esm({
           limit: topK,
           with_payload: true,
           params: {
-            hnsw_ef: ((_a = this.config.options) == null ? void 0 : _a.efSearch) || Math.max(topK * 20, 128),
+            hnsw_ef: ((_a2 = this.config.options) == null ? void 0 : _a2.efSearch) || Math.max(topK * 20, 128),
             exact: false
           },
           filter: must.length > 0 ? { must } : void 0
@@ -1501,13 +1501,13 @@ var init_ChromaDBProvider = __esm({
        * Get or create the ChromaDB collection.
        */
       async initialize() {
-        var _a;
+        var _a2;
         try {
           const { data } = await this.http.get(`/api/v1/collections/${this.indexName}`);
           this.collectionId = data.id;
           console.log(`[ChromaDBProvider] \u2705 Collection "${this.indexName}" found (id: ${this.collectionId})`);
         } catch (err) {
-          if (axios5.isAxiosError(err) && ((_a = err.response) == null ? void 0 : _a.status) === 404) {
+          if (axios5.isAxiosError(err) && ((_a2 = err.response) == null ? void 0 : _a2.status) === 404) {
             console.log(`[ChromaDBProvider] \u23F3 Collection "${this.indexName}" not found. Creating...`);
             const { data } = await this.http.post("/api/v1/collections", {
               name: this.indexName
@@ -1728,7 +1728,7 @@ var init_WeaviateProvider = __esm({
         await this.http.post("/v1/batch/objects", payload);
       }
       async query(vector, topK, namespace, _filter) {
-        var _a, _b;
+        var _a2, _b;
         const queryText = _filter == null ? void 0 : _filter.queryText;
         const sanitizedFilter = this.sanitizeFilter(_filter);
         const searchParams = queryText ? `hybrid: { query: ${JSON.stringify(queryText)}, alpha: 0.5 }` : `nearVector: { vector: ${JSON.stringify(vector)} }`;
@@ -1754,7 +1754,7 @@ var init_WeaviateProvider = __esm({
       `
         };
         const { data } = await this.http.post("/v1/graphql", graphqlQuery);
-        const results = ((_b = (_a = data.data) == null ? void 0 : _a.Get) == null ? void 0 : _b[this.indexName]) || [];
+        const results = ((_b = (_a2 = data.data) == null ? void 0 : _a2.Get) == null ? void 0 : _b[this.indexName]) || [];
         return results.map((res) => ({
           id: res["_additional"].id,
           score: 1 - res["_additional"].distance,
@@ -1807,10 +1807,10 @@ var init_WeaviateProvider = __esm({
         return `{ operator: And, operands: [${operands.join(", ")}] }`;
       }
       weaviateOperand(key, value) {
-        const path = `path: [${JSON.stringify(key)}], operator: Equal`;
-        if (typeof value === "number") return `{ ${path}, valueNumber: ${value} }`;
-        if (typeof value === "boolean") return `{ ${path}, valueBoolean: ${value} }`;
-        return `{ ${path}, valueString: ${JSON.stringify(value)} }`;
+        const path2 = `path: [${JSON.stringify(key)}], operator: Equal`;
+        if (typeof value === "number") return `{ ${path2}, valueNumber: ${value} }`;
+        if (typeof value === "boolean") return `{ ${path2}, valueBoolean: ${value} }`;
+        return `{ ${path2}, valueString: ${JSON.stringify(value)} }`;
       }
     };
   }
@@ -1837,21 +1837,21 @@ var init_UniversalVectorProvider = __esm({
         }
       }
       async initialize() {
-        var _a;
+        var _a2;
         this.http = axios8.create({
           baseURL: this.opts.baseUrl,
           headers: __spreadValues({
             "Content-Type": "application/json"
           }, this.opts.headers),
-          timeout: (_a = this.opts.timeout) != null ? _a : 3e4
+          timeout: (_a2 = this.opts.timeout) != null ? _a2 : 3e4
         });
         if (!await this.ping()) {
           throw new Error(`[UniversalVectorProvider] Failed to connect to ${this.opts.baseUrl}`);
         }
       }
       async upsert(doc, namespace) {
-        var _a, _b, _c;
-        const endpoint = (_a = this.opts.upsertEndpoint) != null ? _a : "/upsert";
+        var _a2, _b, _c;
+        const endpoint = (_a2 = this.opts.upsertEndpoint) != null ? _a2 : "/upsert";
         const template = (_b = this.opts.upsertTemplate) != null ? _b : JSON.stringify({
           id: "{{id}}",
           vector: "{{vector}}",
@@ -1884,8 +1884,8 @@ var init_UniversalVectorProvider = __esm({
         }
       }
       async query(vector, topK, namespace, filter) {
-        var _a, _b;
-        const endpoint = (_a = this.opts.queryEndpoint) != null ? _a : "/query";
+        var _a2, _b;
+        const endpoint = (_a2 = this.opts.queryEndpoint) != null ? _a2 : "/query";
         const template = (_b = this.opts.queryTemplate) != null ? _b : JSON.stringify({
           vector: "{{vector}}",
           limit: "{{topK}}",
@@ -1911,12 +1911,12 @@ var init_UniversalVectorProvider = __esm({
             );
           }
           return results.map((item) => {
-            var _a2, _b2, _c, _d, _e, _f, _g;
+            var _a3, _b2, _c, _d, _e, _f, _g2;
             return {
-              id: item[(_a2 = this.opts.queryIdField) != null ? _a2 : "id"],
+              id: item[(_a3 = this.opts.queryIdField) != null ? _a3 : "id"],
               score: (_c = item[(_b2 = this.opts.queryScoreField) != null ? _b2 : "score"]) != null ? _c : 0,
               content: (_e = item[(_d = this.opts.queryContentField) != null ? _d : "content"]) != null ? _e : "",
-              metadata: (_g = item[(_f = this.opts.queryMetadataField) != null ? _f : "metadata"]) != null ? _g : {}
+              metadata: (_g2 = item[(_f = this.opts.queryMetadataField) != null ? _f : "metadata"]) != null ? _g2 : {}
             };
           });
         } catch (error) {
@@ -1926,8 +1926,8 @@ var init_UniversalVectorProvider = __esm({
         }
       }
       async delete(id, namespace) {
-        var _a, _b;
-        const endpoint = (_a = this.opts.deleteEndpoint) != null ? _a : "/delete";
+        var _a2, _b;
+        const endpoint = (_a2 = this.opts.deleteEndpoint) != null ? _a2 : "/delete";
         const template = (_b = this.opts.deleteTemplate) != null ? _b : JSON.stringify({
           id: "{{id}}",
           namespace: "{{namespace}}"
@@ -1945,8 +1945,8 @@ var init_UniversalVectorProvider = __esm({
         }
       }
       async deleteNamespace(namespace) {
-        var _a;
-        const endpoint = (_a = this.opts.deleteNamespaceEndpoint) != null ? _a : "/delete-namespace";
+        var _a2;
+        const endpoint = (_a2 = this.opts.deleteNamespaceEndpoint) != null ? _a2 : "/delete-namespace";
         try {
           await this.http.post(endpoint, { namespace });
         } catch (error) {
@@ -1956,9 +1956,9 @@ var init_UniversalVectorProvider = __esm({
         }
       }
       async ping() {
-        var _a;
+        var _a2;
         try {
-          const endpoint = (_a = this.opts.pingEndpoint) != null ? _a : "/health";
+          const endpoint = (_a2 = this.opts.pingEndpoint) != null ? _a2 : "/health";
           const response = await this.http.get(endpoint, { timeout: 5e3 });
           return response.status >= 200 && response.status < 300;
         } catch (e) {
@@ -2069,6 +2069,8 @@ var LLM_PROVIDERS = [
   "anthropic",
   "ollama",
   "gemini",
+  "groq",
+  "qwen",
   "rest",
   "universal_rest",
   "custom"
@@ -2077,6 +2079,7 @@ var EMBEDDING_PROVIDERS = [
   "openai",
   "ollama",
   "gemini",
+  "qwen",
   "rest",
   "universal_rest",
   "custom"
@@ -2085,6 +2088,7 @@ var PROVIDERS_WITH_EMBEDDINGS = [
   "openai",
   "ollama",
   "gemini",
+  "qwen",
   "rest",
   "universal_rest"
 ];
@@ -2092,8 +2096,8 @@ var UI_BORDER_RADIUS_OPTIONS = ["none", "sm", "md", "lg", "xl", "full"];
 
 // src/config/serverConfig.ts
 function readString(env, name) {
-  var _a;
-  const value = (_a = env[name]) == null ? void 0 : _a.trim();
+  var _a2;
+  const value = (_a2 = env[name]) == null ? void 0 : _a2.trim();
   return value ? value : void 0;
 }
 function readNumber(env, name, fallback) {
@@ -2106,8 +2110,8 @@ function readNumber(env, name, fallback) {
   return parsed;
 }
 function readEnum(env, name, fallback, allowed) {
-  var _a;
-  const value = (_a = readString(env, name)) != null ? _a : fallback;
+  var _a2;
+  const value = (_a2 = readString(env, name)) != null ? _a2 : fallback;
   if (allowed.includes(value)) {
     return value;
   }
@@ -2117,15 +2121,15 @@ function getRagConfig(baseConfig, env = process.env) {
   return getEnvConfig(env, baseConfig);
 }
 function getEnvConfig(env = process.env, base) {
-  var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j, _k, _l, _m, _n, _o, _p, _q, _r, _s, _t, _u, _v, _w, _x, _y, _z, _A, _B, _C, _D, _E, _F, _G, _H, _I, _J, _K, _L, _M, _N, _O, _P, _Q, _R, _S, _T, _U, _V, _W, _X, _Y, _Z, __, _$, _aa, _ba, _ca, _da, _ea, _fa, _ga, _ha, _ia, _ja, _ka, _la, _ma, _na, _oa, _pa, _qa, _ra, _sa, _ta, _ua, _va, _wa, _xa, _ya, _za, _Aa, _Ba, _Ca, _Da, _Ea, _Fa, _Ga, _Ha, _Ia, _Ja;
-  const projectId = (_c = (_b = (_a = readString(env, "RAG_PROJECT_ID")) != null ? _a : readString(env, "NEXT_PUBLIC_PROJECT_ID")) != null ? _b : base == null ? void 0 : base.projectId) != null ? _c : "__default__";
+  var _a2, _b, _c, _d, _e, _f, _g2, _h, _i, _j, _k, _l, _m, _n, _o, _p, _q, _r, _s, _t, _u, _v, _w, _x, _y, _z, _A, _B, _C, _D, _E, _F, _G, _H, _I, _J, _K, _L, _M, _N, _O, _P, _Q, _R, _S, _T, _U, _V, _W, _X, _Y, _Z, __, _$, _aa, _ba, _ca, _da, _ea, _fa, _ga, _ha, _ia, _ja, _ka, _la, _ma, _na, _oa, _pa, _qa, _ra, _sa, _ta, _ua, _va, _wa, _xa, _ya, _za, _Aa, _Ba, _Ca, _Da, _Ea, _Fa, _Ga, _Ha, _Ia, _Ja, _Ka, _La, _Ma, _Na, _Oa, _Pa, _Qa;
+  const projectId = (_c = (_b = (_a2 = readString(env, "RAG_PROJECT_ID")) != null ? _a2 : readString(env, "NEXT_PUBLIC_PROJECT_ID")) != null ? _b : base == null ? void 0 : base.projectId) != null ? _c : "__default__";
   const vectorProvider = readEnum(env, "VECTOR_DB_PROVIDER", "pinecone", VECTOR_DB_PROVIDERS);
   const llmProvider = readEnum(env, "LLM_PROVIDER", "openai", LLM_PROVIDERS);
   const embeddingProvider = readEnum(env, "EMBEDDING_PROVIDER", "openai", EMBEDDING_PROVIDERS);
   const embeddingDimensions = readNumber(env, "EMBEDDING_DIMENSIONS", 1536);
   const vectorDbOptions = {};
   if (vectorProvider === "pinecone") {
-    vectorDbOptions.apiKey = (_g = (_f = readString(env, "PINECONE_API_KEY")) != null ? _f : (_e = (_d = base == null ? void 0 : base.vectorDb) == null ? void 0 : _d.options) == null ? void 0 : _e.apiKey) != null ? _g : "";
+    vectorDbOptions.apiKey = (_g2 = (_f = readString(env, "PINECONE_API_KEY")) != null ? _f : (_e = (_d = base == null ? void 0 : base.vectorDb) == null ? void 0 : _d.options) == null ? void 0 : _e.apiKey) != null ? _g2 : "";
     vectorDbOptions.indexName = (_l = (_i = readString(env, "PINECONE_INDEX")) != null ? _i : (_h = base == null ? void 0 : base.vectorDb) == null ? void 0 : _h.indexName) != null ? _l : (_k = (_j = base == null ? void 0 : base.vectorDb) == null ? void 0 : _j.options) == null ? void 0 : _k.indexName;
   } else if (vectorProvider === "pgvector" || vectorProvider === "postgresql") {
     vectorDbOptions.connectionString = (_q = (_p = (_m = readString(env, "PGVECTOR_CONNECTION_STRING")) != null ? _m : readString(env, "POSTGRES_URL")) != null ? _p : (_o = (_n = base == null ? void 0 : base.vectorDb) == null ? void 0 : _n.options) == null ? void 0 : _o.connectionString) != null ? _q : "";
@@ -2187,17 +2191,18 @@ function getEnvConfig(env = process.env, base) {
     rest: "default",
     custom: "default"
   };
-  return __spreadValues({
+  return __spreadProps(__spreadValues({
     projectId,
+    licenseKey: (_ha = (_ga = readString(env, "RETRIVORA_LICENSE_KEY")) != null ? _ga : readString(env, "LICENSE_KEY")) != null ? _ha : base == null ? void 0 : base.licenseKey,
     vectorDb: {
       provider: vectorProvider,
-      indexName: (_ha = (_ga = readString(env, "VECTOR_DB_INDEX")) != null ? _ga : vectorDbOptions.indexName) != null ? _ha : "rag-index",
+      indexName: (_ja = (_ia = readString(env, "VECTOR_DB_INDEX")) != null ? _ia : vectorDbOptions.indexName) != null ? _ja : "rag-index",
       options: vectorDbOptions
     },
     llm: {
       provider: llmProvider,
-      model: (_ja = (_ia = readString(env, "LLM_MODEL")) != null ? _ia : DEFAULT_MODEL_BY_PROVIDER[llmProvider]) != null ? _ja : "gpt-4o",
-      apiKey: (_ka = llmApiKeyByProvider[llmProvider]) != null ? _ka : "",
+      model: (_la = (_ka = readString(env, "LLM_MODEL")) != null ? _ka : DEFAULT_MODEL_BY_PROVIDER[llmProvider]) != null ? _la : "gpt-4o",
+      apiKey: (_ma = llmApiKeyByProvider[llmProvider]) != null ? _ma : "",
       baseUrl: readString(env, "LLM_BASE_URL"),
       systemPrompt: readString(env, "LLM_SYSTEM_PROMPT"),
       maxTokens: readNumber(env, "LLM_MAX_TOKENS", 4096),
@@ -2210,7 +2215,7 @@ function getEnvConfig(env = process.env, base) {
     },
     embedding: {
       provider: embeddingProvider,
-      model: (_la = readString(env, "EMBEDDING_MODEL")) != null ? _la : "text-embedding-3-small",
+      model: (_na = readString(env, "EMBEDDING_MODEL")) != null ? _na : "text-embedding-3-small",
       apiKey: embeddingApiKeyByProvider[embeddingProvider],
       baseUrl: readString(env, "EMBEDDING_BASE_URL"),
       dimensions: embeddingDimensions,
@@ -2221,30 +2226,30 @@ function getEnvConfig(env = process.env, base) {
       }
     },
     ui: {
-      title: (_na = (_ma = readString(env, "NEXT_PUBLIC_UI_TITLE")) != null ? _ma : readString(env, "UI_TITLE")) != null ? _na : "AI Assistant",
-      subtitle: (_pa = (_oa = readString(env, "NEXT_PUBLIC_UI_SUBTITLE")) != null ? _oa : readString(env, "UI_SUBTITLE")) != null ? _pa : "Powered by RAG",
-      primaryColor: (_ra = (_qa = readString(env, "NEXT_PUBLIC_PRIMARY_COLOR")) != null ? _qa : readString(env, "UI_PRIMARY_COLOR")) != null ? _ra : "#10b981",
-      accentColor: (_ta = (_sa = readString(env, "NEXT_PUBLIC_ACCENT_COLOR")) != null ? _sa : readString(env, "UI_ACCENT_COLOR")) != null ? _ta : "#3b82f6",
-      logoUrl: (_ua = readString(env, "NEXT_PUBLIC_LOGO_URL")) != null ? _ua : readString(env, "UI_LOGO_URL"),
-      placeholder: (_wa = (_va = readString(env, "NEXT_PUBLIC_PLACEHOLDER")) != null ? _va : readString(env, "UI_PLACEHOLDER")) != null ? _wa : "Ask me anything\u2026",
-      showSources: ((_ya = (_xa = readString(env, "NEXT_PUBLIC_SHOW_SOURCES")) != null ? _xa : readString(env, "UI_SHOW_SOURCES")) != null ? _ya : "true") !== "false",
-      welcomeMessage: (_Aa = (_za = readString(env, "NEXT_PUBLIC_WELCOME_MESSAGE")) != null ? _za : readString(env, "UI_WELCOME_MESSAGE")) != null ? _Aa : "Hello! I'm your AI assistant. Ask me anything about your documents.",
-      visualStyle: (_Ca = (_Ba = readString(env, "NEXT_PUBLIC_UI_VISUAL_STYLE")) != null ? _Ba : readString(env, "UI_VISUAL_STYLE")) != null ? _Ca : "glass",
-      borderRadius: (_Ea = (_Da = readString(env, "NEXT_PUBLIC_UI_BORDER_RADIUS")) != null ? _Da : readString(env, "UI_BORDER_RADIUS")) != null ? _Ea : "xl",
-      allowUpload: ((_Ga = (_Fa = readString(env, "NEXT_PUBLIC_ALLOW_UPLOAD")) != null ? _Fa : readString(env, "UI_ALLOW_UPLOAD")) != null ? _Ga : "false") === "true"
+      title: (_pa = (_oa = readString(env, "NEXT_PUBLIC_UI_TITLE")) != null ? _oa : readString(env, "UI_TITLE")) != null ? _pa : "AI Assistant",
+      subtitle: (_ra = (_qa = readString(env, "NEXT_PUBLIC_UI_SUBTITLE")) != null ? _qa : readString(env, "UI_SUBTITLE")) != null ? _ra : "Powered by RAG",
+      primaryColor: (_ta = (_sa = readString(env, "NEXT_PUBLIC_PRIMARY_COLOR")) != null ? _sa : readString(env, "UI_PRIMARY_COLOR")) != null ? _ta : "#10b981",
+      accentColor: (_va = (_ua = readString(env, "NEXT_PUBLIC_ACCENT_COLOR")) != null ? _ua : readString(env, "UI_ACCENT_COLOR")) != null ? _va : "#3b82f6",
+      logoUrl: (_wa = readString(env, "NEXT_PUBLIC_LOGO_URL")) != null ? _wa : readString(env, "UI_LOGO_URL"),
+      placeholder: (_ya = (_xa = readString(env, "NEXT_PUBLIC_PLACEHOLDER")) != null ? _xa : readString(env, "UI_PLACEHOLDER")) != null ? _ya : "Ask me anything\u2026",
+      showSources: ((_Aa = (_za = readString(env, "NEXT_PUBLIC_SHOW_SOURCES")) != null ? _za : readString(env, "UI_SHOW_SOURCES")) != null ? _Aa : "true") !== "false",
+      welcomeMessage: (_Ca = (_Ba = readString(env, "NEXT_PUBLIC_WELCOME_MESSAGE")) != null ? _Ba : readString(env, "UI_WELCOME_MESSAGE")) != null ? _Ca : "Hello! I'm your AI assistant. Ask me anything about your documents.",
+      visualStyle: (_Ea = (_Da = readString(env, "NEXT_PUBLIC_UI_VISUAL_STYLE")) != null ? _Da : readString(env, "UI_VISUAL_STYLE")) != null ? _Ea : "glass",
+      borderRadius: (_Ga = (_Fa = readString(env, "NEXT_PUBLIC_UI_BORDER_RADIUS")) != null ? _Fa : readString(env, "UI_BORDER_RADIUS")) != null ? _Ga : "xl",
+      allowUpload: ((_Ia = (_Ha = readString(env, "NEXT_PUBLIC_ALLOW_UPLOAD")) != null ? _Ha : readString(env, "UI_ALLOW_UPLOAD")) != null ? _Ia : "false") === "true"
     },
     rag: {
       topK: readNumber(env, "RAG_TOP_K", 5),
       scoreThreshold: readNumber(env, "RAG_SCORE_THRESHOLD", 0),
       chunkSize: readNumber(env, "RAG_CHUNK_SIZE", 1e3),
       chunkOverlap: readNumber(env, "RAG_CHUNK_OVERLAP", 200),
-      filterableFields: (_Ha = readString(env, "RAG_FILTERABLE_FIELDS")) == null ? void 0 : _Ha.split(",").map((f) => f.trim()),
+      filterableFields: (_Ja = readString(env, "RAG_FILTERABLE_FIELDS")) == null ? void 0 : _Ja.split(",").map((f) => f.trim()),
       // Query pipeline toggles — read from .env.local
       useQueryTransformation: readString(env, "RAG_USE_QUERY_TRANSFORMATION") === "true",
       useReranking: readString(env, "RAG_USE_RERANKING") === "true",
       useGraphRetrieval: readString(env, "RAG_USE_GRAPH_RETRIEVAL") === "true",
-      architecture: (_Ia = readString(env, "RAG_ARCHITECTURE")) != null ? _Ia : "simple",
-      chunkingStrategy: (_Ja = readString(env, "RAG_CHUNKING_STRATEGY")) != null ? _Ja : "recursive",
+      architecture: (_Ka = readString(env, "RAG_ARCHITECTURE")) != null ? _Ka : "simple",
+      chunkingStrategy: (_La = readString(env, "RAG_CHUNKING_STRATEGY")) != null ? _La : "recursive",
       uiMapping: (() => {
         const raw = readString(env, "RAG_UI_MAPPING");
         if (!raw) return void 0;
@@ -2254,6 +2259,10 @@ function getEnvConfig(env = process.env, base) {
           return void 0;
         }
       })()
+    },
+    telemetry: {
+      enabled: readString(env, "TELEMETRY_ENABLED") === "true" || readString(env, "NEXT_PUBLIC_TELEMETRY_ENABLED") === "true" || ((_Ma = base == null ? void 0 : base.telemetry) == null ? void 0 : _Ma.enabled) || false,
+      url: (_Qa = (_Pa = (_Na = readString(env, "TELEMETRY_URL")) != null ? _Na : readString(env, "NEXT_PUBLIC_TELEMETRY_URL")) != null ? _Pa : (_Oa = base == null ? void 0 : base.telemetry) == null ? void 0 : _Oa.url) != null ? _Qa : "/api/telemetry"
     }
   }, readString(env, "GRAPH_DB_PROVIDER") ? {
     graphDb: {
@@ -2264,7 +2273,19 @@ function getEnvConfig(env = process.env, base) {
         password: readString(env, "GRAPH_DB_PASSWORD")
       }
     }
-  } : {});
+  } : {}), {
+    mcpServers: (() => {
+      var _a3;
+      const raw = (_a3 = readString(env, "MCP_SERVERS")) != null ? _a3 : readString(env, "NEXT_PUBLIC_MCP_SERVERS");
+      if (!raw) return void 0;
+      try {
+        return JSON.parse(raw);
+      } catch (e) {
+        console.warn("[serverConfig] Failed to parse MCP_SERVERS env:", e);
+        return void 0;
+      }
+    })()
+  });
 }
 
 // src/core/ConfigResolver.ts
@@ -2290,7 +2311,8 @@ var ConfigResolver = class {
         options: __spreadValues(__spreadValues({}, envConfig.embedding.options || {}), hostConfig.embedding.options || {})
       }) : envConfig.embedding,
       ui: hostConfig.ui ? mergeDefined(envConfig.ui, hostConfig.ui) : envConfig.ui,
-      rag: hostConfig.rag ? __spreadValues(__spreadValues({}, envConfig.rag), hostConfig.rag) : envConfig.rag
+      rag: hostConfig.rag ? __spreadValues(__spreadValues({}, envConfig.rag), hostConfig.rag) : envConfig.rag,
+      telemetry: hostConfig.telemetry ? __spreadValues(__spreadValues({}, envConfig.telemetry), hostConfig.telemetry) : envConfig.telemetry
     });
   }
   /**
@@ -2299,10 +2321,10 @@ var ConfigResolver = class {
    * fallback behavior.
    */
   static resolveUniversal(hostConfig, env = process.env) {
-    var _a;
+    var _a2;
     if (!hostConfig) return this.resolve(void 0, env);
     const normalized = __spreadProps(__spreadValues({}, hostConfig), {
-      vectorDb: (_a = hostConfig.vectorDb) != null ? _a : hostConfig.vectorDatabase,
+      vectorDb: (_a2 = hostConfig.vectorDb) != null ? _a2 : hostConfig.vectorDatabase,
       rag: this.mergeRetrievalWorkflow(hostConfig.rag, hostConfig.retrieval, hostConfig.workflow)
     });
     return this.resolve(normalized, env);
@@ -2322,11 +2344,11 @@ var ConfigResolver = class {
     }
   }
   static mergeRetrievalWorkflow(rag, retrieval, workflow) {
-    var _a, _b, _c, _d, _e;
+    var _a2, _b, _c, _d, _e;
     if (!rag && !retrieval && !workflow) return void 0;
     const normalized = __spreadValues({}, rag != null ? rag : {});
     if (retrieval) {
-      normalized.topK = (_a = retrieval.topK) != null ? _a : normalized.topK;
+      normalized.topK = (_a2 = retrieval.topK) != null ? _a2 : normalized.topK;
       normalized.scoreThreshold = (_b = retrieval.scoreThreshold) != null ? _b : normalized.scoreThreshold;
       normalized.useReranking = (_c = retrieval.useReranking) != null ? _c : normalized.useReranking;
       if (retrieval.strategy === "hybrid") normalized.architecture = (_d = normalized.architecture) != null ? _d : "hybrid";
@@ -2408,8 +2430,8 @@ var OpenAIProvider = class {
         const apiKey = config.apiKey;
         const modelName = config.model;
         try {
-          const OpenAI2 = await import("openai");
-          const client = new OpenAI2.default({ apiKey });
+          const OpenAI4 = await import("openai");
+          const client = new OpenAI4.default({ apiKey });
           const models = await client.models.list();
           const hasModel = models.data.some((m) => m.id === modelName);
           return {
@@ -2430,8 +2452,8 @@ var OpenAIProvider = class {
     };
   }
   async chat(messages, context, options) {
-    var _a, _b, _c, _d, _e, _f, _g, _h, _i;
-    const basePrompt = (_b = (_a = options == null ? void 0 : options.systemPrompt) != null ? _a : this.llmConfig.systemPrompt) != null ? _b : "You are a helpful assistant. Answer questions based on the provided context.";
+    var _a2, _b, _c, _d, _e, _f, _g2, _h, _i;
+    const basePrompt = (_b = (_a2 = options == null ? void 0 : options.systemPrompt) != null ? _a2 : this.llmConfig.systemPrompt) != null ? _b : "You are a helpful assistant. Answer questions based on the provided context.";
     const systemMessage = {
       role: "system",
       content: buildSystemContent(basePrompt, context)
@@ -2450,12 +2472,12 @@ var OpenAIProvider = class {
       temperature: (_f = (_e = options == null ? void 0 : options.temperature) != null ? _e : this.llmConfig.temperature) != null ? _f : 0.7,
       stop: options == null ? void 0 : options.stop
     });
-    return (_i = (_h = (_g = completion.choices[0]) == null ? void 0 : _g.message) == null ? void 0 : _h.content) != null ? _i : "";
+    return (_i = (_h = (_g2 = completion.choices[0]) == null ? void 0 : _g2.message) == null ? void 0 : _h.content) != null ? _i : "";
   }
   chatStream(messages, context, options) {
     return __asyncGenerator(this, null, function* () {
-      var _a, _b, _c, _d, _e, _f, _g, _h;
-      const basePrompt = (_b = (_a = options == null ? void 0 : options.systemPrompt) != null ? _a : this.llmConfig.systemPrompt) != null ? _b : "You are a helpful assistant. Answer questions based on the provided context.";
+      var _a2, _b, _c, _d, _e, _f, _g2, _h;
+      const basePrompt = (_b = (_a2 = options == null ? void 0 : options.systemPrompt) != null ? _a2 : this.llmConfig.systemPrompt) != null ? _b : "You are a helpful assistant. Answer questions based on the provided context.";
       const systemMessage = {
         role: "system",
         content: buildSystemContent(basePrompt, context)
@@ -2481,7 +2503,7 @@ var OpenAIProvider = class {
       try {
         for (var iter = __forAwait(stream), more, temp, error; more = !(temp = yield new __await(iter.next())).done; more = false) {
           const chunk = temp.value;
-          const content = ((_h = (_g = chunk.choices[0]) == null ? void 0 : _g.delta) == null ? void 0 : _h.content) || "";
+          const content = ((_h = (_g2 = chunk.choices[0]) == null ? void 0 : _g2.delta) == null ? void 0 : _h.content) || "";
           if (content) yield content;
         }
       } catch (temp) {
@@ -2501,8 +2523,8 @@ var OpenAIProvider = class {
     return results[0];
   }
   async batchEmbed(texts, options) {
-    var _a, _b, _c, _d, _e;
-    const model = (_c = (_b = options == null ? void 0 : options.model) != null ? _b : (_a = this.embeddingConfig) == null ? void 0 : _a.model) != null ? _c : "text-embedding-3-small";
+    var _a2, _b, _c, _d, _e;
+    const model = (_c = (_b = options == null ? void 0 : options.model) != null ? _b : (_a2 = this.embeddingConfig) == null ? void 0 : _a2.model) != null ? _c : "text-embedding-3-small";
     const apiKey = (_e = (_d = this.embeddingConfig) == null ? void 0 : _d.apiKey) != null ? _e : this.llmConfig.apiKey;
     const client = apiKey !== this.llmConfig.apiKey ? new OpenAI({ apiKey }) : this.client;
     const response = await client.embeddings.create({ model, input: texts });
@@ -2579,8 +2601,8 @@ var AnthropicProvider = class {
     };
   }
   async chat(messages, context, options) {
-    var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j, _k;
-    const basePrompt = (_b = (_a = options == null ? void 0 : options.systemPrompt) != null ? _a : this.llmConfig.systemPrompt) != null ? _b : "You are a helpful assistant. Use the context below to answer the user's question accurately.";
+    var _a2, _b, _c, _d, _e, _f, _g2, _h, _i, _j, _k;
+    const basePrompt = (_b = (_a2 = options == null ? void 0 : options.systemPrompt) != null ? _a2 : this.llmConfig.systemPrompt) != null ? _b : "You are a helpful assistant. Use the context below to answer the user's question accurately.";
     const system = buildSystemContent(basePrompt, context);
     const anthropicMessages = messages.map((m) => ({
       role: m.role === "assistant" ? "assistant" : "user",
@@ -2591,7 +2613,7 @@ var AnthropicProvider = class {
     const extraParams = {};
     if (isThinkingEnabled && isClaude37) {
       extraParams.betas = ["interleaved-thinking-2025-05-14"];
-      const maxTokens = (_g = (_f = options == null ? void 0 : options.maxTokens) != null ? _f : this.llmConfig.maxTokens) != null ? _g : 4096;
+      const maxTokens = (_g2 = (_f = options == null ? void 0 : options.maxTokens) != null ? _f : this.llmConfig.maxTokens) != null ? _g2 : 4096;
       const budget = Math.min(
         typeof ((_h = this.llmConfig.options) == null ? void 0 : _h.thinkingBudget) === "number" ? (_i = this.llmConfig.options) == null ? void 0 : _i.thinkingBudget : 2048,
         maxTokens - 1
@@ -2618,8 +2640,8 @@ var AnthropicProvider = class {
   }
   chatStream(messages, context, options) {
     return __asyncGenerator(this, null, function* () {
-      var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j, _k;
-      const basePrompt = (_b = (_a = options == null ? void 0 : options.systemPrompt) != null ? _a : this.llmConfig.systemPrompt) != null ? _b : "You are a helpful assistant. Use the context below to answer the user's question accurately.";
+      var _a2, _b, _c, _d, _e, _f, _g2, _h, _i, _j, _k;
+      const basePrompt = (_b = (_a2 = options == null ? void 0 : options.systemPrompt) != null ? _a2 : this.llmConfig.systemPrompt) != null ? _b : "You are a helpful assistant. Use the context below to answer the user's question accurately.";
       const system = buildSystemContent(basePrompt, context);
       const anthropicMessages = messages.map((m) => ({
         role: m.role === "assistant" ? "assistant" : "user",
@@ -2630,7 +2652,7 @@ var AnthropicProvider = class {
       const extraParams = {};
       if (isThinkingEnabled && isClaude37) {
         extraParams.betas = ["interleaved-thinking-2025-05-14"];
-        const maxTokens = (_g = (_f = options == null ? void 0 : options.maxTokens) != null ? _f : this.llmConfig.maxTokens) != null ? _g : 4096;
+        const maxTokens = (_g2 = (_f = options == null ? void 0 : options.maxTokens) != null ? _f : this.llmConfig.maxTokens) != null ? _g2 : 4096;
         const budget = Math.min(
           typeof ((_h = this.llmConfig.options) == null ? void 0 : _h.thinkingBudget) === "number" ? (_i = this.llmConfig.options) == null ? void 0 : _i.thinkingBudget : 2048,
           maxTokens - 1
@@ -2703,8 +2725,8 @@ var AnthropicProvider = class {
 import axios from "axios";
 var OllamaProvider = class {
   constructor(llmConfig, embeddingConfig) {
-    var _a, _b;
-    const baseURL = (_a = llmConfig.baseUrl) != null ? _a : "http://localhost:11434";
+    var _a2, _b;
+    const baseURL = (_a2 = llmConfig.baseUrl) != null ? _a2 : "http://localhost:11434";
     const timeout = Number((_b = llmConfig.options) == null ? void 0 : _b.timeout) || 3e5;
     this.http = axios.create({ baseURL, timeout });
     this.llmConfig = llmConfig;
@@ -2762,8 +2784,8 @@ var OllamaProvider = class {
     };
   }
   async chat(messages, context, options) {
-    var _a, _b, _c, _d, _e, _f;
-    const basePrompt = (_b = (_a = options == null ? void 0 : options.systemPrompt) != null ? _a : this.llmConfig.systemPrompt) != null ? _b : "You are a helpful assistant. Use the provided context to answer the user's question.";
+    var _a2, _b, _c, _d, _e, _f;
+    const basePrompt = (_b = (_a2 = options == null ? void 0 : options.systemPrompt) != null ? _a2 : this.llmConfig.systemPrompt) != null ? _b : "You are a helpful assistant. Use the provided context to answer the user's question.";
     const system = buildSystemContent(basePrompt, context);
     const { data } = await this.http.post("/api/chat", {
       model: this.llmConfig.model,
@@ -2781,8 +2803,8 @@ var OllamaProvider = class {
   }
   chatStream(messages, context, options) {
     return __asyncGenerator(this, null, function* () {
-      var _a, _b, _c, _d, _e, _f, _g, _h;
-      const basePrompt = (_b = (_a = options == null ? void 0 : options.systemPrompt) != null ? _a : this.llmConfig.systemPrompt) != null ? _b : "You are a helpful assistant. Use the provided context to answer the user's question.";
+      var _a2, _b, _c, _d, _e, _f, _g2, _h;
+      const basePrompt = (_b = (_a2 = options == null ? void 0 : options.systemPrompt) != null ? _a2 : this.llmConfig.systemPrompt) != null ? _b : "You are a helpful assistant. Use the provided context to answer the user's question.";
       const system = buildSystemContent(basePrompt, context);
       const response = yield new __await(this.http.post("/api/chat", {
         model: this.llmConfig.model,
@@ -2811,7 +2833,7 @@ var OllamaProvider = class {
             if (!line.trim()) continue;
             try {
               const json = JSON.parse(line);
-              if ((_g = json.message) == null ? void 0 : _g.content) {
+              if ((_g2 = json.message) == null ? void 0 : _g2.content) {
                 yield json.message.content;
               }
               if (json.done) return;
@@ -2840,10 +2862,10 @@ var OllamaProvider = class {
     });
   }
   async embed(text, options) {
-    var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j, _k;
-    const model = (_c = (_b = options == null ? void 0 : options.model) != null ? _b : (_a = this.embeddingConfig) == null ? void 0 : _a.model) != null ? _c : "nomic-embed-text";
+    var _a2, _b, _c, _d, _e, _f, _g2, _h, _i, _j, _k;
+    const model = (_c = (_b = options == null ? void 0 : options.model) != null ? _b : (_a2 = this.embeddingConfig) == null ? void 0 : _a2.model) != null ? _c : "nomic-embed-text";
     const baseURL = (_f = (_e = (_d = this.embeddingConfig) == null ? void 0 : _d.baseUrl) != null ? _e : this.llmConfig.baseUrl) != null ? _f : "http://localhost:11434";
-    const client = baseURL !== ((_g = this.llmConfig.baseUrl) != null ? _g : "http://localhost:11434") ? axios.create({ baseURL, timeout: 6e4 }) : this.http;
+    const client = baseURL !== ((_g2 = this.llmConfig.baseUrl) != null ? _g2 : "http://localhost:11434") ? axios.create({ baseURL, timeout: 6e4 }) : this.http;
     let prompt = text;
     const queryPrefix = (_i = (_h = this.embeddingConfig) == null ? void 0 : _h.queryPrefix) != null ? _i : model.includes("nomic") ? "search_query: " : "";
     const docPrefix = (_k = (_j = this.embeddingConfig) == null ? void 0 : _j.documentPrefix) != null ? _k : model.includes("nomic") ? "search_document: " : "";
@@ -2942,9 +2964,9 @@ var GeminiProvider = class {
   static getHealthChecker() {
     return {
       async check(config) {
-        var _a, _b;
+        var _a2, _b;
         const timestamp = Date.now();
-        const apiKey = (_b = (_a = config.apiKey) != null ? _a : process.env.GOOGLE_GENAI_API_KEY) != null ? _b : "";
+        const apiKey = (_b = (_a2 = config.apiKey) != null ? _a2 : process.env.GOOGLE_GENAI_API_KEY) != null ? _b : "";
         const modelName = sanitizeModel(config.model);
         try {
           const { GoogleGenerativeAI: GoogleGenerativeAI2 } = await import("@google/generative-ai");
@@ -2974,16 +2996,16 @@ var GeminiProvider = class {
   /** Resolve the embedding client — uses a separate client when the embedding
    *  API key differs from the LLM API key. */
   get embeddingClient() {
-    var _a;
-    if (((_a = this.embeddingConfig) == null ? void 0 : _a.apiKey) && this.embeddingConfig.apiKey !== this.llmConfig.apiKey) {
+    var _a2;
+    if (((_a2 = this.embeddingConfig) == null ? void 0 : _a2.apiKey) && this.embeddingConfig.apiKey !== this.llmConfig.apiKey) {
       return buildClient(this.embeddingConfig.apiKey);
     }
     return this.client;
   }
   /** Resolve the embedding model to use, in order of specificity. */
   resolveEmbeddingModel(optionsModel) {
-    var _a, _b;
-    return sanitizeModel((_b = optionsModel != null ? optionsModel : (_a = this.embeddingConfig) == null ? void 0 : _a.model) != null ? _b : DEFAULT_EMBEDDING_MODEL);
+    var _a2, _b;
+    return sanitizeModel((_b = optionsModel != null ? optionsModel : (_a2 = this.embeddingConfig) == null ? void 0 : _a2.model) != null ? _b : DEFAULT_EMBEDDING_MODEL);
   }
   /**
    * Convert ChatMessage[] to the Gemini contents format.
@@ -3003,8 +3025,8 @@ var GeminiProvider = class {
   // ILLMProvider — chat
   // -------------------------------------------------------------------------
   async chat(messages, context, options) {
-    var _a, _b, _c, _d, _e, _f;
-    const basePrompt = (_b = (_a = options == null ? void 0 : options.systemPrompt) != null ? _a : this.llmConfig.systemPrompt) != null ? _b : "You are a helpful assistant. Answer questions based on the provided context.";
+    var _a2, _b, _c, _d, _e, _f;
+    const basePrompt = (_b = (_a2 = options == null ? void 0 : options.systemPrompt) != null ? _a2 : this.llmConfig.systemPrompt) != null ? _b : "You are a helpful assistant. Answer questions based on the provided context.";
     const model = this.client.getGenerativeModel({
       model: this.llmConfig.model,
       systemInstruction: buildSystemContent(basePrompt, context)
@@ -3021,8 +3043,8 @@ var GeminiProvider = class {
   }
   chatStream(messages, context, options) {
     return __asyncGenerator(this, null, function* () {
-      var _a, _b, _c, _d, _e, _f;
-      const basePrompt = (_b = (_a = options == null ? void 0 : options.systemPrompt) != null ? _a : this.llmConfig.systemPrompt) != null ? _b : "You are a helpful assistant. Answer questions based on the provided context.";
+      var _a2, _b, _c, _d, _e, _f;
+      const basePrompt = (_b = (_a2 = options == null ? void 0 : options.systemPrompt) != null ? _a2 : this.llmConfig.systemPrompt) != null ? _b : "You are a helpful assistant. Answer questions based on the provided context.";
       const model = this.client.getGenerativeModel({
         model: this.llmConfig.model,
         systemInstruction: buildSystemContent(basePrompt, context)
@@ -3057,11 +3079,11 @@ var GeminiProvider = class {
   // ILLMProvider — embeddings
   // -------------------------------------------------------------------------
   async embed(text, options) {
-    var _a, _b;
+    var _a2, _b;
     const content = applyPrefix(
       text,
       options == null ? void 0 : options.taskType,
-      (_a = this.embeddingConfig) == null ? void 0 : _a.queryPrefix,
+      (_a2 = this.embeddingConfig) == null ? void 0 : _a2.queryPrefix,
       (_b = this.embeddingConfig) == null ? void 0 : _b.documentPrefix
     );
     const modelName = this.resolveEmbeddingModel(options == null ? void 0 : options.model);
@@ -3078,7 +3100,7 @@ var GeminiProvider = class {
     const modelName = this.resolveEmbeddingModel(options == null ? void 0 : options.model);
     const model = this.embeddingClient.getGenerativeModel({ model: modelName });
     const requests = texts.map((text) => {
-      var _a, _b;
+      var _a2, _b;
       return {
         content: {
           role: "user",
@@ -3086,7 +3108,7 @@ var GeminiProvider = class {
             text: applyPrefix(
               text,
               options == null ? void 0 : options.taskType,
-              (_a = this.embeddingConfig) == null ? void 0 : _a.queryPrefix,
+              (_a2 = this.embeddingConfig) == null ? void 0 : _a2.queryPrefix,
               (_b = this.embeddingConfig) == null ? void 0 : _b.documentPrefix
             )
           }]
@@ -3103,8 +3125,8 @@ var GeminiProvider = class {
       throw err;
     });
     return response.embeddings.map((e) => {
-      var _a;
-      return (_a = e.values) != null ? _a : [];
+      var _a2;
+      return (_a2 = e.values) != null ? _a2 : [];
     });
   }
   // -------------------------------------------------------------------------
@@ -3124,6 +3146,325 @@ var GeminiProvider = class {
   }
 };
 
+// src/llm/providers/GroqProvider.ts
+import OpenAI2 from "openai";
+var GroqProvider = class {
+  constructor(llmConfig, embeddingConfig) {
+    void embeddingConfig;
+    const apiKey = llmConfig.apiKey || process.env.GROQ_API_KEY;
+    if (!apiKey) throw new Error("[GroqProvider] llmConfig.apiKey or GROQ_API_KEY environment variable is required");
+    this.client = new OpenAI2({
+      apiKey,
+      baseURL: llmConfig.baseUrl || "https://api.groq.com/openai/v1"
+    });
+    this.llmConfig = llmConfig;
+  }
+  static getValidator() {
+    return {
+      validate(config) {
+        const errors = [];
+        if (!config.apiKey && !process.env.GROQ_API_KEY) {
+          errors.push({
+            field: "llm.apiKey",
+            message: "Groq API key is required",
+            suggestion: "Set GROQ_API_KEY environment variable",
+            severity: "error"
+          });
+        }
+        if (!config.model) {
+          errors.push({
+            field: "llm.model",
+            message: "Groq model name is required",
+            suggestion: 'e.g., "llama-3.3-70b-versatile"',
+            severity: "error"
+          });
+        }
+        return errors;
+      }
+    };
+  }
+  static getHealthChecker() {
+    return {
+      async check(config) {
+        const timestamp = Date.now();
+        const apiKey = config.apiKey || process.env.GROQ_API_KEY || "";
+        const modelName = config.model;
+        try {
+          const OpenAI4 = await import("openai");
+          const client = new OpenAI4.default({
+            apiKey,
+            baseURL: config.baseUrl || "https://api.groq.com/openai/v1"
+          });
+          const models = await client.models.list();
+          const hasModel = models.data.some((m) => m.id === modelName);
+          return {
+            healthy: true,
+            provider: "groq",
+            capabilities: { model: modelName, available: hasModel, totalModels: models.data.length },
+            timestamp
+          };
+        } catch (error) {
+          return {
+            healthy: false,
+            provider: "groq",
+            error: `Connection failed: ${error instanceof Error ? error.message : String(error)}`,
+            timestamp
+          };
+        }
+      }
+    };
+  }
+  async chat(messages, context, options) {
+    var _a2, _b, _c, _d, _e, _f, _g2, _h, _i;
+    const basePrompt = (_b = (_a2 = options == null ? void 0 : options.systemPrompt) != null ? _a2 : this.llmConfig.systemPrompt) != null ? _b : "You are a helpful assistant. Answer questions based on the provided context.";
+    const systemMessage = {
+      role: "system",
+      content: buildSystemContent(basePrompt, context)
+    };
+    const formattedMessages = [
+      systemMessage,
+      ...messages.map((m) => ({
+        role: m.role,
+        content: m.content
+      }))
+    ];
+    const completion = await this.client.chat.completions.create({
+      model: this.llmConfig.model,
+      messages: formattedMessages,
+      max_tokens: (_d = (_c = options == null ? void 0 : options.maxTokens) != null ? _c : this.llmConfig.maxTokens) != null ? _d : 1024,
+      temperature: (_f = (_e = options == null ? void 0 : options.temperature) != null ? _e : this.llmConfig.temperature) != null ? _f : 0.7,
+      stop: options == null ? void 0 : options.stop
+    });
+    return (_i = (_h = (_g2 = completion.choices[0]) == null ? void 0 : _g2.message) == null ? void 0 : _h.content) != null ? _i : "";
+  }
+  chatStream(messages, context, options) {
+    return __asyncGenerator(this, null, function* () {
+      var _a2, _b, _c, _d, _e, _f, _g2, _h;
+      const basePrompt = (_b = (_a2 = options == null ? void 0 : options.systemPrompt) != null ? _a2 : this.llmConfig.systemPrompt) != null ? _b : "You are a helpful assistant. Answer questions based on the provided context.";
+      const systemMessage = {
+        role: "system",
+        content: buildSystemContent(basePrompt, context)
+      };
+      const formattedMessages = [
+        systemMessage,
+        ...messages.map((m) => ({
+          role: m.role,
+          content: m.content
+        }))
+      ];
+      const stream = yield new __await(this.client.chat.completions.create({
+        model: this.llmConfig.model,
+        messages: formattedMessages,
+        max_tokens: (_d = (_c = options == null ? void 0 : options.maxTokens) != null ? _c : this.llmConfig.maxTokens) != null ? _d : 1024,
+        temperature: (_f = (_e = options == null ? void 0 : options.temperature) != null ? _e : this.llmConfig.temperature) != null ? _f : 0.7,
+        stop: options == null ? void 0 : options.stop,
+        stream: true
+      }));
+      if (!stream) {
+        throw new Error("[GroqProvider] completions.create stream is undefined");
+      }
+      try {
+        for (var iter = __forAwait(stream), more, temp, error; more = !(temp = yield new __await(iter.next())).done; more = false) {
+          const chunk = temp.value;
+          const content = ((_h = (_g2 = chunk.choices[0]) == null ? void 0 : _g2.delta) == null ? void 0 : _h.content) || "";
+          if (content) yield content;
+        }
+      } catch (temp) {
+        error = [temp];
+      } finally {
+        try {
+          more && (temp = iter.return) && (yield new __await(temp.call(iter)));
+        } finally {
+          if (error)
+            throw error[0];
+        }
+      }
+    });
+  }
+  async embed(text, options) {
+    void text;
+    void options;
+    throw new Error('[GroqProvider] Groq does not provide a native embedding API. Please configure "openai" or "gemini" for embeddings in RagConfig.');
+  }
+  async batchEmbed(texts, options) {
+    void texts;
+    void options;
+    throw new Error('[GroqProvider] Groq does not provide a native embedding API. Please configure "openai" or "gemini" for embeddings in RagConfig.');
+  }
+  async ping() {
+    try {
+      await this.client.models.list();
+      return true;
+    } catch (err) {
+      console.error("[GroqProvider] Ping failed:", err);
+      return false;
+    }
+  }
+};
+
+// src/llm/providers/QwenProvider.ts
+import OpenAI3 from "openai";
+var QwenProvider = class {
+  constructor(llmConfig, embeddingConfig) {
+    const apiKey = llmConfig.apiKey || process.env.DASHSCOPE_API_KEY || process.env.QWEN_API_KEY;
+    if (!apiKey) throw new Error("[QwenProvider] llmConfig.apiKey, DASHSCOPE_API_KEY, or QWEN_API_KEY environment variable is required");
+    this.client = new OpenAI3({
+      apiKey,
+      baseURL: llmConfig.baseUrl || "https://dashscope.aliyuncs.com/compatible-mode/v1"
+    });
+    this.llmConfig = llmConfig;
+    this.embeddingConfig = embeddingConfig;
+  }
+  static getValidator() {
+    return {
+      validate(config) {
+        const errors = [];
+        const isEmbedding = config.provider === "qwen" && "dimensions" in config;
+        const prefix = isEmbedding ? "embedding" : "llm";
+        if (!config.apiKey && !process.env.DASHSCOPE_API_KEY && !process.env.QWEN_API_KEY) {
+          errors.push({
+            field: `${prefix}.apiKey`,
+            message: "Qwen API key is required",
+            suggestion: "Set DASHSCOPE_API_KEY or QWEN_API_KEY environment variable",
+            severity: "error"
+          });
+        }
+        if (!config.model) {
+          errors.push({
+            field: `${prefix}.model`,
+            message: "Qwen model name is required",
+            suggestion: isEmbedding ? 'e.g., "text-embedding-v3"' : 'e.g., "qwen-plus" or "qwen-max"',
+            severity: "error"
+          });
+        }
+        return errors;
+      }
+    };
+  }
+  static getHealthChecker() {
+    return {
+      async check(config) {
+        const timestamp = Date.now();
+        const apiKey = config.apiKey || process.env.DASHSCOPE_API_KEY || process.env.QWEN_API_KEY || "";
+        const modelName = config.model;
+        try {
+          const OpenAI4 = await import("openai");
+          const client = new OpenAI4.default({
+            apiKey,
+            baseURL: config.baseUrl || "https://dashscope.aliyuncs.com/compatible-mode/v1"
+          });
+          const models = await client.models.list();
+          const hasModel = models.data.some((m) => m.id === modelName);
+          return {
+            healthy: true,
+            provider: "qwen",
+            capabilities: { model: modelName, available: hasModel, totalModels: models.data.length },
+            timestamp
+          };
+        } catch (error) {
+          return {
+            healthy: false,
+            provider: "qwen",
+            error: `Connection failed: ${error instanceof Error ? error.message : String(error)}`,
+            timestamp
+          };
+        }
+      }
+    };
+  }
+  async chat(messages, context, options) {
+    var _a2, _b, _c, _d, _e, _f, _g2, _h, _i;
+    const basePrompt = (_b = (_a2 = options == null ? void 0 : options.systemPrompt) != null ? _a2 : this.llmConfig.systemPrompt) != null ? _b : "You are a helpful assistant. Answer questions based on the provided context.";
+    const systemMessage = {
+      role: "system",
+      content: buildSystemContent(basePrompt, context)
+    };
+    const formattedMessages = [
+      systemMessage,
+      ...messages.map((m) => ({
+        role: m.role,
+        content: m.content
+      }))
+    ];
+    const completion = await this.client.chat.completions.create({
+      model: this.llmConfig.model,
+      messages: formattedMessages,
+      max_tokens: (_d = (_c = options == null ? void 0 : options.maxTokens) != null ? _c : this.llmConfig.maxTokens) != null ? _d : 1024,
+      temperature: (_f = (_e = options == null ? void 0 : options.temperature) != null ? _e : this.llmConfig.temperature) != null ? _f : 0.7,
+      stop: options == null ? void 0 : options.stop
+    });
+    return (_i = (_h = (_g2 = completion.choices[0]) == null ? void 0 : _g2.message) == null ? void 0 : _h.content) != null ? _i : "";
+  }
+  chatStream(messages, context, options) {
+    return __asyncGenerator(this, null, function* () {
+      var _a2, _b, _c, _d, _e, _f, _g2, _h;
+      const basePrompt = (_b = (_a2 = options == null ? void 0 : options.systemPrompt) != null ? _a2 : this.llmConfig.systemPrompt) != null ? _b : "You are a helpful assistant. Answer questions based on the provided context.";
+      const systemMessage = {
+        role: "system",
+        content: buildSystemContent(basePrompt, context)
+      };
+      const formattedMessages = [
+        systemMessage,
+        ...messages.map((m) => ({
+          role: m.role,
+          content: m.content
+        }))
+      ];
+      const stream = yield new __await(this.client.chat.completions.create({
+        model: this.llmConfig.model,
+        messages: formattedMessages,
+        max_tokens: (_d = (_c = options == null ? void 0 : options.maxTokens) != null ? _c : this.llmConfig.maxTokens) != null ? _d : 1024,
+        temperature: (_f = (_e = options == null ? void 0 : options.temperature) != null ? _e : this.llmConfig.temperature) != null ? _f : 0.7,
+        stop: options == null ? void 0 : options.stop,
+        stream: true
+      }));
+      if (!stream) {
+        throw new Error("[QwenProvider] completions.create stream is undefined");
+      }
+      try {
+        for (var iter = __forAwait(stream), more, temp, error; more = !(temp = yield new __await(iter.next())).done; more = false) {
+          const chunk = temp.value;
+          const content = ((_h = (_g2 = chunk.choices[0]) == null ? void 0 : _g2.delta) == null ? void 0 : _h.content) || "";
+          if (content) yield content;
+        }
+      } catch (temp) {
+        error = [temp];
+      } finally {
+        try {
+          more && (temp = iter.return) && (yield new __await(temp.call(iter)));
+        } finally {
+          if (error)
+            throw error[0];
+        }
+      }
+    });
+  }
+  async embed(text, options) {
+    const results = await this.batchEmbed([text], options);
+    return results[0];
+  }
+  async batchEmbed(texts, options) {
+    var _a2, _b, _c, _d, _e, _f;
+    const model = (_c = (_b = options == null ? void 0 : options.model) != null ? _b : (_a2 = this.embeddingConfig) == null ? void 0 : _a2.model) != null ? _c : "text-embedding-v1";
+    const apiKey = (_e = (_d = this.embeddingConfig) == null ? void 0 : _d.apiKey) != null ? _e : this.llmConfig.apiKey;
+    const client = apiKey !== this.llmConfig.apiKey ? new OpenAI3({
+      apiKey,
+      baseURL: ((_f = this.embeddingConfig) == null ? void 0 : _f.baseUrl) || "https://dashscope.aliyuncs.com/compatible-mode/v1"
+    }) : this.client;
+    const response = await client.embeddings.create({ model, input: texts });
+    return response.data.map((d) => d.embedding);
+  }
+  async ping() {
+    try {
+      await this.client.models.list();
+      return true;
+    } catch (err) {
+      console.error("[QwenProvider] Ping failed:", err);
+      return false;
+    }
+  }
+};
+
 // src/llm/providers/UniversalLLMAdapter.ts
 init_templateUtils();
 import axios2 from "axios";
@@ -3207,10 +3548,10 @@ var VECTOR_PROFILES = {
 // src/llm/providers/UniversalLLMAdapter.ts
 var UniversalLLMAdapter = class {
   constructor(config) {
-    var _a, _b, _c, _d, _e, _f, _g, _h;
+    var _a2, _b, _c, _d, _e, _f, _g2, _h;
     this.model = config.model;
     const llmConfig = config;
-    const options = (_a = llmConfig.options) != null ? _a : {};
+    const options = (_a2 = llmConfig.options) != null ? _a2 : {};
     let profile = {};
     if (typeof options.profile === "string") {
       profile = LLM_PROFILES[options.profile] || {};
@@ -3222,7 +3563,7 @@ var UniversalLLMAdapter = class {
     this.maxTokens = (_d = llmConfig.maxTokens) != null ? _d : 1024;
     this.temperature = (_e = llmConfig.temperature) != null ? _e : 0;
     this.apiKey = config.apiKey;
-    this.baseUrl = (_g = (_f = llmConfig.baseUrl) != null ? _f : this.opts.baseUrl) != null ? _g : "";
+    this.baseUrl = (_g2 = (_f = llmConfig.baseUrl) != null ? _f : this.opts.baseUrl) != null ? _g2 : "";
     if (!this.baseUrl) {
       throw new Error("[UniversalLLMAdapter] baseUrl is required in config or config.options");
     }
@@ -3236,8 +3577,8 @@ var UniversalLLMAdapter = class {
     });
   }
   async chat(messages, context) {
-    var _a, _b;
-    const path = (_a = this.opts.chatPath) != null ? _a : "/chat/completions";
+    var _a2, _b;
+    const path2 = (_a2 = this.opts.chatPath) != null ? _a2 : "/chat/completions";
     const formattedMessages = [
       { role: "system", content: `${this.systemPrompt}
 
@@ -3261,7 +3602,7 @@ ${context != null ? context : "None"}` },
         temperature: this.temperature
       };
     }
-    const { data } = await this.http.post(path, payload);
+    const { data } = await this.http.post(path2, payload);
     const extractPath = (_b = this.opts.responseExtractPath) != null ? _b : "choices[0].message.content";
     const result = resolvePath(data, extractPath);
     if (result === void 0) {
@@ -3276,9 +3617,9 @@ ${context != null ? context : "None"}` },
    */
   chatStream(messages, context) {
     return __asyncGenerator(this, null, function* () {
-      var _a, _b, _c;
-      const path = (_a = this.opts.chatPath) != null ? _a : "/chat/completions";
-      const url = `${this.baseUrl.replace(/\/$/, "")}${path}`;
+      var _a2, _b, _c;
+      const path2 = (_a2 = this.opts.chatPath) != null ? _a2 : "/chat/completions";
+      const url = `${this.baseUrl.replace(/\/$/, "")}${path2}`;
       const extractPath = ((_b = this.opts.responseExtractPath) != null ? _b : "choices[0].message.content").replace("message.content", "delta.content");
       const formattedMessages = [
         { role: "system", content: `${this.systemPrompt}
@@ -3356,8 +3697,8 @@ ${context != null ? context : "None"}` },
     });
   }
   async embed(text) {
-    var _a, _b;
-    const path = (_a = this.opts.embedPath) != null ? _a : "/embeddings";
+    var _a2, _b;
+    const path2 = (_a2 = this.opts.embedPath) != null ? _a2 : "/embeddings";
     let payload;
     if (this.opts.embedPayloadTemplate) {
       payload = buildPayload(this.opts.embedPayloadTemplate, {
@@ -3370,7 +3711,7 @@ ${context != null ? context : "None"}` },
         input: text
       };
     }
-    const { data } = await this.http.post(path, payload);
+    const { data } = await this.http.post(path2, payload);
     const extractPath = (_b = this.opts.embedExtractPath) != null ? _b : "data[0].embedding";
     const vector = resolvePath(data, extractPath);
     if (!Array.isArray(vector)) {
@@ -3438,13 +3779,13 @@ var AuthenticationException = class extends RetrivoraError {
   }
 };
 function wrapError(err, defaultCode, defaultMessage) {
-  var _a;
+  var _a2;
   if (err instanceof RetrivoraError) {
     return err;
   }
   const error = err;
   const message = (error == null ? void 0 : error.message) || defaultMessage || String(err);
-  const status = (error == null ? void 0 : error.status) || (error == null ? void 0 : error.statusCode) || ((_a = error == null ? void 0 : error.response) == null ? void 0 : _a.status);
+  const status = (error == null ? void 0 : error.status) || (error == null ? void 0 : error.statusCode) || ((_a2 = error == null ? void 0 : error.response) == null ? void 0 : _a2.status);
   const code = error == null ? void 0 : error.code;
   if (status === 429 || /rate[- ]?limit/i.test(message) || code === "RATE_LIMIT_EXCEEDED") {
     return new RateLimitException(message, err);
@@ -3508,6 +3849,8 @@ var LLMFactory = class _LLMFactory {
       "anthropic",
       "ollama",
       "gemini",
+      "groq",
+      "qwen",
       "rest",
       "universal_rest",
       "custom",
@@ -3515,7 +3858,7 @@ var LLMFactory = class _LLMFactory {
     ];
   }
   static create(llmConfig, embeddingConfig) {
-    var _a, _b, _c;
+    var _a2, _b, _c;
     switch (llmConfig.provider) {
       case "openai":
         return new OpenAIProvider(llmConfig, embeddingConfig);
@@ -3525,12 +3868,16 @@ var LLMFactory = class _LLMFactory {
         return new OllamaProvider(llmConfig, embeddingConfig);
       case "gemini":
         return new GeminiProvider(llmConfig, embeddingConfig);
+      case "groq":
+        return new GroqProvider(llmConfig, embeddingConfig);
+      case "qwen":
+        return new QwenProvider(llmConfig, embeddingConfig);
       case "rest":
       case "universal_rest":
       case "custom":
         return new UniversalLLMAdapter(llmConfig);
       default: {
-        const providerName = String((_a = llmConfig.provider) != null ? _a : "").toLowerCase();
+        const providerName = String((_a2 = llmConfig.provider) != null ? _a2 : "").toLowerCase();
         const customFactory = customProviders.get(providerName);
         if (customFactory) {
           return customFactory(llmConfig);
@@ -3567,6 +3914,10 @@ var LLMFactory = class _LLMFactory {
         return OllamaProvider;
       case "gemini":
         return GeminiProvider;
+      case "groq":
+        return GroqProvider;
+      case "qwen":
+        return QwenProvider;
       case "rest":
       case "universal_rest":
       case "custom":
@@ -3628,7 +3979,7 @@ var ProviderRegistry = class {
     return null;
   }
   static async loadVectorProviderClass(provider) {
-    var _a;
+    var _a2;
     if (this.vectorProviders[provider]) return this.vectorProviders[provider];
     switch (provider) {
       case "pinecone": {
@@ -3637,7 +3988,7 @@ var ProviderRegistry = class {
       }
       case "pgvector":
       case "postgresql": {
-        const postgresMode = ((_a = process.env.POSTGRES_MODE) != null ? _a : "multi").toLowerCase();
+        const postgresMode = ((_a2 = process.env.POSTGRES_MODE) != null ? _a2 : "multi").toLowerCase();
         if (postgresMode === "single") {
           const { PostgreSQLProvider: PostgreSQLProvider2 } = await Promise.resolve().then(() => (init_PostgreSQLProvider(), PostgreSQLProvider_exports));
           return PostgreSQLProvider2;
@@ -3857,6 +4208,125 @@ var ConfigValidator = class {
   }
 };
 
+// src/core/LicenseVerifier.ts
+import * as crypto2 from "crypto";
+var LicenseVerifier = class {
+  /**
+   * Decodes, verifies signature, and checks metadata of a license key.
+   *
+   * @param licenseKey - Base64url signed JWT license key.
+   * @param currentProjectId - Project namespace ID.
+   * @param publicKeyOverride - Optional override for unit/integration tests.
+   */
+  static verify(licenseKey, currentProjectId, provider, publicKeyOverride) {
+    const isProduction = process.env.NODE_ENV === "production";
+    if (!licenseKey) {
+      if (isProduction) {
+        throw new ConfigurationException(
+          "[Retrivora SDK] License key (licenseKey) is required in production environments."
+        );
+      }
+      console.warn(
+        `\x1B[33m[Retrivora SDK] WARNING: Running without a license key. This namespace (${currentProjectId}) is permitted for local development only.\x1B[0m`
+      );
+      return {
+        projectId: currentProjectId,
+        expiresAt: Math.floor(Date.now() / 1e3) + 86400,
+        // Valid for 24h
+        tier: "hobby"
+      };
+    }
+    try {
+      const parts = licenseKey.split(".");
+      if (parts.length !== 3) {
+        throw new Error("Malformed token structure (expected 3 parts).");
+      }
+      const [headerB64, payloadB64, signatureB64] = parts;
+      const dataToVerify = `${headerB64}.${payloadB64}`;
+      const publicKey = publicKeyOverride != null ? publicKeyOverride : this.PUBLIC_KEY;
+      const signature = Buffer.from(signatureB64, "base64url");
+      const data = Buffer.from(dataToVerify);
+      const isValid = crypto2.verify(
+        "sha256",
+        data,
+        publicKey,
+        signature
+      );
+      if (!isValid) {
+        throw new Error("Signature verification failed.");
+      }
+      const payload = JSON.parse(
+        Buffer.from(payloadB64, "base64url").toString("utf8")
+      );
+      if (!payload.projectId) {
+        throw new Error('License payload is missing "projectId".');
+      }
+      if (payload.projectId !== currentProjectId) {
+        throw new Error(
+          `Project ID mismatch. License is bound to project namespace "${payload.projectId}" but configuration has "${currentProjectId}".`
+        );
+      }
+      if (!payload.expiresAt) {
+        throw new Error('License payload is missing expiration ("expiresAt").');
+      }
+      const currentTimestampSec = Math.floor(Date.now() / 1e3);
+      if (currentTimestampSec > payload.expiresAt) {
+        throw new Error(
+          `License key has expired. Expiration: ${new Date(
+            payload.expiresAt * 1e3
+          ).toDateString()}`
+        );
+      }
+      if (isProduction && provider) {
+        const tier = (payload.tier || "").toLowerCase();
+        const normalizedProvider = provider.toLowerCase();
+        if (tier === "hobby") {
+          const allowedHobby = ["postgresql", "pgvector", "supabase"];
+          if (!allowedHobby.includes(normalizedProvider)) {
+            throw new Error(
+              `The database provider "${provider}" is not allowed on the Hobby tier. Hobby tier is restricted to PostgreSQL and Supabase. Please upgrade to a Developer Pro or Enterprise plan.`
+            );
+          }
+        } else if (tier === "pro" || tier === "developer_pro" || tier === "premium" || tier === "growth") {
+          const allowedPro = [
+            "postgresql",
+            "pgvector",
+            "supabase",
+            "pinecone",
+            "qdrant",
+            "mongodb",
+            "milvus",
+            "chroma",
+            "chromadb"
+          ];
+          if (!allowedPro.includes(normalizedProvider)) {
+            throw new Error(
+              `The database provider "${provider}" is not allowed on the Developer Pro tier. Please upgrade to an Enterprise plan.`
+            );
+          }
+        }
+      }
+      return payload;
+    } catch (err) {
+      const message = err instanceof Error ? err.message : String(err);
+      throw new ConfigurationException(
+        `[Retrivora SDK] License key validation failed: ${message}`
+      );
+    }
+  }
+};
+// Retrivora's Public Key used to verify the license key signature.
+// In production, this matches the private key held by Retrivora SaaS.
+LicenseVerifier.PUBLIC_KEY = `-----BEGIN PUBLIC KEY-----
+MIIBIjANBgkqhkiG9w0BAQEFAAOCAQ8AMIIBCgKCAQEArMieCkCBtTfByRNOserm
+XM4T0Am29JyNzB4XQPe+WJJ56CN73H1HLXx9n1ByFcxkEJ9po+SURj5DnvUXwEy+
+xstx33wDPpVmcMQe33j5CRYS0S4gICiNqIRN7XkTlJtrcXqrmh1/hdOAwfRbjO1T
+NVtORHwUwWfI/lBLyxUPGtKPed+0fQ7NtcW4o00JXxs3EjCB5BV9CbnXoTjQb/4h
+iwZof4l8rb7oaNzOsVg0RSyxsETX7Ex5sI0CjBz1p2mYp4oVhw/A/h8Ls7H0XzjC
++Eb6BLtsytt5JHoSp0jExLlCpDmpys2L+kETcBVx+IqiXtdN1n6KbkksvEDUVzLI
+MwIDAQAB
+-----END PUBLIC KEY-----`;
+
 // src/rag/DocumentChunker.ts
 var DocumentChunker = class {
   constructor(chunkSize = 1e3, chunkOverlap = 200, separators = ["\n# ", "\n## ", "\n### ", "\n#### ", "\n\n", "\n", " ", ""]) {
@@ -4086,11 +4556,11 @@ var LlamaIndexIngestor = class {
    * than standard character-count splitting.
    */
   async chunk(text, options = {}) {
-    var _a, _b;
+    var _a2, _b;
     try {
       const { Document, SentenceSplitter, MetadataMode } = await import(`${"llamaindex"}`);
       const splitter = new SentenceSplitter({
-        chunkSize: (_a = options.chunkSize) != null ? _a : 1e3,
+        chunkSize: (_a2 = options.chunkSize) != null ? _a2 : 1e3,
         chunkOverlap: (_b = options.chunkOverlap) != null ? _b : 200
       });
       const doc = new Document({ text, metadata: options.metadata || {} });
@@ -4199,7 +4669,7 @@ ${error instanceof Error ? error.message : String(error)}`
    * The agent returns `{ messages: [...] }` — the last message is the final answer.
    */
   async run(input, chatHistory = []) {
-    var _a, _b, _c;
+    var _a2, _b, _c;
     if (!this.agent) {
       throw new Error("[LangChainAgent] Agent not initialized. Call initialize() first.");
     }
@@ -4210,14 +4680,445 @@ ${error instanceof Error ? error.message : String(error)}`
     const response = await this.agent.invoke({
       messages: [...historyMessages, new HumanMessage(input)]
     });
-    const lastMessage = (_a = response == null ? void 0 : response.messages) == null ? void 0 : _a.at(-1);
-    if (lastMessage && typeof lastMessage.content === "string") {
-      return lastMessage.content;
-    }
-    if (Array.isArray(lastMessage == null ? void 0 : lastMessage.content)) {
-      return lastMessage.content.filter((c) => c.type === "text").map((c) => c.text).join("");
-    }
-    return String((_c = (_b = response == null ? void 0 : response.output) != null ? _b : response) != null ? _c : "");
+    const lastMessage = (_a2 = response == null ? void 0 : response.messages) == null ? void 0 : _a2.at(-1);
+    if (lastMessage && typeof lastMessage.content === "string") {
+      return lastMessage.content;
+    }
+    if (Array.isArray(lastMessage == null ? void 0 : lastMessage.content)) {
+      return lastMessage.content.filter((c) => c.type === "text").map((c) => c.text).join("");
+    }
+    return String((_c = (_b = response == null ? void 0 : response.output) != null ? _b : response) != null ? _c : "");
+  }
+};
+
+// src/core/mcp.ts
+import { spawn } from "child_process";
+var MCPClient = class {
+  constructor(config) {
+    this.config = config;
+    this.messageIdCounter = 0;
+    this.pendingRequests = /* @__PURE__ */ new Map();
+    this.stdioBuffer = "";
+    this.initialized = false;
+  }
+  getNextMessageId() {
+    return ++this.messageIdCounter;
+  }
+  /**
+   * Connect and initialize the MCP server connection.
+   */
+  async connect() {
+    if (this.initialized) return;
+    if (this.config.transport === "stdio") {
+      await this.connectStdio();
+    } else {
+      await this.connectSSE();
+    }
+    try {
+      const response = await this.sendJsonRpc("initialize", {
+        protocolVersion: "2024-11-05",
+        capabilities: {},
+        clientInfo: {
+          name: "retrivora-sdk",
+          version: "1.0.0"
+        }
+      });
+      await this.sendNotification("notifications/initialized");
+      this.initialized = true;
+      console.log(`[MCPClient] Connected and initialized server "${this.config.name}" successfully.`);
+    } catch (err) {
+      this.disconnect();
+      throw new Error(`[MCPClient] Failed to initialize server "${this.config.name}": ${err.message}`);
+    }
+  }
+  async connectStdio() {
+    var _a2;
+    const cmd = this.config.command;
+    if (!cmd) {
+      throw new Error(`[MCPClient] Command option is required for stdio transport on "${this.config.name}"`);
+    }
+    const args = this.config.args || [];
+    const env = __spreadValues(__spreadValues({}, process.env), this.config.env || {});
+    this.childProcess = spawn(cmd, args, { env, shell: true });
+    if (!this.childProcess || !this.childProcess.stdout || !this.childProcess.stdin) {
+      throw new Error(`[MCPClient] Failed to spawn stdio child process for "${this.config.name}"`);
+    }
+    this.childProcess.stdout.on("data", (data) => {
+      this.stdioBuffer += data.toString("utf-8");
+      this.processStdioLines();
+    });
+    (_a2 = this.childProcess.stderr) == null ? void 0 : _a2.on("data", (data) => {
+      console.warn(`[MCPClient] [stderr] [${this.config.name}]:`, data.toString("utf-8"));
+    });
+    this.childProcess.on("close", (code) => {
+      console.log(`[MCPClient] Connection closed for "${this.config.name}" with exit code: ${code}`);
+      this.disconnect();
+    });
+    await new Promise((resolve) => setTimeout(resolve, 100));
+  }
+  processStdioLines() {
+    let newlineIndex;
+    while ((newlineIndex = this.stdioBuffer.indexOf("\n")) !== -1) {
+      const line = this.stdioBuffer.substring(0, newlineIndex).trim();
+      this.stdioBuffer = this.stdioBuffer.substring(newlineIndex + 1);
+      if (!line) continue;
+      try {
+        const payload = JSON.parse(line);
+        if (payload.id !== void 0) {
+          const req = this.pendingRequests.get(Number(payload.id));
+          if (req) {
+            this.pendingRequests.delete(Number(payload.id));
+            if (payload.error) {
+              req.reject(new Error(payload.error.message || JSON.stringify(payload.error)));
+            } else {
+              req.resolve(payload.result);
+            }
+          }
+        }
+      } catch (e) {
+        console.warn(`[MCPClient] Error parsing stdio JSON-RPC line from "${this.config.name}":`, e, "Line content:", line);
+      }
+    }
+  }
+  async connectSSE() {
+    if (!this.config.url) {
+      throw new Error(`[MCPClient] URL is required for sse transport on "${this.config.name}"`);
+    }
+    this.sseUrl = this.config.url;
+  }
+  async sendJsonRpc(method, params) {
+    const id = this.getNextMessageId();
+    const request = {
+      jsonrpc: "2.0",
+      id,
+      method,
+      params
+    };
+    if (this.config.transport === "stdio") {
+      if (!this.childProcess || !this.childProcess.stdin) {
+        throw new Error(`[MCPClient] Server "${this.config.name}" is not connected (stdio process missing)`);
+      }
+      return new Promise((resolve, reject) => {
+        this.pendingRequests.set(id, { resolve, reject });
+        this.childProcess.stdin.write(JSON.stringify(request) + "\n", "utf-8");
+      });
+    } else {
+      if (!this.sseUrl) {
+        throw new Error(`[MCPClient] Server "${this.config.name}" SSE URL is not initialized`);
+      }
+      const response = await fetch(this.sseUrl, {
+        method: "POST",
+        headers: { "Content-Type": "application/json" },
+        body: JSON.stringify(request)
+      });
+      if (!response.ok) {
+        throw new Error(`SSE JSON-RPC POST failed with status ${response.status}`);
+      }
+      const payload = await response.json();
+      if (payload.error) {
+        throw new Error(payload.error.message || JSON.stringify(payload.error));
+      }
+      return payload.result;
+    }
+  }
+  async sendNotification(method, params) {
+    var _a2;
+    const notification = {
+      jsonrpc: "2.0",
+      method,
+      params
+    };
+    if (this.config.transport === "stdio") {
+      if ((_a2 = this.childProcess) == null ? void 0 : _a2.stdin) {
+        this.childProcess.stdin.write(JSON.stringify(notification) + "\n", "utf-8");
+      }
+    } else if (this.sseUrl) {
+      await fetch(this.sseUrl, {
+        method: "POST",
+        headers: { "Content-Type": "application/json" },
+        body: JSON.stringify(notification)
+      }).catch((err) => console.warn("[MCPClient] Failed to send notification over SSE:", err));
+    }
+  }
+  /**
+   * List all tools offered by the MCP server.
+   */
+  async listTools() {
+    await this.connect();
+    const result = await this.sendJsonRpc("tools/list", {});
+    return (result == null ? void 0 : result.tools) || [];
+  }
+  /**
+   * Execute a tool on the MCP server.
+   */
+  async callTool(name, args = {}) {
+    await this.connect();
+    return await this.sendJsonRpc("tools/call", { name, arguments: args });
+  }
+  disconnect() {
+    this.initialized = false;
+    this.pendingRequests.forEach((req) => req.reject(new Error("MCPClient disconnected")));
+    this.pendingRequests.clear();
+    if (this.childProcess) {
+      try {
+        this.childProcess.kill();
+      } catch (e) {
+      }
+      this.childProcess = void 0;
+    }
+  }
+};
+var MCPRegistry = class {
+  constructor(servers = []) {
+    this.clients = [];
+    this.clients = servers.map((cfg) => new MCPClient(cfg));
+  }
+  async getAllTools() {
+    const list = [];
+    await Promise.all(
+      this.clients.map(async (client) => {
+        try {
+          const tools = await client.listTools();
+          tools.forEach((t) => list.push({ serverName: client["config"].name, tool: t }));
+        } catch (e) {
+          console.warn(`[MCPRegistry] Failed to fetch tools from server "${client["config"].name}":`, e);
+        }
+      })
+    );
+    return list;
+  }
+  async callTool(toolName, args = {}) {
+    for (const client of this.clients) {
+      try {
+        const tools = await client.listTools();
+        if (tools.some((t) => t.name === toolName)) {
+          return await client.callTool(toolName, args);
+        }
+      } catch (e) {
+      }
+    }
+    throw new Error(`[MCPRegistry] Tool "${toolName}" not found on any registered MCP server.`);
+  }
+  disconnectAll() {
+    this.clients.forEach((c) => c.disconnect());
+  }
+};
+
+// src/core/MultiAgentCoordinator.ts
+var MultiAgentCoordinator = class {
+  constructor(options) {
+    var _a2;
+    this.llmProvider = options.llmProvider;
+    this.mcpRegistry = options.mcpRegistry;
+    this.documentSearch = options.documentSearch;
+    this.maxIterations = (_a2 = options.maxIterations) != null ? _a2 : 5;
+  }
+  /**
+   * Run the multi-agent coordination loop synchronously and return the final response.
+   */
+  async run(question, history = []) {
+    const generator = this.runStream(question, history);
+    let finalReply = "";
+    const sources = [];
+    try {
+      for (var iter = __forAwait(generator), more, temp, error; more = !(temp = await iter.next()).done; more = false) {
+        const chunk = temp.value;
+        if (typeof chunk === "string") {
+          finalReply += chunk;
+        } else if (chunk && typeof chunk === "object") {
+          if ("reply" in chunk && chunk.reply) {
+            finalReply += chunk.reply;
+          }
+          if ("sources" in chunk && chunk.sources) {
+            sources.push(...chunk.sources);
+          }
+        }
+      }
+    } catch (temp) {
+      error = [temp];
+    } finally {
+      try {
+        more && (temp = iter.return) && await temp.call(iter);
+      } finally {
+        if (error)
+          throw error[0];
+      }
+    }
+    return {
+      reply: finalReply,
+      sources
+    };
+  }
+  /**
+   * Run the multi-agent coordination loop as an async stream, yielding intermediate thought blocks and tool execution events.
+   */
+  runStream(_0) {
+    return __asyncGenerator(this, arguments, function* (question, history = []) {
+      const mcpTools = yield new __await(this.mcpRegistry.getAllTools());
+      let systemPrompt = `You are a goal-driven supervisor agent coordinating a RAG search engine and external MCP servers to solve the user's task.
+You must think step-by-step before acting. Write your internal thinking process inside a \`<think>...</think>\` block first, then act or answer.
+
+Available Tools:
+- document_search(query: string): Search the vector database and knowledge base for documents.
+
+${mcpTools.length > 0 ? "MCP Server Tools:" : ""}
+${mcpTools.map((mt) => {
+        var _a2;
+        return `- ${mt.tool.name}(${JSON.stringify(((_a2 = mt.tool.inputSchema) == null ? void 0 : _a2.properties) || {})}): ${mt.tool.description || "No description provided."}`;
+      }).join("\n")}
+
+Tool Calling Protocol:
+If you need to call a tool, write:
+TOOL_CALL: <toolName>({ "argName": "value" })
+And then STOP generating immediately.
+
+Once you have gathered enough information to fully satisfy the user's query, write your final response directly after the closing \`</think>\` block.
+`;
+      const supervisorHistory = [
+        { role: "system", content: systemPrompt },
+        ...history,
+        { role: "user", content: question }
+      ];
+      let iteration = 0;
+      const allSources = [];
+      while (iteration < this.maxIterations) {
+        iteration++;
+        const context = "";
+        let fullModelResponse = "";
+        let textBuffer = "";
+        let inThinking = false;
+        let toolCallDetected = null;
+        if (!this.llmProvider.chatStream) {
+          const reply = yield new __await(this.llmProvider.chat(supervisorHistory, context));
+          fullModelResponse = reply;
+          const thinkIndex = reply.indexOf("<think>");
+          const endThinkIndex = reply.indexOf("</think>");
+          let thinkingText = "";
+          let answerText = reply;
+          if (thinkIndex !== -1 && endThinkIndex !== -1) {
+            thinkingText = reply.substring(thinkIndex + 7, endThinkIndex);
+            answerText = reply.substring(endThinkIndex + 8);
+            yield { type: "thinking", text: thinkingText };
+          }
+          const toolMatch = answerText.match(/TOOL_CALL:\s*(\w+)\s*\(([\s\S]*)\)/);
+          if (toolMatch) {
+            const toolName = toolMatch[1];
+            const toolArgsStr = toolMatch[2];
+            try {
+              const toolArgs = JSON.parse(toolArgsStr.trim());
+              toolCallDetected = { name: toolName, args: toolArgs };
+            } catch (e) {
+              console.error("[MultiAgentCoordinator] Failed to parse tool args:", toolArgsStr, e);
+            }
+          } else {
+            yield answerText;
+          }
+        } else {
+          const stream = this.llmProvider.chatStream(supervisorHistory, context);
+          try {
+            for (var iter = __forAwait(stream), more, temp, error; more = !(temp = yield new __await(iter.next())).done; more = false) {
+              const chunk = temp.value;
+              fullModelResponse += chunk;
+              textBuffer += chunk;
+              if (!inThinking && textBuffer.includes("<think>")) {
+                inThinking = true;
+                const idx = textBuffer.indexOf("<think>");
+                textBuffer = textBuffer.slice(idx + 7);
+              }
+              if (inThinking && textBuffer.includes("</think>")) {
+                inThinking = false;
+                const idx = textBuffer.indexOf("</think>");
+                const thinkingDelta = textBuffer.slice(0, idx);
+                yield { type: "thinking", text: thinkingDelta };
+                textBuffer = textBuffer.slice(idx + 8);
+              }
+              if (inThinking && textBuffer.length > 0) {
+                yield { type: "thinking", text: textBuffer };
+                textBuffer = "";
+              }
+              if (!inThinking && textBuffer.includes("TOOL_CALL:")) {
+                const idx = textBuffer.indexOf("TOOL_CALL:");
+                const precedingText = textBuffer.slice(0, idx);
+                if (precedingText.trim()) {
+                  yield precedingText;
+                }
+                textBuffer = textBuffer.slice(idx);
+              }
+              if (!inThinking && !textBuffer.includes("TOOL_CALL:") && textBuffer.length > 0) {
+                yield textBuffer;
+                textBuffer = "";
+              }
+            }
+          } catch (temp) {
+            error = [temp];
+          } finally {
+            try {
+              more && (temp = iter.return) && (yield new __await(temp.call(iter)));
+            } finally {
+              if (error)
+                throw error[0];
+            }
+          }
+          if (textBuffer.trim()) {
+            const toolMatch = textBuffer.match(/TOOL_CALL:\s*(\w+)\s*\(([\s\S]*)\)/);
+            if (toolMatch) {
+              const toolName = toolMatch[1];
+              const toolArgsStr = toolMatch[2];
+              try {
+                const toolArgs = JSON.parse(toolArgsStr.trim());
+                toolCallDetected = { name: toolName, args: toolArgs };
+              } catch (e) {
+                console.error("[MultiAgentCoordinator] Failed to parse tool args:", toolArgsStr, e);
+              }
+            } else {
+              yield textBuffer;
+            }
+          }
+        }
+        supervisorHistory.push({ role: "assistant", content: fullModelResponse });
+        if (toolCallDetected) {
+          const { name: toolName, args: toolArgs } = toolCallDetected;
+          yield {
+            type: "thinking",
+            text: `
+[Agent Call] Executing tool "${toolName}" with parameters ${JSON.stringify(toolArgs)}...
+`
+          };
+          let toolResultText = "";
+          try {
+            if (toolName === "document_search") {
+              const searchRes = yield new __await(this.documentSearch(toolArgs.query || ""));
+              toolResultText = `document_search returned:
+${searchRes.reply}
+
+Sources count: ${searchRes.sources.length}`;
+              if (searchRes.sources && searchRes.sources.length > 0) {
+                allSources.push(...searchRes.sources);
+                yield { reply: "", sources: searchRes.sources };
+              }
+            } else {
+              const mcpRes = yield new __await(this.mcpRegistry.callTool(toolName, toolArgs));
+              toolResultText = `MCP Tool "${toolName}" returned:
+${JSON.stringify(mcpRes.content)}`;
+            }
+          } catch (err) {
+            toolResultText = `Error calling tool "${toolName}": ${err.message}`;
+          }
+          supervisorHistory.push({
+            role: "user",
+            content: `TOOL_RESULT for ${toolName}:
+${toolResultText}`
+          });
+          yield {
+            type: "thinking",
+            text: `[Agent Output] Tool "${toolName}" executed. Continuing reasoning...
+`
+          };
+        } else {
+          break;
+        }
+      }
+    });
   }
 };
 
@@ -4547,7 +5448,7 @@ var QueryProcessor = class {
    * @param validFields Optional list of known filterable fields to look for
    */
   static extractQueryFieldHints(question, validFields = []) {
-    var _a, _b, _c, _d;
+    var _a2, _b, _c, _d;
     if (!question.trim()) return [];
     const hints = /* @__PURE__ */ new Map();
     const addHint = (value, field) => {
@@ -4627,7 +5528,7 @@ var QueryProcessor = class {
     ];
     for (const p of universalPatterns) {
       for (const match of question.matchAll(p.regex)) {
-        const val = p.group ? (_a = match[p.group]) != null ? _a : match[0] : match[0];
+        const val = p.group ? (_a2 = match[p.group]) != null ? _a2 : match[0] : match[0];
         if (!val) continue;
         if (p.field) addHint(val, p.field);
         else addHint(val);
@@ -4851,11 +5752,11 @@ var LLMRouter = class {
    *   When provided it is used directly as the 'default' role without re-constructing.
    */
   async initialize(prebuiltDefault) {
-    var _a;
+    var _a2;
     const defaultModel = prebuiltDefault != null ? prebuiltDefault : LLMFactory.create(this.config.llm, this.config.embedding);
     this.models.set("default", defaultModel);
     const envFastModel = process.env.FAST_LLM_MODEL;
-    const providerFastDefault = (_a = FAST_MODEL_DEFAULTS[this.config.llm.provider]) != null ? _a : "";
+    const providerFastDefault = (_a2 = FAST_MODEL_DEFAULTS[this.config.llm.provider]) != null ? _a2 : "";
     const fastModelName = envFastModel || providerFastDefault;
     if (fastModelName && fastModelName !== this.config.llm.model) {
       console.log(`[LLMRouter] Fast role \u2192 provider="${this.config.llm.provider}" model="${fastModelName}"`);
@@ -4874,8 +5775,8 @@ var LLMRouter = class {
    * Falls back to 'default' if the requested role is not registered.
    */
   get(role) {
-    var _a;
-    const provider = (_a = this.models.get(role)) != null ? _a : this.models.get("default");
+    var _a2;
+    const provider = (_a2 = this.models.get(role)) != null ? _a2 : this.models.get("default");
     if (!provider) {
       throw new Error(`[LLMRouter] No provider registered for role: "${role}". Did you call initialize()?`);
     }
@@ -4893,7 +5794,7 @@ var UITransformer = class _UITransformer {
    * Prefer `analyzeAndDecide()` in production.
    */
   static transform(userQuery, retrievedData, config, trainedSchema, intent) {
-    var _a, _b, _c;
+    var _a2, _b, _c;
     if (!retrievedData || retrievedData.length === 0) {
       return this.createTextResponse("No data available", "No relevant data found for your query.");
     }
@@ -4913,7 +5814,7 @@ var UITransformer = class _UITransformer {
       return this.transformToBarChart(filteredData, profile, userQuery, resolvedIntent.visualizationHint === "ranking");
     }
     if (resolvedIntent.visualizationHint === "distribution") {
-      return (_a = this.transformToHistogram(profile, userQuery)) != null ? _a : this.transformToBarChart(filteredData, profile, userQuery);
+      return (_a2 = this.transformToHistogram(profile, userQuery)) != null ? _a2 : this.transformToBarChart(filteredData, profile, userQuery);
     }
     if (resolvedIntent.visualizationHint === "correlation") {
       return (_b = this.transformToScatterPlot(profile, userQuery)) != null ? _b : this.transformToBarChart(filteredData, profile, userQuery);
@@ -5238,11 +6139,11 @@ ${schemaProfileText}` : ""}`;
     };
   }
   static transformToPieChart(data, profile, query = "") {
-    var _a;
+    var _a2;
     const dimension = profile ? this.selectDimensionField(profile, query) : void 0;
     const categories = dimension ? Array.from(new Set(profile.records.map((record) => {
-      var _a2;
-      return String((_a2 = record.fields[dimension.key]) != null ? _a2 : "");
+      var _a3;
+      return String((_a3 = record.fields[dimension.key]) != null ? _a3 : "");
     }).filter(Boolean))) : this.detectCategories(data);
     if (categories.length === 0) return null;
     const categoryData = dimension && profile ? this.aggregateProfileByDimension(profile, dimension.key) : this.aggregateByCategory(data, categories);
@@ -5252,7 +6153,7 @@ ${schemaProfileText}` : ""}`;
     });
     return {
       type: "pie_chart",
-      title: `Distribution by ${(_a = dimension == null ? void 0 : dimension.label) != null ? _a : "Category"}`,
+      title: `Distribution by ${(_a2 = dimension == null ? void 0 : dimension.label) != null ? _a2 : "Category"}`,
       description: `Showing breakdown across ${pieData.length} categories`,
       data: pieData
     };
@@ -5262,8 +6163,8 @@ ${schemaProfileText}` : ""}`;
     const valueField = profile.numericFields[0];
     const buckets = /* @__PURE__ */ new Map();
     profile.records.forEach((record) => {
-      var _a, _b, _c;
-      const timestamp = String((_a = record.fields[dateField.key]) != null ? _a : "");
+      var _a2, _b, _c;
+      const timestamp = String((_a2 = record.fields[dateField.key]) != null ? _a2 : "");
       const value = (_b = this.toFiniteNumber(record.fields[valueField.key])) != null ? _b : 0;
       buckets.set(timestamp, ((_c = buckets.get(timestamp)) != null ? _c : 0) + value);
     });
@@ -5276,23 +6177,23 @@ ${schemaProfileText}` : ""}`;
     };
   }
   static transformToBarChart(data, profile, query = "", horizontal = false) {
-    var _a;
+    var _a2;
     const dimension = profile ? this.selectDimensionField(profile, query) : void 0;
     const measure = profile ? this.selectNumericField(profile, query) : void 0;
     const aggregate = dimension && profile ? this.aggregateProfileByDimension(profile, dimension.key, measure == null ? void 0 : measure.key) : this.aggregateByCategory(data, this.detectCategories(data));
     const barData = Object.entries(aggregate).map(([category, value]) => ({ category, value: Number(value) })).sort((a, b) => horizontal ? b.value - a.value : 0).slice(0, 12);
     const fallbackData = barData.length > 0 ? barData : data.slice(0, 12).map((item, index) => {
-      var _a2, _b, _c, _d, _e;
+      var _a3, _b, _c, _d, _e;
       const meta = item.metadata || {};
       const label = String(
-        (_c = (_b = (_a2 = this.getDynamicVal(meta, "name")) != null ? _a2 : meta.label) != null ? _b : meta.title) != null ? _c : `Result ${index + 1}`
+        (_c = (_b = (_a3 = this.getDynamicVal(meta, "name")) != null ? _a3 : meta.label) != null ? _b : meta.title) != null ? _c : `Result ${index + 1}`
       );
       const value = (_e = (_d = this.extractNumericValue(meta)) != null ? _d : item.score) != null ? _e : 0;
       return { category: label, value: Number(value) };
     });
     return {
       type: horizontal ? "horizontal_bar" : "bar_chart",
-      title: dimension ? `${(_a = measure == null ? void 0 : measure.label) != null ? _a : "Count"} by ${dimension.label}` : "Comparison",
+      title: dimension ? `${(_a2 = measure == null ? void 0 : measure.label) != null ? _a2 : "Count"} by ${dimension.label}` : "Comparison",
       description: `Showing ${fallbackData.length} comparable values`,
       data: fallbackData
     };
@@ -5327,11 +6228,11 @@ ${schemaProfileText}` : ""}`;
     if (fields.length < 2) return null;
     const [xField, yField] = fields;
     const points = profile.records.map((record) => {
-      var _a;
+      var _a2;
       const x = this.toFiniteNumber(record.fields[xField.key]);
       const y = this.toFiniteNumber(record.fields[yField.key]);
       if (x === null || y === null) return null;
-      return { x, y, label: String((_a = this.getRecordLabel(record)) != null ? _a : record.id) };
+      return { x, y, label: String((_a2 = this.getRecordLabel(record)) != null ? _a2 : record.id) };
     }).filter((point) => point !== null).slice(0, 100);
     if (points.length === 0) return null;
     return {
@@ -5361,9 +6262,9 @@ ${schemaProfileText}` : ""}`;
   static transformToRadarChart(data) {
     const attributeMap = {};
     data.forEach((item) => {
-      var _a, _b, _c;
+      var _a2, _b, _c;
       const meta = item.metadata || {};
-      const seriesName = String((_c = (_b = (_a = meta.name) != null ? _a : meta.product) != null ? _b : item.id) != null ? _c : "Item");
+      const seriesName = String((_c = (_b = (_a2 = meta.name) != null ? _a2 : meta.product) != null ? _b : item.id) != null ? _c : "Item");
       Object.entries(meta).forEach(([key, val]) => {
         if (typeof val === "number" && !["price", "id"].includes(key.toLowerCase())) {
           if (!attributeMap[key]) attributeMap[key] = {};
@@ -5445,22 +6346,22 @@ ${schemaProfileText}` : ""}`;
     return null;
   }
   static normalizeTransformation(payload) {
-    var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j;
+    var _a2, _b, _c, _d, _e, _f, _g2, _h, _i, _j;
     if (!payload || typeof payload !== "object") return null;
     const p = payload;
     const type = this.normalizeVisualizationType(
-      String((_c = (_b = (_a = p.type) != null ? _a : p.view) != null ? _b : p.chartType) != null ? _c : "")
+      String((_c = (_b = (_a2 = p.type) != null ? _a2 : p.view) != null ? _b : p.chartType) != null ? _c : "")
     );
     if (!type) return null;
     const title = String((_e = (_d = p.title) != null ? _d : p.heading) != null ? _e : "Visualization");
     const description = p.description ? String(p.description) : void 0;
-    const rawData = (_j = (_i = (_h = (_g = (_f = p.data) != null ? _f : p.table) != null ? _g : p.rows) != null ? _h : p.items) != null ? _i : p.content) != null ? _j : null;
+    const rawData = (_j = (_i = (_h = (_g2 = (_f = p.data) != null ? _f : p.table) != null ? _g2 : p.rows) != null ? _h : p.items) != null ? _i : p.content) != null ? _j : null;
     const data = type === "text" && typeof rawData === "string" ? { content: rawData } : rawData;
     const transformation = { type, title, description, data };
     return this.validateTransformation(transformation) ? transformation : null;
   }
   static normalizeVisualizationType(type) {
-    var _a;
+    var _a2;
     const mapping = {
       pie: "pie_chart",
       pie_chart: "pie_chart",
@@ -5488,7 +6389,7 @@ ${schemaProfileText}` : ""}`;
       product_carousel: "product_carousel",
       carousel: "carousel"
     };
-    return (_a = mapping[type.toLowerCase()]) != null ? _a : null;
+    return (_a2 = mapping[type.toLowerCase()]) != null ? _a2 : null;
   }
   static validateTransformation(t) {
     const { type, data } = t;
@@ -5690,16 +6591,16 @@ ${schemaProfileText}` : ""}`;
     return null;
   }
   static selectDimensionField(profile, query) {
-    var _a, _b;
+    var _a2, _b;
     const productCategory = profile.categoricalFields.find(
       (field) => /category|department|collection|type|group|segment|region|country|state|city/i.test(field.key)
     );
     const ranked = this.rankFieldsByQuery(profile.categoricalFields, query);
-    return (_b = (_a = ranked[0]) != null ? _a : productCategory) != null ? _b : profile.categoricalFields[0];
+    return (_b = (_a2 = ranked[0]) != null ? _a2 : productCategory) != null ? _b : profile.categoricalFields[0];
   }
   static selectNumericField(profile, query) {
-    var _a;
-    return (_a = this.rankFieldsByQuery(profile.numericFields, query)[0]) != null ? _a : profile.numericFields[0];
+    var _a2;
+    return (_a2 = this.rankFieldsByQuery(profile.numericFields, query)[0]) != null ? _a2 : profile.numericFields[0];
   }
   static rankFieldsByQuery(fields, query) {
     const q = query.toLowerCase();
@@ -5728,8 +6629,8 @@ ${schemaProfileText}` : ""}`;
   static aggregateProfileByDimension(profile, dimensionKey, measureKey) {
     const result = {};
     profile.records.forEach((record) => {
-      var _a, _b, _c;
-      const category = String((_a = record.fields[dimensionKey]) != null ? _a : "Other").trim() || "Other";
+      var _a2, _b, _c;
+      const category = String((_a2 = record.fields[dimensionKey]) != null ? _a2 : "Other").trim() || "Other";
       const value = measureKey ? (_b = this.toFiniteNumber(record.fields[measureKey])) != null ? _b : 0 : 1;
       result[category] = ((_c = result[category]) != null ? _c : 0) + value;
     });
@@ -5808,8 +6709,8 @@ ${schemaProfileText}` : ""}`;
   static aggregateByCategory(data, categories) {
     const result = Object.fromEntries(categories.map((c) => [c, 0]));
     data.forEach((item) => {
-      var _a;
-      const cat = (_a = this.getProductCategory(item)) != null ? _a : "Other";
+      var _a2;
+      const cat = (_a2 = this.getProductCategory(item)) != null ? _a2 : "Other";
       if (Object.prototype.hasOwnProperty.call(result, cat)) result[cat]++;
       else result["Other"] = (result["Other"] || 0) + 1;
     });
@@ -5817,17 +6718,17 @@ ${schemaProfileText}` : ""}`;
   }
   static extractTimeSeriesData(data) {
     return data.map((item) => {
-      var _a, _b, _c, _d, _e;
+      var _a2, _b, _c, _d, _e;
       const meta = item.metadata || {};
       return {
-        timestamp: (_b = (_a = meta.timestamp) != null ? _a : meta.date) != null ? _b : (/* @__PURE__ */ new Date()).toISOString(),
+        timestamp: (_b = (_a2 = meta.timestamp) != null ? _a2 : meta.date) != null ? _b : (/* @__PURE__ */ new Date()).toISOString(),
         value: (_d = (_c = meta.value) != null ? _c : item.score) != null ? _d : 0,
         label: (_e = meta.label) != null ? _e : item.content.substring(0, 50)
       };
     });
   }
   static extractNumericValue(meta) {
-    var _a;
+    var _a2;
     const preferredKeys = [
       "value",
       "count",
@@ -5842,7 +6743,7 @@ ${schemaProfileText}` : ""}`;
       "price"
     ];
     for (const key of preferredKeys) {
-      const raw = (_a = resolveMetadataValue(meta, key)) != null ? _a : meta[key];
+      const raw = (_a2 = resolveMetadataValue(meta, key)) != null ? _a2 : meta[key];
       const value = typeof raw === "number" ? raw : Number(String(raw != null ? raw : "").replace(/[$,% ,]/g, ""));
       if (Number.isFinite(value)) return value;
     }
@@ -5988,8 +6889,8 @@ ${schemaProfileText}` : ""}`;
     let inStock = 0;
     let outOfStock = 0;
     data.forEach((d) => {
-      var _a, _b;
-      const cat = (_a = this.getProductCategory(d)) != null ? _a : "Other";
+      var _a2, _b;
+      const cat = (_a2 = this.getProductCategory(d)) != null ? _a2 : "Other";
       if (cat === category) {
         const quantity = (_b = this.extractStockQuantity(d)) != null ? _b : 1;
         if (this.determineStockStatus(d)) inStock += quantity;
@@ -6033,9 +6934,9 @@ ${schemaProfileText}` : ""}`;
   }
   // ─── Product Extraction ───────────────────────────────────────────────────
   static getDynamicVal(meta, uiKey, config, trainedSchema) {
-    var _a;
+    var _a2;
     if (!meta) return void 0;
-    const mapping = (_a = config == null ? void 0 : config.rag) == null ? void 0 : _a.uiMapping;
+    const mapping = (_a2 = config == null ? void 0 : config.rag) == null ? void 0 : _a2.uiMapping;
     if ((mapping == null ? void 0 : mapping[uiKey]) && meta[mapping[uiKey]] !== void 0) return meta[mapping[uiKey]];
     if (trainedSchema && typeof trainedSchema === "object") {
       const trainedKey = trainedSchema[uiKey];
@@ -6044,13 +6945,13 @@ ${schemaProfileText}` : ""}`;
     return resolveMetadataValue(meta, uiKey);
   }
   static extractProductInfo(item, config, trainedSchema) {
-    var _a;
+    var _a2;
     const meta = item.metadata || {};
     const name = this.getDynamicVal(meta, "name", config, trainedSchema);
     const price = this.getDynamicVal(meta, "price", config, trainedSchema);
     const brand = this.getDynamicVal(meta, "brand", config, trainedSchema);
     const description = this.cleanProductDescription(
-      (_a = this.extractProductDescriptionFromContent(item.content)) != null ? _a : this.getProductDescriptionValue(meta, config, trainedSchema)
+      (_a2 = this.extractProductDescriptionFromContent(item.content)) != null ? _a2 : this.getProductDescriptionValue(meta, config, trainedSchema)
     );
     if (name || this.isProductData(item)) {
       let finalName = name ? String(name) : void 0;
@@ -6158,10 +7059,10 @@ RULES:
   }
   static buildContextSummary(sources, maxChars = 6e3) {
     const items = sources.map((s, i) => {
-      var _a, _b, _c, _d;
+      var _a2, _b, _c, _d;
       return {
         index: i + 1,
-        content: (_b = (_a = s.content) == null ? void 0 : _a.substring(0, 400)) != null ? _b : "",
+        content: (_b = (_a2 = s.content) == null ? void 0 : _a2.substring(0, 400)) != null ? _b : "",
         metadata: (_c = s.metadata) != null ? _c : {},
         score: (_d = s.score) != null ? _d : 0
       };
@@ -6354,9 +7255,11 @@ var Pipeline = class {
     /** LRU-bounded cache: avoids re-embedding identical queries within the same process. */
     this.embeddingCache = new LRUEmbeddingCache(500);
     this.initialised = false;
-    var _a, _b, _c, _d, _e;
+    /** Namespace-specific static cold context cache for CAG */
+    this.coldContexts = /* @__PURE__ */ new Map();
+    var _a2, _b, _c, _d, _e;
     this.chunker = new DocumentChunker(
-      (_b = (_a = config.rag) == null ? void 0 : _a.chunkSize) != null ? _b : 1e3,
+      (_b = (_a2 = config.rag) == null ? void 0 : _a2.chunkSize) != null ? _b : 1e3,
       (_d = (_c = config.rag) == null ? void 0 : _c.chunkOverlap) != null ? _d : 200
     );
     if (((_e = config.rag) == null ? void 0 : _e.chunkingStrategy) === "llamaindex") {
@@ -6372,7 +7275,7 @@ var Pipeline = class {
     return this.initialised ? this.llmProvider : void 0;
   }
   async initialize() {
-    var _a;
+    var _a2, _b, _c;
     if (this.initialised) return;
     const chartInstruction = `You are a helpful assistant. Use the provided context to answer questions accurately.
 
@@ -6415,12 +7318,47 @@ var Pipeline = class {
       this.entityExtractor = new EntityExtractor(this.llmProvider);
     }
     await this.vectorDB.initialize();
-    if (((_a = this.config.rag) == null ? void 0 : _a.architecture) === "agentic") {
+    if (((_a2 = this.config.rag) == null ? void 0 : _a2.architecture) === "agentic" || this.config.mcpServers && this.config.mcpServers.length > 0) {
+      this.mcpRegistry = new MCPRegistry(this.config.mcpServers || []);
+      this.multiAgentCoordinator = new MultiAgentCoordinator({
+        llmProvider: this.llmProvider,
+        mcpRegistry: this.mcpRegistry,
+        documentSearch: async (query) => {
+          return this.runNormalQuery(query);
+        }
+      });
       this.agent = new LangChainAgent(this, this.config);
       await this.agent.initialize(this.llmProvider);
     }
+    if ((_c = (_b = this.config.rag) == null ? void 0 : _b.cag) == null ? void 0 : _c.enabled) {
+      await this.loadColdContext(this.config.projectId);
+    }
     this.initialised = true;
   }
+  async loadColdContext(ns) {
+    var _a2, _b;
+    const cagConfig = (_a2 = this.config.rag) == null ? void 0 : _a2.cag;
+    if (!cagConfig || !cagConfig.enabled) return;
+    try {
+      const coldNs = cagConfig.coldNamespace || ns;
+      const limit = (_b = cagConfig.coldLimit) != null ? _b : 20;
+      const queryText = cagConfig.coldQuery || "documentation";
+      const queryVector = await this.embeddingProvider.embed(queryText, { taskType: "query" });
+      const coldMatches = await this.vectorDB.query(queryVector, limit, coldNs);
+      if (coldMatches.length > 0) {
+        const formatted = coldMatches.map((m, i) => `[Cold Source ${i + 1}]
+${m.content}`).join("\n\n---\n\n");
+        this.coldContexts.set(ns, formatted);
+        console.log(`[Pipeline] Cold RAG (CAG) loaded ${coldMatches.length} documents into cached context for namespace: ${ns}`);
+      } else {
+        this.coldContexts.set(ns, "No cold context found.");
+        console.warn(`[Pipeline] Cold RAG (CAG) initialized but no documents were found in namespace: ${coldNs}`);
+      }
+    } catch (err) {
+      console.error(`[Pipeline] Failed to load Cold RAG (CAG) context for namespace ${ns}:`, err);
+      this.coldContexts.set(ns, "Failed to load cold context.");
+    }
+  }
   /**
    * Ingest documents with automatic chunking, embedding, and batch upsert.
    */
@@ -6452,12 +7390,12 @@ var Pipeline = class {
   }
   /** Step 1: Chunk the document content. */
   async prepareChunks(doc) {
-    var _a, _b;
+    var _a2, _b;
     if (this.llamaIngestor) {
       return this.llamaIngestor.chunk(doc.content, {
         docId: doc.docId,
         metadata: doc.metadata,
-        chunkSize: (_a = this.config.rag) == null ? void 0 : _a.chunkSize,
+        chunkSize: (_a2 = this.config.rag) == null ? void 0 : _a2.chunkSize,
         chunkOverlap: (_b = this.config.rag) == null ? void 0 : _b.chunkOverlap
       });
     }
@@ -6528,12 +7466,37 @@ var Pipeline = class {
       extractionOptions
     );
   }
+  async runNormalQuery(question, history = [], namespace) {
+    const stream = this.askStreamInternal(question, history, namespace);
+    let reply = "";
+    let sources = [];
+    try {
+      for (var iter = __forAwait(stream), more, temp, error; more = !(temp = await iter.next()).done; more = false) {
+        const chunk = temp.value;
+        if (typeof chunk === "string") {
+          reply += chunk;
+        } else if (typeof chunk === "object" && chunk !== null) {
+          if ("reply" in chunk && chunk.reply) reply += chunk.reply;
+          if ("sources" in chunk && chunk.sources) sources = chunk.sources;
+        }
+      }
+    } catch (temp) {
+      error = [temp];
+    } finally {
+      try {
+        more && (temp = iter.return) && await temp.call(iter);
+      } finally {
+        if (error)
+          throw error[0];
+      }
+    }
+    return { reply, sources };
+  }
   async ask(question, history = [], namespace) {
-    var _a;
+    var _a2;
     await this.initialize();
-    if (((_a = this.config.rag) == null ? void 0 : _a.architecture) === "agentic" && this.agent) {
-      const agentReply = await this.agent.run(question, history);
-      return { reply: agentReply, sources: [] };
+    if ((((_a2 = this.config.rag) == null ? void 0 : _a2.architecture) === "agentic" || this.config.mcpServers && this.config.mcpServers.length > 0) && this.multiAgentCoordinator) {
+      return await this.multiAgentCoordinator.run(question, history);
     }
     const stream = this.askStream(question, history, namespace);
     let reply = "";
@@ -6565,6 +7528,47 @@ var Pipeline = class {
     }
     return { reply, sources, graphData, ui_transformation: uiTransformation, trace };
   }
+  askStream(_0) {
+    return __asyncGenerator(this, arguments, function* (question, history = [], namespace) {
+      var _a2;
+      yield new __await(this.initialize());
+      if ((((_a2 = this.config.rag) == null ? void 0 : _a2.architecture) === "agentic" || this.config.mcpServers && this.config.mcpServers.length > 0) && this.multiAgentCoordinator) {
+        const stream2 = this.multiAgentCoordinator.runStream(question, history);
+        try {
+          for (var iter = __forAwait(stream2), more, temp, error; more = !(temp = yield new __await(iter.next())).done; more = false) {
+            const chunk = temp.value;
+            yield chunk;
+          }
+        } catch (temp) {
+          error = [temp];
+        } finally {
+          try {
+            more && (temp = iter.return) && (yield new __await(temp.call(iter)));
+          } finally {
+            if (error)
+              throw error[0];
+          }
+        }
+        return;
+      }
+      const stream = this.askStreamInternal(question, history, namespace);
+      try {
+        for (var iter2 = __forAwait(stream), more2, temp2, error2; more2 = !(temp2 = yield new __await(iter2.next())).done; more2 = false) {
+          const chunk = temp2.value;
+          yield chunk;
+        }
+      } catch (temp2) {
+        error2 = [temp2];
+      } finally {
+        try {
+          more2 && (temp2 = iter2.return) && (yield new __await(temp2.call(iter2)));
+        } finally {
+          if (error2)
+            throw error2[0];
+        }
+      }
+    });
+  }
   /**
    * High-performance streaming RAG flow.
    * Yields text chunks first, then the retrieval metadata + observability trace at the end.
@@ -6575,12 +7579,12 @@ var Pipeline = class {
    * - UITransformation is computed after text streaming and emitted with metadata
    * - SchemaMapper.train runs while answer generation streams
    */
-  askStream(_0) {
+  askStreamInternal(_0) {
     return __asyncGenerator(this, arguments, function* (question, history = [], namespace) {
-      var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j, _k, _l, _m, _n, _o, _p, _q, _r, _s, _t, _u, _v;
+      var _a2, _b, _c, _d, _e, _f, _g2, _h, _i, _j, _k, _l, _m, _n, _o, _p, _q, _r, _s, _t, _u, _v, _w, _x, _y, _z;
       yield new __await(this.initialize());
       const ns = namespace != null ? namespace : this.config.projectId;
-      const topK = (_b = (_a = this.config.rag) == null ? void 0 : _a.topK) != null ? _b : 5;
+      const topK = (_b = (_a2 = this.config.rag) == null ? void 0 : _a2.topK) != null ? _b : 5;
       const scoreThreshold = (_d = (_c = this.config.rag) == null ? void 0 : _c.scoreThreshold) != null ? _d : 0.3;
       const requestId = `req_${Date.now()}_${Math.random().toString(36).slice(2, 8)}`;
       const requestStart = performance.now();
@@ -6593,7 +7597,7 @@ var Pipeline = class {
         }
         const hints = QueryProcessor.extractQueryFieldHints(question, (_f = this.config.rag) == null ? void 0 : _f.filterableFields);
         const filter = QueryProcessor.buildQueryFilter(question, hints);
-        const numericPredicates = QueryProcessor.extractNumericPredicates(question, (_g = this.config.rag) == null ? void 0 : _g.filterableFields);
+        const numericPredicates = QueryProcessor.extractNumericPredicates(question, (_g2 = this.config.rag) == null ? void 0 : _g2.filterableFields);
         if (numericPredicates.length > 0) {
           filter.__numericPredicates = numericPredicates;
         }
@@ -6659,6 +7663,32 @@ ${graphContext}
 VECTOR CONTEXT:
 ${context}`;
         }
+        if ((_n = (_m = this.config.rag) == null ? void 0 : _m.cag) == null ? void 0 : _n.enabled) {
+          if (!this.coldContexts.has(ns)) {
+            yield new __await(this.loadColdContext(ns));
+          }
+          const coldContext = this.coldContexts.get(ns);
+          if (coldContext && coldContext !== "No cold context found." && coldContext !== "Failed to load cold context.") {
+            context = `COLD CONTEXT (STATIC KNOWLEDGE):
+${coldContext}
+
+RETRIEVED HOT CONTEXT (REAL-TIME KNOWLEDGE):
+${context}`;
+          }
+        }
+        yield {
+          reply: "",
+          sources: sources.map((s) => {
+            var _a3;
+            return {
+              id: s.id,
+              score: s.score,
+              content: s.content,
+              metadata: (_a3 = s.metadata) != null ? _a3 : {},
+              namespace: ns
+            };
+          })
+        };
         const allMetadataKeys = Array.from(new Set(sources.flatMap((s) => Object.keys(s.metadata || {}))));
         const cachedSchema = allMetadataKeys.length > 0 ? SchemaMapper.getCached(ns, allMetadataKeys) : void 0;
         if (allMetadataKeys.length > 0 && !cachedSchema) {
@@ -6686,10 +7716,10 @@ ${context}`;
         let hallucinationScoringPromise = Promise.resolve(void 0);
         const restrictionSuffix = "\n\n(IMPORTANT: Format your response beautifully using rich Markdown! Use bolding for emphasis, headings (##) for structure, bullet points for lists, and proper line breaks between paragraphs to make it highly readable. However, NEVER generate Markdown tables, HTML figures, or text charts (the UI renders requested charts separately, so NEVER say you cannot create, display, draw, render, or provide a chart/visualization). If listing products, use simple markdown bullet points or comma-separated names. NEVER use plus signs (+) to separate product names, category names, or list items. For product description/detail questions, provide customer-facing prose only and do NOT list internal catalog fields such as Handle, Body (HTML), Vendor, Type, Tags, Published, Option, Variant, SKU, or raw metadata labels. You CAN use numbers for years/counts like 2006 or 5800, but NEVER put a dollar sign ($) before them.)";
         const isClaude37 = this.config.llm.model.includes("claude-3-7-sonnet");
-        const isNativeThinking = isClaude37 && (((_m = this.config.llm.options) == null ? void 0 : _m.thinking) === true || ((_n = this.config.llm.options) == null ? void 0 : _n.thinking) === "enabled" || ((_o = this.config.llm.options) == null ? void 0 : _o.thinking) !== false);
+        const isNativeThinking = isClaude37 && (((_o = this.config.llm.options) == null ? void 0 : _o.thinking) === true || ((_p = this.config.llm.options) == null ? void 0 : _p.thinking) === "enabled" || ((_q = this.config.llm.options) == null ? void 0 : _q.thinking) !== false);
         const modelNameLower = this.config.llm.model.toLowerCase();
         const isReasoningModel = modelNameLower.includes("-r1") || modelNameLower.includes("deepseek-r1") || modelNameLower.includes("reasoning") || modelNameLower.includes("thinking");
-        const isSimulatedThinking = !isNativeThinking && (((_p = this.config.llm.options) == null ? void 0 : _p.thinking) === true || ((_q = this.config.llm.options) == null ? void 0 : _q.thinking) === "enabled" || isReasoningModel && ((_r = this.config.llm.options) == null ? void 0 : _r.thinking) !== false);
+        const isSimulatedThinking = !isNativeThinking && (((_r = this.config.llm.options) == null ? void 0 : _r.thinking) === true || ((_s = this.config.llm.options) == null ? void 0 : _s.thinking) === "enabled" || isReasoningModel && ((_t = this.config.llm.options) == null ? void 0 : _t.thinking) !== false);
         let finalRestrictionSuffix = restrictionSuffix;
         if (isSimulatedThinking) {
           finalRestrictionSuffix += "\n\n(IMPORTANT: You must think step-by-step before answering. Write your thinking process inside a `<think>...</think>` block. Write the `<think>` block first, then follow it with your final response. Keep the thinking process thorough and detailed. Example: `<think>Thinking details here...</think>Final answer text here.`)";
@@ -6697,7 +7727,7 @@ ${context}`;
         const hardenedHistory = [...history];
         const userQuestion = { role: "user", content: question + finalRestrictionSuffix };
         const messages = [...hardenedHistory, userQuestion];
-        const systemPrompt = (_s = this.config.llm.systemPrompt) != null ? _s : "";
+        const systemPrompt = (_u = this.config.llm.systemPrompt) != null ? _u : "";
         const userPrompt = messages.map((m) => `${m.role}: ${m.content}`).join("\n");
         let fullReply = "";
         let thinkingText = "";
@@ -6808,7 +7838,7 @@ ${context}`;
           }
           yield fullReply;
         }
-        const runHallucination = ((_t = this.config.llm.options) == null ? void 0 : _t.hallucinationScoring) === true || this.config.llm.provider !== "ollama" && ((_u = this.config.llm.options) == null ? void 0 : _u.hallucinationScoring) !== false;
+        const runHallucination = ((_v = this.config.llm.options) == null ? void 0 : _v.hallucinationScoring) === true || this.config.llm.provider !== "ollama" && ((_w = this.config.llm.options) == null ? void 0 : _w.hallucinationScoring) !== false;
         if (runHallucination) {
           hallucinationScoringPromise = scoreHallucination(this.llmProvider, fullReply, context).catch(() => void 0);
         }
@@ -6817,7 +7847,7 @@ ${context}`;
         const latency = {
           embedMs: Math.round(embedMs),
           retrieveMs: Math.round(retrieveMs),
-          rerankMs: ((_v = this.config.rag) == null ? void 0 : _v.useReranking) ? Math.round(rerankMs) : void 0,
+          rerankMs: ((_x = this.config.rag) == null ? void 0 : _x.useReranking) ? Math.round(rerankMs) : void 0,
           generateMs: Math.round(generateMs),
           totalMs: Math.round(totalMs)
         };
@@ -6830,33 +7860,63 @@ ${context}`;
           totalTokens: promptTokens + completionTokens,
           estimatedCostUsd: estimateCostUsd(promptTokens, completionTokens, this.config.llm.model)
         };
+        const awaitHallucination = ((_y = this.config.llm.options) == null ? void 0 : _y.awaitHallucination) === true;
         const [uiTransformation, hallucinationResult] = yield new __await(Promise.all([
           uiTransformationPromise,
-          hallucinationScoringPromise
+          awaitHallucination ? hallucinationScoringPromise : Promise.resolve(void 0)
         ]));
-        const trace = __spreadValues({
+        const buildTrace = (hScore) => __spreadValues({
           requestId,
           query: question,
           rewrittenQuery,
           systemPrompt,
           userPrompt: question + finalRestrictionSuffix,
           chunks: sources.map((s) => {
-            var _a2;
+            var _a3;
             return {
               id: s.id,
               score: s.score,
               content: s.content,
-              metadata: (_a2 = s.metadata) != null ? _a2 : {},
+              metadata: (_a3 = s.metadata) != null ? _a3 : {},
               namespace: ns
             };
           }),
           latency,
           tokens,
           timestamp: (/* @__PURE__ */ new Date()).toISOString()
-        }, hallucinationResult ? {
-          hallucinationScore: hallucinationResult.score,
-          hallucinationReason: hallucinationResult.reason
+        }, hScore ? {
+          hallucinationScore: hScore.score,
+          hallucinationReason: hScore.reason
         } : {});
+        const trace = buildTrace(hallucinationResult);
+        if ((_z = this.config.telemetry) == null ? void 0 : _z.enabled) {
+          const telemetryUrl = this.config.telemetry.url || "/api/telemetry";
+          const absoluteUrl = telemetryUrl.startsWith("http") ? telemetryUrl : (process.env.NEXT_PUBLIC_APP_URL || `http://localhost:${process.env.PORT || 3e3}`) + telemetryUrl;
+          (async () => {
+            try {
+              let finalTrace = trace;
+              if (!awaitHallucination && runHallucination) {
+                const backgroundScoreResult = await hallucinationScoringPromise;
+                if (backgroundScoreResult) {
+                  finalTrace = buildTrace(backgroundScoreResult);
+                }
+              }
+              await fetch(absoluteUrl, {
+                method: "POST",
+                headers: {
+                  "Content-Type": "application/json"
+                },
+                body: JSON.stringify({
+                  trace: finalTrace,
+                  licenseKey: this.config.licenseKey,
+                  projectId: ns
+                })
+              });
+            } catch (err) {
+              console.warn(`[Pipeline Telemetry] Failed to send trace async to ${absoluteUrl}:`, err.message);
+            }
+          })();
+        }
         yield {
           reply: "",
           sources,
@@ -6876,7 +7936,7 @@ ${context}`;
    * Uses an LRU-bounded embedding cache to avoid re-embedding the same query.
    */
   async generateUiTransformation(question, sources, cachedSchema, forceDeterministic = false) {
-    var _a;
+    var _a2;
     if (!sources || sources.length === 0) {
       return UITransformer.transform(question, sources, this.config, cachedSchema);
     }
@@ -6890,7 +7950,7 @@ ${context}`;
       );
     }
     const isLocalProvider = this.config.llm.provider === "ollama";
-    const disableLlmUiTransform = ((_a = this.config.llm.options) == null ? void 0 : _a.disableLlmUiTransform) === true;
+    const disableLlmUiTransform = ((_a2 = this.config.llm.options) == null ? void 0 : _a2.disableLlmUiTransform) === true;
     if (forceDeterministic || isLocalProvider || disableLlmUiTransform) {
       return UITransformer.transform(question, sources, this.config, cachedSchema);
     }
@@ -6908,9 +7968,9 @@ ${context}`;
       const value = this.resolveNumericPredicateValue(source, predicate);
       return value !== null && this.matchesNumericPredicate(value, predicate);
     })).sort((a, b) => {
-      var _a, _b;
+      var _a2, _b;
       const primary = predicates[0];
-      const aValue = (_a = this.resolveNumericPredicateValue(a, primary)) != null ? _a : 0;
+      const aValue = (_a2 = this.resolveNumericPredicateValue(a, primary)) != null ? _a2 : 0;
       const bValue = (_b = this.resolveNumericPredicateValue(b, primary)) != null ? _b : 0;
       return primary.operator === "lt" || primary.operator === "lte" ? aValue - bValue : bValue - aValue;
     });
@@ -6982,8 +8042,8 @@ ${context}`;
     return Number.isFinite(numeric) ? numeric : null;
   }
   async retrieve(query, options) {
-    var _a, _b, _c, _d, _e, _f, _g;
-    const ns = (_a = options.namespace) != null ? _a : this.config.projectId;
+    var _a2, _b, _c, _d, _e, _f, _g2;
+    const ns = (_a2 = options.namespace) != null ? _a2 : this.config.projectId;
     const topK = (_b = options.topK) != null ? _b : 5;
     const cacheKey = `${ns}::${query}`;
     let queryVector = this.embeddingCache.get(cacheKey);
@@ -7015,7 +8075,7 @@ ${context}`;
     ) : [];
     const resolvedSources = [];
     for (const source of sources) {
-      const parentId = (_g = source.metadata) == null ? void 0 : _g.parent_id;
+      const parentId = (_g2 = source.metadata) == null ? void 0 : _g2.parent_id;
       if (parentId) {
         console.log(`[Pipeline] Multi-Vector: Found child chunk. Parent ID: ${parentId}`);
         resolvedSources.push(source);
@@ -7097,8 +8157,14 @@ var Retrivora = class {
     this.pipeline = new Pipeline(this.config);
   }
   async initialize() {
+    var _a2;
     try {
       await ConfigValidator.validateAndThrow(this.config);
+      LicenseVerifier.verify(
+        this.config.licenseKey,
+        this.config.projectId,
+        (_a2 = this.config.vectorDb) == null ? void 0 : _a2.provider
+      );
     } catch (err) {
       throw wrapError(err, "CONFIGURATION_ERROR");
     }
@@ -7272,11 +8338,20 @@ var ProviderHealthCheck = class {
 // src/core/VectorPlugin.ts
 var VectorPlugin = class {
   constructor(hostConfig) {
-    this.config = ConfigResolver.resolve(hostConfig);
-    this.validationPromise = ConfigValidator.validateAndThrow(this.config);
+    const resolvedConfig = ConfigResolver.resolve(hostConfig);
+    this.config = resolvedConfig;
+    this.validationPromise = (async () => {
+      var _a2;
+      await ConfigValidator.validateAndThrow(resolvedConfig);
+      LicenseVerifier.verify(
+        resolvedConfig.licenseKey,
+        resolvedConfig.projectId,
+        (_a2 = resolvedConfig.vectorDb) == null ? void 0 : _a2.provider
+      );
+    })();
     this.validationPromise.catch(() => {
     });
-    this.pipeline = new Pipeline(this.config);
+    this.pipeline = new Pipeline(resolvedConfig);
   }
   /**
    * Get the current resolved configuration.
@@ -7412,13 +8487,13 @@ var ConfigBuilder = class {
    * Configure the vector database provider
    */
   vectorDb(provider, options) {
-    var _a;
+    var _a2;
     if (provider === "auto") {
       this._vectorDb = this._autoDetectVectorDb();
     } else {
       this._vectorDb = {
         provider,
-        indexName: (_a = options == null ? void 0 : options.indexName) != null ? _a : "default",
+        indexName: (_a2 = options == null ? void 0 : options.indexName) != null ? _a2 : "default",
         options: __spreadValues({}, options)
       };
     }
@@ -7428,7 +8503,7 @@ var ConfigBuilder = class {
    * Configure the LLM provider for chat
    */
   llm(provider, model, apiKey, options) {
-    var _a, _b;
+    var _a2, _b;
     if (provider === "auto") {
       this._llm = this._autoDetectLLM();
     } else {
@@ -7437,7 +8512,7 @@ var ConfigBuilder = class {
         model: model != null ? model : "default-model",
         apiKey,
         systemPrompt: options == null ? void 0 : options.systemPrompt,
-        maxTokens: (_a = options == null ? void 0 : options.maxTokens) != null ? _a : 1024,
+        maxTokens: (_a2 = options == null ? void 0 : options.maxTokens) != null ? _a2 : 1024,
         temperature: (_b = options == null ? void 0 : options.temperature) != null ? _b : 0.7,
         baseUrl: options == null ? void 0 : options.baseUrl,
         options
@@ -7480,8 +8555,8 @@ var ConfigBuilder = class {
    * Set RAG-specific pipeline parameters
    */
   rag(options) {
-    var _a;
-    this._rag = __spreadValues(__spreadValues({}, (_a = this._rag) != null ? _a : {}), options);
+    var _a2;
+    this._rag = __spreadValues(__spreadValues({}, (_a2 = this._rag) != null ? _a2 : {}), options);
     return this;
   }
   /**
@@ -7497,7 +8572,7 @@ var ConfigBuilder = class {
    * Throws if required fields (projectId, vectorDb, llm, embedding) are not set.
    */
   build() {
-    var _a;
+    var _a2;
     const missing = [];
     if (!this._projectId) missing.push('projectId (call .projectId("my-app"))');
     if (!this._vectorDb) missing.push('vectorDb (call .vectorDb("pinecone", { ... }))');
@@ -7509,7 +8584,7 @@ var ConfigBuilder = class {
   ${missing.join("\n  ")}`
       );
     }
-    const ragConfig = (_a = this._rag) != null ? _a : { chunkSize: 1e3, chunkOverlap: 200, topK: 5 };
+    const ragConfig = (_a2 = this._rag) != null ? _a2 : { chunkSize: 1e3, chunkOverlap: 200, topK: 5 };
     if (process.env.RAG_USE_GRAPH_RETRIEVAL === "true") {
       ragConfig.useGraphRetrieval = true;
     }
@@ -7605,8 +8680,8 @@ var DocumentParser = class {
    * Extract text from a File or Buffer based on its type.
    */
   static async parse(file, fileName, mimeType) {
-    var _a;
-    const extension = (_a = fileName.split(".").pop()) == null ? void 0 : _a.toLowerCase();
+    var _a2;
+    const extension = (_a2 = fileName.split(".").pop()) == null ? void 0 : _a2.toLowerCase();
     if (extension === "txt" || extension === "md" || mimeType === "text/plain" || mimeType === "text/markdown") {
       return this.readAsText(file);
     }
@@ -7673,6 +8748,497 @@ init_UniversalVectorProvider();
 
 // src/handlers/index.ts
 import { NextResponse } from "next/server";
+
+// src/core/DatabaseStorage.ts
+import * as fs from "fs";
+import * as path from "path";
+var DatabaseStorage = class {
+  constructor(config) {
+    // Dynamic PG Pool
+    this.pgPool = null;
+    // Dynamic MongoDB Client
+    this.mongoClient = null;
+    this.mongoDbName = "";
+    // Filesystem fallback configuration
+    this.fallbackDir = path.join(process.cwd(), ".retrivora");
+    this.historyFile = path.join(this.fallbackDir, "history.json");
+    this.feedbackFile = path.join(this.fallbackDir, "feedback.json");
+    var _a2, _b, _c, _d, _e;
+    this.config = config;
+    this.provider = ((_a2 = config.vectorDb) == null ? void 0 : _a2.provider) || "fs";
+    const rawHistoryTable = ((_c = (_b = config.rag) == null ? void 0 : _b.history) == null ? void 0 : _c.tableName) || "retrivora_history";
+    const rawFeedbackTable = ((_e = (_d = config.rag) == null ? void 0 : _d.feedback) == null ? void 0 : _e.tableName) || "retrivora_feedback";
+    this.historyTableName = rawHistoryTable.replace(/[^a-zA-Z0-9_]/g, "_");
+    this.feedbackTableName = rawFeedbackTable.replace(/[^a-zA-Z0-9_]/g, "_");
+  }
+  /**
+   * Asserts the user is running a valid Premium/Enterprise license in production.
+   * In non-production (development / test) environments this is a complete no-op —
+   * all History and Feedback features are freely accessible for local development.
+   */
+  checkPremiumSubscription() {
+    if (process.env.NODE_ENV !== "production") {
+      return;
+    }
+    if (!this.config.licenseKey) {
+      throw new Error(
+        "[Retrivora SDK] Chat History and Feedback features require a Premium or Enterprise license key in production."
+      );
+    }
+    try {
+      const payload = LicenseVerifier.verify(this.config.licenseKey, this.config.projectId);
+      const tier = (payload.tier || "").toLowerCase();
+      if (tier !== "premium" && tier !== "enterprise" && tier !== "growth" && tier !== "pro" && tier !== "developer_pro") {
+        throw new Error(
+          `[Retrivora SDK] Chat History and Feedback features are not available on the "${tier}" tier. Please upgrade to a Developer Pro or Enterprise plan.`
+        );
+      }
+    } catch (err) {
+      throw new Error(
+        `[Retrivora SDK] Subscription check failed: ${err instanceof Error ? err.message : String(err)}`
+      );
+    }
+  }
+  checkHistoryEnabled() {
+    var _a2, _b;
+    this.checkPremiumSubscription();
+    if (((_b = (_a2 = this.config.rag) == null ? void 0 : _a2.history) == null ? void 0 : _b.enabled) === false) {
+      throw new Error("[Retrivora SDK] Chat History is disabled in RagConfig.");
+    }
+  }
+  checkFeedbackEnabled() {
+    var _a2, _b;
+    this.checkPremiumSubscription();
+    if (((_b = (_a2 = this.config.rag) == null ? void 0 : _a2.feedback) == null ? void 0 : _b.enabled) === false) {
+      throw new Error("[Retrivora SDK] User Feedback is disabled in RagConfig.");
+    }
+  }
+  async getPGPool() {
+    const globalKey = `__retrivora_pg_pool_${this.config.projectId}`;
+    const _g2 = global;
+    if (_g2[globalKey]) {
+      this.pgPool = _g2[globalKey];
+      return this.pgPool;
+    }
+    const opts = this.config.vectorDb.options;
+    if (!opts.connectionString) {
+      throw new Error("[DatabaseStorage] pg connectionString option is missing.");
+    }
+    const { Pool: Pool3 } = await import("pg");
+    this.pgPool = new Pool3({ connectionString: opts.connectionString });
+    _g2[globalKey] = this.pgPool;
+    const client = await this.pgPool.connect();
+    try {
+      await client.query(`
+        CREATE TABLE IF NOT EXISTS ${this.historyTableName} (
+          id                TEXT PRIMARY KEY,
+          session_id        TEXT NOT NULL,
+          role              TEXT NOT NULL,
+          content           TEXT NOT NULL,
+          sources           JSONB,
+          ui_transformation JSONB,
+          trace             JSONB,
+          created_at        TIMESTAMP WITH TIME ZONE DEFAULT CURRENT_TIMESTAMP
+        )
+      `);
+      await client.query(`
+        CREATE TABLE IF NOT EXISTS ${this.feedbackTableName} (
+          id          TEXT PRIMARY KEY,
+          message_id  TEXT NOT NULL UNIQUE,
+          session_id  TEXT NOT NULL,
+          rating      TEXT NOT NULL,
+          comment     TEXT,
+          created_at  TIMESTAMP WITH TIME ZONE DEFAULT CURRENT_TIMESTAMP
+        )
+      `);
+    } finally {
+      client.release();
+    }
+    return this.pgPool;
+  }
+  async getMongoClient() {
+    const globalKey = `__retrivora_mongo_client_${this.config.projectId}`;
+    const _g2 = global;
+    if (_g2[globalKey]) {
+      this.mongoClient = _g2[globalKey];
+      this.mongoDbName = this.config.vectorDb.options.database;
+      return this.mongoClient;
+    }
+    const opts = this.config.vectorDb.options;
+    if (!opts.uri || !opts.database) {
+      throw new Error("[DatabaseStorage] mongo uri and database options are missing.");
+    }
+    const { MongoClient: MongoClient2 } = await import("mongodb");
+    this.mongoClient = new MongoClient2(opts.uri);
+    await this.mongoClient.connect();
+    _g2[globalKey] = this.mongoClient;
+    this.mongoDbName = opts.database;
+    return this.mongoClient;
+  }
+  getMongoDb() {
+    return this.mongoClient.db(this.mongoDbName);
+  }
+  // Helper for FS fallback reads
+  readLocalFile(filePath) {
+    if (!fs.existsSync(filePath)) return [];
+    try {
+      const raw = fs.readFileSync(filePath, "utf-8");
+      return JSON.parse(raw) || [];
+    } catch (e) {
+      return [];
+    }
+  }
+  // Helper for FS fallback writes
+  writeLocalFile(filePath, data) {
+    if (!fs.existsSync(this.fallbackDir)) {
+      fs.mkdirSync(this.fallbackDir, { recursive: true });
+    }
+    fs.writeFileSync(filePath, JSON.stringify(data, null, 2), "utf-8");
+  }
+  // ─── Message History Operations ──────────────────────────────────────────
+  async saveMessage(sessionId, message) {
+    this.checkHistoryEnabled();
+    const createdAt = (/* @__PURE__ */ new Date()).toISOString();
+    const msgId = message.id || `msg_${Date.now()}_${Math.random().toString(36).slice(2, 6)}`;
+    if (this.provider === "postgresql" || this.provider === "pgvector") {
+      const pool = await this.getPGPool();
+      await pool.query(
+        `INSERT INTO ${this.historyTableName} (id, session_id, role, content, sources, ui_transformation, trace, created_at)
+         VALUES ($1, $2, $3, $4, $5, $6, $7, $8)
+         ON CONFLICT (id) DO UPDATE 
+         SET content = EXCLUDED.content, sources = EXCLUDED.sources, ui_transformation = EXCLUDED.ui_transformation, trace = EXCLUDED.trace`,
+        [
+          msgId,
+          sessionId,
+          message.role,
+          message.content,
+          message.sources ? JSON.stringify(message.sources) : null,
+          message.uiTransformation ? JSON.stringify(message.uiTransformation) : null,
+          message.trace ? JSON.stringify(message.trace) : null,
+          createdAt
+        ]
+      );
+    } else if (this.provider === "mongodb") {
+      await this.getMongoClient();
+      const db = this.getMongoDb();
+      const col = db.collection(this.historyTableName);
+      await col.updateOne(
+        { id: msgId },
+        {
+          $set: {
+            id: msgId,
+            session_id: sessionId,
+            role: message.role,
+            content: message.content,
+            sources: message.sources || null,
+            ui_transformation: message.uiTransformation || null,
+            trace: message.trace || null,
+            created_at: createdAt
+          }
+        },
+        { upsert: true }
+      );
+    } else {
+      const history = this.readLocalFile(this.historyFile);
+      const index = history.findIndex((h) => h.id === msgId);
+      const item = {
+        id: msgId,
+        session_id: sessionId,
+        role: message.role,
+        content: message.content,
+        sources: message.sources || null,
+        ui_transformation: message.uiTransformation || null,
+        trace: message.trace || null,
+        created_at: createdAt
+      };
+      if (index !== -1) {
+        history[index] = item;
+      } else {
+        history.push(item);
+      }
+      this.writeLocalFile(this.historyFile, history);
+    }
+  }
+  async getHistory(sessionId) {
+    this.checkHistoryEnabled();
+    if (this.provider === "postgresql" || this.provider === "pgvector") {
+      const pool = await this.getPGPool();
+      const res = await pool.query(
+        `SELECT id, role, content, sources, ui_transformation as "uiTransformation", trace, created_at as "createdAt"
+         FROM ${this.historyTableName} 
+         WHERE session_id = $1 
+         ORDER BY created_at ASC`,
+        [sessionId]
+      );
+      return res.rows;
+    } else if (this.provider === "mongodb") {
+      await this.getMongoClient();
+      const db = this.getMongoDb();
+      const col = db.collection(this.historyTableName);
+      const items = await col.find({ session_id: sessionId }).sort({ created_at: 1 }).toArray();
+      return items.map((item) => ({
+        id: item.id,
+        role: item.role,
+        content: item.content,
+        sources: item.sources,
+        uiTransformation: item.ui_transformation,
+        trace: item.trace,
+        createdAt: item.created_at
+      }));
+    } else {
+      const history = this.readLocalFile(this.historyFile);
+      return history.filter((h) => h.session_id === sessionId).sort((a, b) => new Date(a.created_at).getTime() - new Date(b.created_at).getTime());
+    }
+  }
+  async clearHistory(sessionId) {
+    this.checkHistoryEnabled();
+    if (this.provider === "postgresql" || this.provider === "pgvector") {
+      const pool = await this.getPGPool();
+      await pool.query(`DELETE FROM ${this.historyTableName} WHERE session_id = $1`, [sessionId]);
+      await pool.query(`DELETE FROM ${this.feedbackTableName} WHERE session_id = $1`, [sessionId]);
+    } else if (this.provider === "mongodb") {
+      await this.getMongoClient();
+      const db = this.getMongoDb();
+      await db.collection(this.historyTableName).deleteMany({ session_id: sessionId });
+      await db.collection(this.feedbackTableName).deleteMany({ session_id: sessionId });
+    } else {
+      const history = this.readLocalFile(this.historyFile);
+      const filteredHistory = history.filter((h) => h.session_id !== sessionId);
+      this.writeLocalFile(this.historyFile, filteredHistory);
+      const feedback = this.readLocalFile(this.feedbackFile);
+      const filteredFeedback = feedback.filter((f) => f.session_id !== sessionId);
+      this.writeLocalFile(this.feedbackFile, filteredFeedback);
+    }
+  }
+  async listSessions() {
+    this.checkHistoryEnabled();
+    if (this.provider === "postgresql" || this.provider === "pgvector") {
+      const pool = await this.getPGPool();
+      const res = await pool.query(`SELECT DISTINCT session_id FROM ${this.historyTableName}`);
+      return res.rows.map((r) => r.session_id);
+    } else if (this.provider === "mongodb") {
+      await this.getMongoClient();
+      const db = this.getMongoDb();
+      return db.collection(this.historyTableName).distinct("session_id");
+    } else {
+      const history = this.readLocalFile(this.historyFile);
+      const sessions = new Set(history.map((h) => h.session_id));
+      return Array.from(sessions);
+    }
+  }
+  // ─── Feedback Operations ──────────────────────────────────────────────────
+  async saveFeedback(feedback) {
+    this.checkFeedbackEnabled();
+    const id = `fb_${Date.now()}_${Math.random().toString(36).slice(2, 6)}`;
+    const createdAt = (/* @__PURE__ */ new Date()).toISOString();
+    if (this.provider === "postgresql" || this.provider === "pgvector") {
+      const pool = await this.getPGPool();
+      await pool.query(
+        `INSERT INTO ${this.feedbackTableName} (id, message_id, session_id, rating, comment, created_at)
+         VALUES ($1, $2, $3, $4, $5, $6)
+         ON CONFLICT (message_id) DO UPDATE 
+         SET rating = EXCLUDED.rating, comment = EXCLUDED.comment`,
+        [id, feedback.messageId, feedback.sessionId, feedback.rating, feedback.comment || null, createdAt]
+      );
+    } else if (this.provider === "mongodb") {
+      await this.getMongoClient();
+      const db = this.getMongoDb();
+      const col = db.collection(this.feedbackTableName);
+      await col.updateOne(
+        { message_id: feedback.messageId },
+        {
+          $set: {
+            id,
+            message_id: feedback.messageId,
+            session_id: feedback.sessionId,
+            rating: feedback.rating,
+            comment: feedback.comment || null,
+            created_at: createdAt
+          }
+        },
+        { upsert: true }
+      );
+    } else {
+      const list = this.readLocalFile(this.feedbackFile);
+      const index = list.findIndex((f) => f.message_id === feedback.messageId);
+      const item = {
+        id,
+        message_id: feedback.messageId,
+        session_id: feedback.sessionId,
+        rating: feedback.rating,
+        comment: feedback.comment || null,
+        created_at: createdAt
+      };
+      if (index !== -1) {
+        list[index] = item;
+      } else {
+        list.push(item);
+      }
+      this.writeLocalFile(this.feedbackFile, list);
+    }
+  }
+  async getFeedback(messageId) {
+    this.checkFeedbackEnabled();
+    if (this.provider === "postgresql" || this.provider === "pgvector") {
+      const pool = await this.getPGPool();
+      const res = await pool.query(
+        `SELECT id, message_id as "messageId", session_id as "sessionId", rating, comment, created_at as "createdAt"
+         FROM ${this.feedbackTableName} 
+         WHERE message_id = $1`,
+        [messageId]
+      );
+      return res.rows[0] || null;
+    } else if (this.provider === "mongodb") {
+      await this.getMongoClient();
+      const db = this.getMongoDb();
+      const col = db.collection(this.feedbackTableName);
+      const item = await col.findOne({ message_id: messageId });
+      if (!item) return null;
+      return {
+        id: item.id,
+        messageId: item.message_id,
+        sessionId: item.session_id,
+        rating: item.rating,
+        comment: item.comment,
+        createdAt: item.created_at
+      };
+    } else {
+      const list = this.readLocalFile(this.feedbackFile);
+      return list.find((f) => f.message_id === messageId) || null;
+    }
+  }
+  async listFeedback() {
+    this.checkFeedbackEnabled();
+    if (this.provider === "postgresql" || this.provider === "pgvector") {
+      const pool = await this.getPGPool();
+      const res = await pool.query(
+        `SELECT id, message_id as "messageId", session_id as "sessionId", rating, comment, created_at as "createdAt"
+         FROM ${this.feedbackTableName} 
+         ORDER BY created_at DESC`
+      );
+      return res.rows;
+    } else if (this.provider === "mongodb") {
+      await this.getMongoClient();
+      const db = this.getMongoDb();
+      const col = db.collection(this.feedbackTableName);
+      const items = await col.find({}).sort({ created_at: -1 }).toArray();
+      return items.map((item) => ({
+        id: item.id,
+        messageId: item.message_id,
+        sessionId: item.session_id,
+        rating: item.rating,
+        comment: item.comment,
+        createdAt: item.created_at
+      }));
+    } else {
+      const list = this.readLocalFile(this.feedbackFile);
+      return [...list].sort((a, b) => new Date(b.created_at).getTime() - new Date(a.created_at).getTime());
+    }
+  }
+  async disconnect() {
+    if (this.pgPool) {
+      await this.pgPool.end();
+      this.pgPool = null;
+    }
+    if (this.mongoClient) {
+      await this.mongoClient.close();
+      this.mongoClient = null;
+    }
+  }
+};
+
+// src/handlers/index.ts
+async function checkAuth(req, onAuthorize) {
+  if (!onAuthorize) return null;
+  try {
+    const res = await onAuthorize(req);
+    if (res === false) {
+      return NextResponse.json({ error: "Unauthorized" }, { status: 401 });
+    }
+    if (res instanceof Response) {
+      return res;
+    }
+    return null;
+  } catch (err) {
+    const msg = err instanceof Error ? err.message : "Authorization check failed";
+    return NextResponse.json({ error: msg }, { status: 401 });
+  }
+}
+var RateLimiter = class {
+  constructor(windowMs = 6e4, max = 30) {
+    this.hits = /* @__PURE__ */ new Map();
+    this.windowMs = windowMs;
+    this.max = max;
+  }
+  /** Returns true if the request should be allowed, false if rate-limited. */
+  allow(key) {
+    const now = Date.now();
+    const cutoff = now - this.windowMs;
+    const existing = (this.hits.get(key) || []).filter((t) => t > cutoff);
+    existing.push(now);
+    this.hits.set(key, existing);
+    if (this.hits.size > 5e3) {
+      for (const [k, timestamps] of this.hits.entries()) {
+        if (timestamps.every((t) => t <= cutoff)) this.hits.delete(k);
+      }
+    }
+    return existing.length <= this.max;
+  }
+  retryAfterSec(key) {
+    const now = Date.now();
+    const cutoff = now - this.windowMs;
+    const existing = (this.hits.get(key) || []).filter((t) => t > cutoff);
+    if (existing.length === 0) return 0;
+    return Math.ceil((existing[0] + this.windowMs - now) / 1e3);
+  }
+};
+var _g = global;
+var _a;
+var rateLimiter = (_a = _g.__retrivoraRateLimiter) != null ? _a : _g.__retrivoraRateLimiter = new RateLimiter(6e4, 30);
+function getRateLimitKey(req) {
+  var _a2, _b;
+  const ip = ((_b = (_a2 = req.headers.get("x-forwarded-for")) == null ? void 0 : _a2.split(",")[0]) == null ? void 0 : _b.trim()) || req.headers.get("x-real-ip") || "127.0.0.1";
+  return `ip:${ip}`;
+}
+function checkRateLimit(req) {
+  const key = getRateLimitKey(req);
+  if (!rateLimiter.allow(key)) {
+    const retryAfter = rateLimiter.retryAfterSec(key);
+    return new Response(
+      JSON.stringify({ error: { code: "RATE_LIMITED", message: "Too many requests. Please slow down." } }),
+      {
+        status: 429,
+        headers: {
+          "Content-Type": "application/json",
+          "Retry-After": String(retryAfter),
+          "X-RateLimit-Limit": "30",
+          "X-RateLimit-Window": "60"
+        }
+      }
+    );
+  }
+  return null;
+}
+var MAX_MESSAGE_LENGTH = 8e3;
+function sanitizeInput(raw) {
+  const stripped = raw.replace(/[\u0000-\u0008\u000B\u000C\u000E-\u001F\u007F-\u009F]/g, "");
+  const trimmed = stripped.trim();
+  if (!trimmed) {
+    return { ok: false, error: "message must not be empty" };
+  }
+  if (trimmed.length > MAX_MESSAGE_LENGTH) {
+    return { ok: false, error: `message must be \u2264 ${MAX_MESSAGE_LENGTH} characters` };
+  }
+  return { ok: true, value: trimmed };
+}
+function getOrCreatePlugin(configOrPlugin) {
+  if (configOrPlugin instanceof VectorPlugin) return configOrPlugin;
+  const cacheKey = "__retrivoraPlugin_" + JSON.stringify(configOrPlugin != null ? configOrPlugin : {}).slice(0, 64);
+  if (!_g[cacheKey]) {
+    _g[cacheKey] = new VectorPlugin(configOrPlugin);
+  }
+  return _g[cacheKey];
+}
 function sseFrame(payload) {
   return `data: ${JSON.stringify(payload)}
 
@@ -7710,47 +9276,88 @@ var SSE_HEADERS = {
   "X-Accel-Buffering": "no"
   // Disable Nginx buffering for streaming
 };
-function createChatHandler(configOrPlugin) {
-  const plugin = configOrPlugin instanceof VectorPlugin ? configOrPlugin : new VectorPlugin(configOrPlugin);
+function createChatHandler(configOrPlugin, options) {
+  const plugin = getOrCreatePlugin(configOrPlugin);
+  const storage = new DatabaseStorage(plugin.getConfig());
+  const onAuthorize = options == null ? void 0 : options.onAuthorize;
   return async function POST(req) {
+    const authResult = await checkAuth(req, onAuthorize);
+    if (authResult) return authResult;
+    const rateLimited = checkRateLimit(req);
+    if (rateLimited) return rateLimited;
     try {
       const body = await req.json();
-      const { message, history = [], namespace } = body;
-      if (!(message == null ? void 0 : message.trim())) {
-        return NextResponse.json({ error: "message is required" }, { status: 400 });
-      }
+      const { message: rawMessage, history = [], namespace, sessionId = "default", messageId, userMessageId } = body;
+      const sanitized = sanitizeInput(rawMessage != null ? rawMessage : "");
+      if (!sanitized.ok) {
+        return NextResponse.json({ error: { code: "INVALID_INPUT", message: sanitized.error } }, { status: 400 });
+      }
+      const message = sanitized.value;
+      const userMsgId = userMessageId || `user_${Date.now()}_${Math.random().toString(36).slice(2, 6)}`;
+      await storage.saveMessage(sessionId, {
+        id: userMsgId,
+        role: "user",
+        content: message
+      }).catch((err) => console.warn("[createChatHandler] Failed to save user message:", err));
       const result = await plugin.chat(message, history, namespace);
-      return NextResponse.json(result);
+      const assistantMsgId = messageId || `assistant_${Date.now()}_${Math.random().toString(36).slice(2, 6)}`;
+      await storage.saveMessage(sessionId, {
+        id: assistantMsgId,
+        role: "assistant",
+        content: result.reply,
+        sources: result.sources,
+        uiTransformation: result.ui_transformation,
+        trace: result.trace
+      }).catch((err) => console.warn("[createChatHandler] Failed to save assistant message:", err));
+      return NextResponse.json(__spreadProps(__spreadValues({}, result), {
+        messageId: assistantMsgId,
+        userMessageId: userMsgId
+      }));
     } catch (err) {
       const message = err instanceof Error ? err.message : "Internal server error";
       return NextResponse.json({ error: message }, { status: 500 });
     }
   };
 }
-function createStreamHandler(configOrPlugin) {
-  const plugin = configOrPlugin instanceof VectorPlugin ? configOrPlugin : new VectorPlugin(configOrPlugin);
+function createStreamHandler(configOrPlugin, options) {
+  const plugin = getOrCreatePlugin(configOrPlugin);
+  const storage = new DatabaseStorage(plugin.getConfig());
+  const onAuthorize = options == null ? void 0 : options.onAuthorize;
   return async function POST(req) {
+    const authResult = await checkAuth(req, onAuthorize);
+    if (authResult) return authResult;
+    const rateLimited = checkRateLimit(req);
+    if (rateLimited) return rateLimited;
     let body;
     try {
       body = await req.json();
     } catch (e) {
-      return new Response(JSON.stringify({ error: "Invalid JSON body" }), {
+      return new Response(JSON.stringify({ error: { code: "INVALID_JSON", message: "Invalid JSON body" } }), {
         status: 400,
         headers: { "Content-Type": "application/json" }
       });
     }
-    const { message, history = [], namespace } = body;
-    if (!(message == null ? void 0 : message.trim())) {
-      return new Response(JSON.stringify({ error: "message is required" }), {
-        status: 400,
-        headers: { "Content-Type": "application/json" }
-      });
+    const { message: rawMessage, history = [], namespace, sessionId = "default", messageId, userMessageId } = body;
+    const sanitized = sanitizeInput(rawMessage != null ? rawMessage : "");
+    if (!sanitized.ok) {
+      return new Response(
+        JSON.stringify({ error: { code: "INVALID_INPUT", message: sanitized.error } }),
+        { status: 400, headers: { "Content-Type": "application/json" } }
+      );
     }
+    const message = sanitized.value;
+    const userMsgId = userMessageId || `user_${Date.now()}_${Math.random().toString(36).slice(2, 6)}`;
+    await storage.saveMessage(sessionId, {
+      id: userMsgId,
+      role: "user",
+      content: message
+    }).catch((err) => console.warn("[createStreamHandler] Failed to save user message:", err));
+    const assistantMsgId = messageId || `assistant_${Date.now()}_${Math.random().toString(36).slice(2, 6)}`;
     const encoder = new TextEncoder();
     let isActive = true;
     const stream = new ReadableStream({
       async start(controller) {
-        var _a;
+        var _a2;
         const enqueue = (text) => {
           if (!isActive) return;
           try {
@@ -7761,11 +9368,13 @@ function createStreamHandler(configOrPlugin) {
         };
         try {
           const pipelineStream = plugin.chatStream(message, history, namespace);
+          let fullReply = "";
           try {
             for (var iter = __forAwait(pipelineStream), more, temp, error; more = !(temp = await iter.next()).done; more = false) {
               const chunk = temp.value;
               if (!isActive) break;
               if (typeof chunk === "string") {
+                fullReply += chunk;
                 enqueue(sseTextFrame(chunk));
               } else if (chunk && typeof chunk === "object" && "type" in chunk && chunk.type === "thinking") {
                 enqueue(`data: ${JSON.stringify(chunk)}
@@ -7778,9 +9387,10 @@ function createStreamHandler(configOrPlugin) {
                 if (responseChunk == null ? void 0 : responseChunk.trace) {
                   enqueue(sseObservabilityFrame(responseChunk.trace));
                 }
+                let uiTransformation = responseChunk == null ? void 0 : responseChunk.ui_transformation;
                 if (sources.length > 0) {
                   try {
-                    const uiTransformation = (_a = responseChunk == null ? void 0 : responseChunk.ui_transformation) != null ? _a : UITransformer.transform(message, sources, plugin.getConfig());
+                    uiTransformation = (_a2 = responseChunk == null ? void 0 : responseChunk.ui_transformation) != null ? _a2 : UITransformer.transform(message, sources, plugin.getConfig());
                     if (uiTransformation) {
                       enqueue(sseUIFrame(uiTransformation));
                     }
@@ -7788,11 +9398,22 @@ function createStreamHandler(configOrPlugin) {
                     console.warn("[createStreamHandler] UI transformation warning:", transformError);
                     try {
                       const fallback = UITransformer.transform(message, sources, plugin.getConfig());
-                      if (fallback) enqueue(sseUIFrame(fallback));
+                      if (fallback) {
+                        uiTransformation = fallback;
+                        enqueue(sseUIFrame(fallback));
+                      }
                     } catch (e) {
                     }
                   }
                 }
+                await storage.saveMessage(sessionId, {
+                  id: assistantMsgId,
+                  role: "assistant",
+                  content: fullReply,
+                  sources,
+                  uiTransformation,
+                  trace: responseChunk == null ? void 0 : responseChunk.trace
+                }).catch((err) => console.warn("[createStreamHandler] Failed to save assistant message:", err));
               }
             }
           } catch (temp) {
@@ -7829,12 +9450,15 @@ function createStreamHandler(configOrPlugin) {
         console.log("[createStreamHandler] Stream connection closed by client:", reason);
       }
     });
-    return new Response(stream, { headers: SSE_HEADERS });
+    return new Response(stream, { headers: __spreadProps(__spreadValues({}, SSE_HEADERS), { "x-message-id": assistantMsgId, "x-user-message-id": userMsgId }) });
   };
 }
-function createIngestHandler(configOrPlugin) {
-  const plugin = configOrPlugin instanceof VectorPlugin ? configOrPlugin : new VectorPlugin(configOrPlugin);
+function createIngestHandler(configOrPlugin, options) {
+  const plugin = getOrCreatePlugin(configOrPlugin);
+  const onAuthorize = options == null ? void 0 : options.onAuthorize;
   return async function POST(req) {
+    const authResult = await checkAuth(req, onAuthorize);
+    if (authResult) return authResult;
     try {
       const body = await req.json();
       const { documents, namespace } = body;
@@ -7849,9 +9473,14 @@ function createIngestHandler(configOrPlugin) {
     }
   };
 }
-function createHealthHandler(configOrPlugin) {
-  const plugin = configOrPlugin instanceof VectorPlugin ? configOrPlugin : new VectorPlugin(configOrPlugin);
-  return async function GET() {
+function createHealthHandler(configOrPlugin, options) {
+  const plugin = getOrCreatePlugin(configOrPlugin);
+  const onAuthorize = options == null ? void 0 : options.onAuthorize;
+  return async function GET(req) {
+    if (req) {
+      const authResult = await checkAuth(req, onAuthorize);
+      if (authResult) return authResult;
+    }
     try {
       const health = await plugin.checkHealth();
       const status = health.allHealthy ? "ok" : "degraded";
@@ -7866,9 +9495,12 @@ function createHealthHandler(configOrPlugin) {
     }
   };
 }
-function createUploadHandler(configOrPlugin) {
-  const plugin = configOrPlugin instanceof VectorPlugin ? configOrPlugin : new VectorPlugin(configOrPlugin);
+function createUploadHandler(configOrPlugin, options) {
+  const plugin = getOrCreatePlugin(configOrPlugin);
+  const onAuthorize = options == null ? void 0 : options.onAuthorize;
   return async function POST(req) {
+    const authResult = await checkAuth(req, onAuthorize);
+    if (authResult) return authResult;
     try {
       const formData = await req.formData();
       const files = formData.getAll("files");
@@ -7942,9 +9574,12 @@ function createUploadHandler(configOrPlugin) {
     }
   };
 }
-function createSuggestionsHandler(configOrPlugin) {
-  const plugin = configOrPlugin instanceof VectorPlugin ? configOrPlugin : new VectorPlugin(configOrPlugin);
+function createSuggestionsHandler(configOrPlugin, options) {
+  const plugin = getOrCreatePlugin(configOrPlugin);
+  const onAuthorize = options == null ? void 0 : options.onAuthorize;
   return async function POST(req) {
+    const authResult = await checkAuth(req, onAuthorize);
+    if (authResult) return authResult;
     try {
       const body = await req.json();
       const { query, namespace } = body;
@@ -7959,13 +9594,108 @@ function createSuggestionsHandler(configOrPlugin) {
     }
   };
 }
-function createRagHandler(configOrPlugin) {
-  const plugin = configOrPlugin instanceof VectorPlugin ? configOrPlugin : new VectorPlugin(configOrPlugin);
-  const chatHandler = createChatHandler(plugin);
-  const streamHandler = createStreamHandler(plugin);
-  const uploadHandler = createUploadHandler(plugin);
-  const healthHandler = createHealthHandler(plugin);
-  const suggestionsHandler = createSuggestionsHandler(plugin);
+function createHistoryHandler(configOrPlugin, options) {
+  const plugin = getOrCreatePlugin(configOrPlugin);
+  const storage = new DatabaseStorage(plugin.getConfig());
+  const onAuthorize = options == null ? void 0 : options.onAuthorize;
+  return async function handler(req) {
+    const authResult = await checkAuth(req, onAuthorize);
+    if (authResult) return authResult;
+    const url = new URL(req.url);
+    const sessionId = url.searchParams.get("sessionId") || "default";
+    if (req.method === "GET") {
+      try {
+        const history = await storage.getHistory(sessionId);
+        return NextResponse.json({ history });
+      } catch (err) {
+        const message = err instanceof Error ? err.message : "Failed to fetch history";
+        return NextResponse.json({ error: message }, { status: 500 });
+      }
+    } else if (req.method === "POST") {
+      try {
+        const body = await req.json().catch(() => ({}));
+        const sid = body.sessionId || sessionId;
+        await storage.clearHistory(sid);
+        return NextResponse.json({ success: true });
+      } catch (err) {
+        const message = err instanceof Error ? err.message : "Failed to clear history";
+        return NextResponse.json({ error: message }, { status: 500 });
+      }
+    }
+    return NextResponse.json({ error: "Method Not Allowed" }, { status: 405 });
+  };
+}
+function createFeedbackHandler(configOrPlugin, options) {
+  const plugin = getOrCreatePlugin(configOrPlugin);
+  const storage = new DatabaseStorage(plugin.getConfig());
+  const onAuthorize = options == null ? void 0 : options.onAuthorize;
+  return async function handler(req) {
+    const authResult = await checkAuth(req, onAuthorize);
+    if (authResult) return authResult;
+    if (req.method === "GET") {
+      try {
+        const url = new URL(req.url);
+        const messageId = url.searchParams.get("messageId");
+        if (!messageId) {
+          const feedbackList = await storage.listFeedback();
+          return NextResponse.json({ feedback: feedbackList });
+        }
+        const feedback = await storage.getFeedback(messageId);
+        return NextResponse.json({ feedback });
+      } catch (err) {
+        const message = err instanceof Error ? err.message : "Failed to fetch feedback";
+        return NextResponse.json({ error: message }, { status: 500 });
+      }
+    } else if (req.method === "POST") {
+      try {
+        const body = await req.json();
+        const { messageId, sessionId = "default", rating, comment } = body;
+        if (!messageId) {
+          return NextResponse.json({ error: "messageId is required" }, { status: 400 });
+        }
+        if (!rating) {
+          return NextResponse.json({ error: "rating is required" }, { status: 400 });
+        }
+        await storage.saveFeedback({ messageId, sessionId, rating, comment });
+        return NextResponse.json({ success: true });
+      } catch (err) {
+        const message = err instanceof Error ? err.message : "Failed to save feedback";
+        return NextResponse.json({ error: message }, { status: 500 });
+      }
+    }
+    return NextResponse.json({ error: "Method Not Allowed" }, { status: 405 });
+  };
+}
+function createSessionsHandler(configOrPlugin, options) {
+  const plugin = getOrCreatePlugin(configOrPlugin);
+  const storage = new DatabaseStorage(plugin.getConfig());
+  const onAuthorize = options == null ? void 0 : options.onAuthorize;
+  return async function GET(req) {
+    if (req) {
+      const authResult = await checkAuth(req, onAuthorize);
+      if (authResult) return authResult;
+    }
+    void req;
+    try {
+      const sessions = await storage.listSessions();
+      return NextResponse.json({ sessions });
+    } catch (err) {
+      const message = err instanceof Error ? err.message : "Failed to list sessions";
+      return NextResponse.json({ error: message }, { status: 500 });
+    }
+  };
+}
+function createRagHandler(configOrPlugin, options) {
+  const plugin = getOrCreatePlugin(configOrPlugin);
+  const onAuthorize = options == null ? void 0 : options.onAuthorize;
+  const chatHandler = createChatHandler(plugin, { onAuthorize });
+  const streamHandler = createStreamHandler(plugin, { onAuthorize });
+  const uploadHandler = createUploadHandler(plugin, { onAuthorize });
+  const healthHandler = createHealthHandler(plugin, { onAuthorize });
+  const suggestionsHandler = createSuggestionsHandler(plugin, { onAuthorize });
+  const historyHandler = createHistoryHandler(plugin, { onAuthorize });
+  const feedbackHandler = createFeedbackHandler(plugin, { onAuthorize });
+  const sessionsHandler = createSessionsHandler(plugin, { onAuthorize });
   async function routePostRequest(req, segment) {
     switch (segment) {
       case "chat":
@@ -7977,22 +9707,35 @@ function createRagHandler(configOrPlugin) {
       case "suggestions":
         return suggestionsHandler(req);
       case "health":
-        return healthHandler();
+        return healthHandler(req);
+      case "history":
+      case "history/clear":
+        return historyHandler(req);
+      case "feedback":
+        return feedbackHandler(req);
       default:
         return NextResponse.json({ error: `Not Found: POST segment "${segment}" not supported.` }, { status: 404 });
     }
   }
   async function routeGetRequest(req, segment) {
-    if (segment === "health") {
-      return healthHandler();
+    switch (segment) {
+      case "health":
+        return healthHandler(req);
+      case "history":
+        return historyHandler(req);
+      case "feedback":
+        return feedbackHandler(req);
+      case "sessions":
+        return sessionsHandler(req);
+      default:
+        return NextResponse.json({ error: `Method Not Allowed: GET segment "${segment}" not supported.` }, { status: 405 });
     }
-    return NextResponse.json({ error: `Method Not Allowed: GET is only supported for "health" segment.` }, { status: 405 });
   }
   async function getSegment(context) {
-    var _a;
-    const resolvedParams = typeof ((_a = context == null ? void 0 : context.params) == null ? void 0 : _a.then) === "function" ? await context.params : context == null ? void 0 : context.params;
+    var _a2;
+    const resolvedParams = typeof ((_a2 = context == null ? void 0 : context.params) == null ? void 0 : _a2.then) === "function" ? await context.params : context == null ? void 0 : context.params;
     const segments = (resolvedParams == null ? void 0 : resolvedParams.retrivora) || [];
-    return segments[0] || "chat";
+    return segments.join("/") || "chat";
   }
   return {
     GET: async (req, context) => {
@@ -8030,8 +9773,10 @@ export {
   EmbeddingFailedException,
   EmbeddingStrategy,
   EmbeddingStrategyResolver,
+  GroqProvider,
   LLMFactory,
   LLM_PROFILES,
+  LicenseVerifier,
   MilvusProvider,
   MongoDBProvider,
   MultiTablePostgresProvider,
@@ -8045,6 +9790,7 @@ export {
   ProviderNotFoundException,
   ProviderRegistry,
   QdrantProvider,
+  QwenProvider,
   RateLimitException,
   RedisProvider,
   RetrievalException,
@@ -8056,10 +9802,13 @@ export {
   VectorPlugin,
   WeaviateProvider,
   createChatHandler,
+  createFeedbackHandler,
   createFromPreset,
   createHealthHandler,
+  createHistoryHandler,
   createIngestHandler,
   createRagHandler,
+  createSessionsHandler,
   createStreamHandler,
   createUploadHandler,
   getRagConfig,