@retrivora-ai/rag-engine 1.1.0 → 1.1.1

package/dist/{QdrantProvider-WWXFX2XF.mjs → QdrantProvider-NMXOULCU.mjs}

@@ -1,6 +1,6 @@
 import {
   QdrantProvider
-} from "./chunk-65S4BQL2.mjs";
+} from "./chunk-63HITIWC.mjs";
 import "./chunk-IMP6FUCY.mjs";
 import "./chunk-X4TOT24V.mjs";
 export {

package/dist/{chunk-65S4BQL2.mjs → chunk-63HITIWC.mjs}

@@ -11,6 +11,7 @@ import crypto from "crypto";
 var QdrantProvider = class extends BaseVectorProvider {
   constructor(config) {
     super(config);
+    this.schemaDiscovered = false;
     const opts = config.options;
     const baseUrl = opts.baseUrl;
     if (!baseUrl) throw new Error("[QdrantProvider] baseUrl is required");
@@ -19,12 +20,15 @@ var QdrantProvider = class extends BaseVectorProvider {
     this.isFlatPayload = !!opts.flatPayload;
     this.http = axios.create({
       baseURL: baseUrl,
+      timeout: 15e3,
+      // 15s timeout for vector DB operations
       headers: __spreadValues({
         "Content-Type": "application/json"
       }, opts.apiKey ? { "api-key": opts.apiKey } : {})
     });
   }
   async initialize() {
+    if (this.schemaDiscovered) return;
     await this.ping();
     await this.ensureCollection();
     console.log(`[QdrantProvider] \u{1F50D} Discovering schema for collection "${this.indexName}"...`);
@@ -34,13 +38,16 @@ var QdrantProvider = class extends BaseVectorProvider {
     const allFields = [.../* @__PURE__ */ new Set([...discoveredFields, ...configFields])];
     if (allFields.length > 0) {
       console.log(`[QdrantProvider] \u{1F6E0} Ensuring indexes for ${allFields.length} discovered fields...`);
-      for (const fieldInfo of allFields) {
-        const [fieldName, schemaType] = fieldInfo.split(":");
-        await this.ensureIndex(fieldName, schemaType || "keyword");
-      }
+      await Promise.all(
+        allFields.map(async (fieldInfo) => {
+          const [fieldName, schemaType] = fieldInfo.split(":");
+          return this.ensureIndex(fieldName, schemaType || "keyword");
+        })
+      );
     } else {
       console.log(`[QdrantProvider] \u2139\uFE0F No fields discovered for indexing.`);
     }
+    this.schemaDiscovered = true;
   }
   /**
    * Samples points from the collection to discover available payload fields.

package/dist/{chunk-G2LVK46T.mjs → chunk-JZ4H7EP6.mjs}

@@ -333,6 +333,53 @@ ${context}`
     });
     return (_h = (_g = (_f = completion.choices[0]) == null ? void 0 : _f.message) == null ? void 0 : _g.content) != null ? _h : "";
   }
+  chatStream(messages, context, options) {
+    return __asyncGenerator(this, null, function* () {
+      var _a, _b, _c, _d, _e, _f, _g;
+      const systemContent = (_a = this.llmConfig.systemPrompt) != null ? _a : `You are a helpful assistant. Answer questions based on the provided context.
+
+Context:
+${context}`;
+      const systemMessage = {
+        role: "system",
+        content: systemContent.includes("{{context}}") ? systemContent.replace("{{context}}", context) : `${systemContent}
+
+Context:
+${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: (_c = (_b = options == null ? void 0 : options.maxTokens) != null ? _b : this.llmConfig.maxTokens) != null ? _c : 1024,
+        temperature: (_e = (_d = options == null ? void 0 : options.temperature) != null ? _d : this.llmConfig.temperature) != null ? _e : 0.7,
+        stop: options == null ? void 0 : options.stop,
+        stream: true
+      }));
+      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 = ((_g = (_f = chunk.choices[0]) == null ? void 0 : _f.delta) == null ? void 0 : _g.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];
@@ -438,6 +485,47 @@ ${context}`;
     const block = response.content[0];
     return block.type === "text" ? block.text : "";
   }
+  chatStream(messages, context, options) {
+    return __asyncGenerator(this, null, function* () {
+      var _a, _b, _c;
+      const systemPrompt = (_a = this.llmConfig.systemPrompt) != null ? _a : `You are a helpful assistant. Use the context below to answer the user's question accurately.
+
+Context:
+${context}`;
+      const system = systemPrompt.includes("{{context}}") ? systemPrompt.replace("{{context}}", context) : `${systemPrompt}
+
+Context:
+${context}`;
+      const anthropicMessages = messages.map((m) => ({
+        role: m.role === "assistant" ? "assistant" : "user",
+        content: m.content
+      }));
+      const stream = yield new __await(this.client.messages.create({
+        model: this.llmConfig.model,
+        max_tokens: (_c = (_b = options == null ? void 0 : options.maxTokens) != null ? _b : this.llmConfig.maxTokens) != null ? _c : 1024,
+        system,
+        messages: anthropicMessages,
+        stream: true
+      }));
+      try {
+        for (var iter = __forAwait(stream), more, temp, error; more = !(temp = yield new __await(iter.next())).done; more = false) {
+          const chunk = temp.value;
+          if (chunk.type === "content_block_delta" && chunk.delta.type === "text_delta") {
+            yield chunk.delta.text;
+          }
+        }
+      } 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;
@@ -543,7 +631,7 @@ var OllamaProvider = class {
   }
   chatStream(messages, context, options) {
     return __asyncGenerator(this, null, function* () {
-      var _a, _b, _c, _d, _e;
+      var _a, _b, _c, _d, _e, _f;
       const system = this.buildSystemPrompt(context);
       const response = yield new __await(this.http.post("/api/chat", {
         model: this.llmConfig.model,
@@ -557,11 +645,15 @@ var OllamaProvider = class {
           ...messages.map((m) => ({ role: m.role, content: m.content }))
         ]
       }, { responseType: "stream" }));
+      let lineBuffer = "";
       try {
         for (var iter = __forAwait(response.data), more, temp, error; more = !(temp = yield new __await(iter.next())).done; more = false) {
           const chunk = temp.value;
-          const lines = chunk.toString().split("\n").filter(Boolean);
+          lineBuffer += chunk.toString();
+          const lines = lineBuffer.split("\n");
+          lineBuffer = lines.pop() || "";
           for (const line of lines) {
+            if (!line.trim()) continue;
             try {
               const json = JSON.parse(line);
               if ((_e = json.message) == null ? void 0 : _e.content) {
@@ -569,6 +661,7 @@ var OllamaProvider = class {
               }
               if (json.done) return;
             } catch (e) {
+              console.warn("[OllamaProvider] Failed to parse streaming line:", line);
             }
           }
         }
@@ -582,6 +675,13 @@ var OllamaProvider = class {
             throw error[0];
         }
       }
+      if (lineBuffer.trim()) {
+        try {
+          const json = JSON.parse(lineBuffer);
+          if ((_f = json.message) == null ? void 0 : _f.content) yield json.message.content;
+        } catch (e) {
+        }
+      }
     });
   }
   buildSystemPrompt(context) {
@@ -727,6 +827,49 @@ ${context}`;
     });
     return (_f = response.text) != null ? _f : "";
   }
+  chatStream(messages, context, options) {
+    return __asyncGenerator(this, null, function* () {
+      var _a, _b, _c, _d, _e;
+      const systemPrompt = (_a = this.llmConfig.systemPrompt) != null ? _a : `You are a helpful assistant. Answer questions based on the provided context.
+
+Context:
+${context}`;
+      const systemInstruction = systemPrompt.includes("{{context}}") ? systemPrompt.replace("{{context}}", context) : `${systemPrompt}
+
+Context:
+${context}`;
+      const geminiMessages = messages.map((m) => ({
+        role: m.role === "assistant" ? "model" : "user",
+        parts: [{ text: m.content }]
+      }));
+      const result = yield new __await(this.client.models.generateContentStream({
+        model: this.llmConfig.model,
+        contents: geminiMessages,
+        config: {
+          systemInstruction,
+          temperature: (_c = (_b = options == null ? void 0 : options.temperature) != null ? _b : this.llmConfig.temperature) != null ? _c : 0.7,
+          maxOutputTokens: (_e = (_d = options == null ? void 0 : options.maxTokens) != null ? _d : this.llmConfig.maxTokens) != null ? _e : 1024,
+          stopSequences: options == null ? void 0 : options.stop
+        }
+      }));
+      try {
+        for (var iter = __forAwait(result.stream), more, temp, error; more = !(temp = yield new __await(iter.next())).done; more = false) {
+          const chunk = temp.value;
+          const text = chunk.text;
+          if (text) yield text;
+        }
+      } 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) {
     var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j, _k;
     const model = this.sanitizeModel(
@@ -1106,7 +1249,7 @@ var ProviderRegistry = class {
         return MilvusProvider;
       }
       case "qdrant": {
-        const { QdrantProvider } = await import("./QdrantProvider-WWXFX2XF.mjs");
+        const { QdrantProvider } = await import("./QdrantProvider-NMXOULCU.mjs");
         return QdrantProvider;
       }
       case "chromadb": {

package/dist/handlers/index.js

@@ -842,6 +842,7 @@ var init_QdrantProvider = __esm({
     QdrantProvider = class extends BaseVectorProvider {
       constructor(config) {
         super(config);
+        this.schemaDiscovered = false;
         const opts = config.options;
         const baseUrl = opts.baseUrl;
         if (!baseUrl) throw new Error("[QdrantProvider] baseUrl is required");
@@ -850,12 +851,15 @@ var init_QdrantProvider = __esm({
         this.isFlatPayload = !!opts.flatPayload;
         this.http = import_axios4.default.create({
           baseURL: baseUrl,
+          timeout: 15e3,
+          // 15s timeout for vector DB operations
           headers: __spreadValues({
             "Content-Type": "application/json"
           }, opts.apiKey ? { "api-key": opts.apiKey } : {})
         });
       }
       async initialize() {
+        if (this.schemaDiscovered) return;
         await this.ping();
         await this.ensureCollection();
         console.log(`[QdrantProvider] \u{1F50D} Discovering schema for collection "${this.indexName}"...`);
@@ -865,13 +869,16 @@ var init_QdrantProvider = __esm({
         const allFields = [.../* @__PURE__ */ new Set([...discoveredFields, ...configFields])];
         if (allFields.length > 0) {
           console.log(`[QdrantProvider] \u{1F6E0} Ensuring indexes for ${allFields.length} discovered fields...`);
-          for (const fieldInfo of allFields) {
-            const [fieldName, schemaType] = fieldInfo.split(":");
-            await this.ensureIndex(fieldName, schemaType || "keyword");
-          }
+          await Promise.all(
+            allFields.map(async (fieldInfo) => {
+              const [fieldName, schemaType] = fieldInfo.split(":");
+              return this.ensureIndex(fieldName, schemaType || "keyword");
+            })
+          );
         } else {
           console.log(`[QdrantProvider] \u2139\uFE0F No fields discovered for indexing.`);
         }
+        this.schemaDiscovered = true;
       }
       /**
        * Samples points from the collection to discover available payload fields.
@@ -1932,6 +1939,53 @@ ${context}`
     });
     return (_h = (_g = (_f = completion.choices[0]) == null ? void 0 : _f.message) == null ? void 0 : _g.content) != null ? _h : "";
   }
+  chatStream(messages, context, options) {
+    return __asyncGenerator(this, null, function* () {
+      var _a, _b, _c, _d, _e, _f, _g;
+      const systemContent = (_a = this.llmConfig.systemPrompt) != null ? _a : `You are a helpful assistant. Answer questions based on the provided context.
+
+Context:
+${context}`;
+      const systemMessage = {
+        role: "system",
+        content: systemContent.includes("{{context}}") ? systemContent.replace("{{context}}", context) : `${systemContent}
+
+Context:
+${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: (_c = (_b = options == null ? void 0 : options.maxTokens) != null ? _b : this.llmConfig.maxTokens) != null ? _c : 1024,
+        temperature: (_e = (_d = options == null ? void 0 : options.temperature) != null ? _d : this.llmConfig.temperature) != null ? _e : 0.7,
+        stop: options == null ? void 0 : options.stop,
+        stream: true
+      }));
+      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 = ((_g = (_f = chunk.choices[0]) == null ? void 0 : _f.delta) == null ? void 0 : _g.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];
@@ -2037,6 +2091,47 @@ ${context}`;
     const block = response.content[0];
     return block.type === "text" ? block.text : "";
   }
+  chatStream(messages, context, options) {
+    return __asyncGenerator(this, null, function* () {
+      var _a, _b, _c;
+      const systemPrompt = (_a = this.llmConfig.systemPrompt) != null ? _a : `You are a helpful assistant. Use the context below to answer the user's question accurately.
+
+Context:
+${context}`;
+      const system = systemPrompt.includes("{{context}}") ? systemPrompt.replace("{{context}}", context) : `${systemPrompt}
+
+Context:
+${context}`;
+      const anthropicMessages = messages.map((m) => ({
+        role: m.role === "assistant" ? "assistant" : "user",
+        content: m.content
+      }));
+      const stream = yield new __await(this.client.messages.create({
+        model: this.llmConfig.model,
+        max_tokens: (_c = (_b = options == null ? void 0 : options.maxTokens) != null ? _b : this.llmConfig.maxTokens) != null ? _c : 1024,
+        system,
+        messages: anthropicMessages,
+        stream: true
+      }));
+      try {
+        for (var iter = __forAwait(stream), more, temp, error; more = !(temp = yield new __await(iter.next())).done; more = false) {
+          const chunk = temp.value;
+          if (chunk.type === "content_block_delta" && chunk.delta.type === "text_delta") {
+            yield chunk.delta.text;
+          }
+        }
+      } 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;
@@ -2142,7 +2237,7 @@ var OllamaProvider = class {
   }
   chatStream(messages, context, options) {
     return __asyncGenerator(this, null, function* () {
-      var _a, _b, _c, _d, _e;
+      var _a, _b, _c, _d, _e, _f;
       const system = this.buildSystemPrompt(context);
       const response = yield new __await(this.http.post("/api/chat", {
         model: this.llmConfig.model,
@@ -2156,11 +2251,15 @@ var OllamaProvider = class {
           ...messages.map((m) => ({ role: m.role, content: m.content }))
         ]
       }, { responseType: "stream" }));
+      let lineBuffer = "";
       try {
         for (var iter = __forAwait(response.data), more, temp, error; more = !(temp = yield new __await(iter.next())).done; more = false) {
           const chunk = temp.value;
-          const lines = chunk.toString().split("\n").filter(Boolean);
+          lineBuffer += chunk.toString();
+          const lines = lineBuffer.split("\n");
+          lineBuffer = lines.pop() || "";
           for (const line of lines) {
+            if (!line.trim()) continue;
             try {
               const json = JSON.parse(line);
               if ((_e = json.message) == null ? void 0 : _e.content) {
@@ -2168,6 +2267,7 @@ var OllamaProvider = class {
               }
               if (json.done) return;
             } catch (e) {
+              console.warn("[OllamaProvider] Failed to parse streaming line:", line);
             }
           }
         }
@@ -2181,6 +2281,13 @@ var OllamaProvider = class {
             throw error[0];
         }
       }
+      if (lineBuffer.trim()) {
+        try {
+          const json = JSON.parse(lineBuffer);
+          if ((_f = json.message) == null ? void 0 : _f.content) yield json.message.content;
+        } catch (e) {
+        }
+      }
     });
   }
   buildSystemPrompt(context) {
@@ -2326,6 +2433,49 @@ ${context}`;
     });
     return (_f = response.text) != null ? _f : "";
   }
+  chatStream(messages, context, options) {
+    return __asyncGenerator(this, null, function* () {
+      var _a, _b, _c, _d, _e;
+      const systemPrompt = (_a = this.llmConfig.systemPrompt) != null ? _a : `You are a helpful assistant. Answer questions based on the provided context.
+
+Context:
+${context}`;
+      const systemInstruction = systemPrompt.includes("{{context}}") ? systemPrompt.replace("{{context}}", context) : `${systemPrompt}
+
+Context:
+${context}`;
+      const geminiMessages = messages.map((m) => ({
+        role: m.role === "assistant" ? "model" : "user",
+        parts: [{ text: m.content }]
+      }));
+      const result = yield new __await(this.client.models.generateContentStream({
+        model: this.llmConfig.model,
+        contents: geminiMessages,
+        config: {
+          systemInstruction,
+          temperature: (_c = (_b = options == null ? void 0 : options.temperature) != null ? _b : this.llmConfig.temperature) != null ? _c : 0.7,
+          maxOutputTokens: (_e = (_d = options == null ? void 0 : options.maxTokens) != null ? _d : this.llmConfig.maxTokens) != null ? _e : 1024,
+          stopSequences: options == null ? void 0 : options.stop
+        }
+      }));
+      try {
+        for (var iter = __forAwait(result.stream), more, temp, error; more = !(temp = yield new __await(iter.next())).done; more = false) {
+          const chunk = temp.value;
+          const text = chunk.text;
+          if (text) yield text;
+        }
+      } 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) {
     var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j, _k;
     const model = this.sanitizeModel(

package/dist/handlers/index.mjs

@@ -8,7 +8,7 @@ import {
   sseFrame,
   sseMetaFrame,
   sseTextFrame
-} from "../chunk-G2LVK46T.mjs";
+} from "../chunk-JZ4H7EP6.mjs";
 import "../chunk-YLTMFW4M.mjs";
 import "../chunk-X4TOT24V.mjs";
 export {

package/dist/server.d.mts

@@ -738,6 +738,7 @@ declare class QdrantProvider extends BaseVectorProvider {
     private contentField;
     private metadataField;
     private isFlatPayload;
+    private schemaDiscovered;
     constructor(config: VectorDBConfig);
     initialize(): Promise<void>;
     /**
@@ -892,6 +893,7 @@ declare class OpenAIProvider implements ILLMProvider {
     static getValidator(): IProviderValidator;
     static getHealthChecker(): IProviderHealthChecker;
     chat(messages: ChatMessage[], context: string, options?: ChatOptions): Promise<string>;
+    chatStream(messages: ChatMessage[], context: string, options?: ChatOptions): AsyncIterable<string>;
     embed(text: string, options?: EmbedOptions): Promise<number[]>;
     batchEmbed(texts: string[], options?: EmbedOptions): Promise<number[][]>;
     ping(): Promise<boolean>;
@@ -909,6 +911,7 @@ declare class AnthropicProvider implements ILLMProvider {
     static getValidator(): IProviderValidator;
     static getHealthChecker(): IProviderHealthChecker;
     chat(messages: ChatMessage[], context: string, options?: ChatOptions): Promise<string>;
+    chatStream(messages: ChatMessage[], context: string, options?: ChatOptions): AsyncIterable<string>;
     embed(text: string, options?: EmbedOptions): Promise<number[]>;
     batchEmbed(texts: string[], options?: EmbedOptions): Promise<number[][]>;
     ping(): Promise<boolean>;

package/dist/server.d.ts

@@ -738,6 +738,7 @@ declare class QdrantProvider extends BaseVectorProvider {
     private contentField;
     private metadataField;
     private isFlatPayload;
+    private schemaDiscovered;
     constructor(config: VectorDBConfig);
     initialize(): Promise<void>;
     /**
@@ -892,6 +893,7 @@ declare class OpenAIProvider implements ILLMProvider {
     static getValidator(): IProviderValidator;
     static getHealthChecker(): IProviderHealthChecker;
     chat(messages: ChatMessage[], context: string, options?: ChatOptions): Promise<string>;
+    chatStream(messages: ChatMessage[], context: string, options?: ChatOptions): AsyncIterable<string>;
     embed(text: string, options?: EmbedOptions): Promise<number[]>;
     batchEmbed(texts: string[], options?: EmbedOptions): Promise<number[][]>;
     ping(): Promise<boolean>;
@@ -909,6 +911,7 @@ declare class AnthropicProvider implements ILLMProvider {
     static getValidator(): IProviderValidator;
     static getHealthChecker(): IProviderHealthChecker;
     chat(messages: ChatMessage[], context: string, options?: ChatOptions): Promise<string>;
+    chatStream(messages: ChatMessage[], context: string, options?: ChatOptions): AsyncIterable<string>;
     embed(text: string, options?: EmbedOptions): Promise<number[]>;
     batchEmbed(texts: string[], options?: EmbedOptions): Promise<number[][]>;
     ping(): Promise<boolean>;

package/dist/server.js

@@ -854,6 +854,7 @@ var init_QdrantProvider = __esm({
     QdrantProvider = class extends BaseVectorProvider {
       constructor(config) {
         super(config);
+        this.schemaDiscovered = false;
         const opts = config.options;
         const baseUrl = opts.baseUrl;
         if (!baseUrl) throw new Error("[QdrantProvider] baseUrl is required");
@@ -862,12 +863,15 @@ var init_QdrantProvider = __esm({
         this.isFlatPayload = !!opts.flatPayload;
         this.http = import_axios4.default.create({
           baseURL: baseUrl,
+          timeout: 15e3,
+          // 15s timeout for vector DB operations
           headers: __spreadValues({
             "Content-Type": "application/json"
           }, opts.apiKey ? { "api-key": opts.apiKey } : {})
         });
       }
       async initialize() {
+        if (this.schemaDiscovered) return;
         await this.ping();
         await this.ensureCollection();
         console.log(`[QdrantProvider] \u{1F50D} Discovering schema for collection "${this.indexName}"...`);
@@ -877,13 +881,16 @@ var init_QdrantProvider = __esm({
         const allFields = [.../* @__PURE__ */ new Set([...discoveredFields, ...configFields])];
         if (allFields.length > 0) {
           console.log(`[QdrantProvider] \u{1F6E0} Ensuring indexes for ${allFields.length} discovered fields...`);
-          for (const fieldInfo of allFields) {
-            const [fieldName, schemaType] = fieldInfo.split(":");
-            await this.ensureIndex(fieldName, schemaType || "keyword");
-          }
+          await Promise.all(
+            allFields.map(async (fieldInfo) => {
+              const [fieldName, schemaType] = fieldInfo.split(":");
+              return this.ensureIndex(fieldName, schemaType || "keyword");
+            })
+          );
         } else {
           console.log(`[QdrantProvider] \u2139\uFE0F No fields discovered for indexing.`);
         }
+        this.schemaDiscovered = true;
       }
       /**
        * Samples points from the collection to discover available payload fields.
@@ -1980,6 +1987,53 @@ ${context}`
     });
     return (_h = (_g = (_f = completion.choices[0]) == null ? void 0 : _f.message) == null ? void 0 : _g.content) != null ? _h : "";
   }
+  chatStream(messages, context, options) {
+    return __asyncGenerator(this, null, function* () {
+      var _a, _b, _c, _d, _e, _f, _g;
+      const systemContent = (_a = this.llmConfig.systemPrompt) != null ? _a : `You are a helpful assistant. Answer questions based on the provided context.
+
+Context:
+${context}`;
+      const systemMessage = {
+        role: "system",
+        content: systemContent.includes("{{context}}") ? systemContent.replace("{{context}}", context) : `${systemContent}
+
+Context:
+${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: (_c = (_b = options == null ? void 0 : options.maxTokens) != null ? _b : this.llmConfig.maxTokens) != null ? _c : 1024,
+        temperature: (_e = (_d = options == null ? void 0 : options.temperature) != null ? _d : this.llmConfig.temperature) != null ? _e : 0.7,
+        stop: options == null ? void 0 : options.stop,
+        stream: true
+      }));
+      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 = ((_g = (_f = chunk.choices[0]) == null ? void 0 : _f.delta) == null ? void 0 : _g.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];
@@ -2085,6 +2139,47 @@ ${context}`;
     const block = response.content[0];
     return block.type === "text" ? block.text : "";
   }
+  chatStream(messages, context, options) {
+    return __asyncGenerator(this, null, function* () {
+      var _a, _b, _c;
+      const systemPrompt = (_a = this.llmConfig.systemPrompt) != null ? _a : `You are a helpful assistant. Use the context below to answer the user's question accurately.
+
+Context:
+${context}`;
+      const system = systemPrompt.includes("{{context}}") ? systemPrompt.replace("{{context}}", context) : `${systemPrompt}
+
+Context:
+${context}`;
+      const anthropicMessages = messages.map((m) => ({
+        role: m.role === "assistant" ? "assistant" : "user",
+        content: m.content
+      }));
+      const stream = yield new __await(this.client.messages.create({
+        model: this.llmConfig.model,
+        max_tokens: (_c = (_b = options == null ? void 0 : options.maxTokens) != null ? _b : this.llmConfig.maxTokens) != null ? _c : 1024,
+        system,
+        messages: anthropicMessages,
+        stream: true
+      }));
+      try {
+        for (var iter = __forAwait(stream), more, temp, error; more = !(temp = yield new __await(iter.next())).done; more = false) {
+          const chunk = temp.value;
+          if (chunk.type === "content_block_delta" && chunk.delta.type === "text_delta") {
+            yield chunk.delta.text;
+          }
+        }
+      } 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;
@@ -2190,7 +2285,7 @@ var OllamaProvider = class {
   }
   chatStream(messages, context, options) {
     return __asyncGenerator(this, null, function* () {
-      var _a, _b, _c, _d, _e;
+      var _a, _b, _c, _d, _e, _f;
       const system = this.buildSystemPrompt(context);
       const response = yield new __await(this.http.post("/api/chat", {
         model: this.llmConfig.model,
@@ -2204,11 +2299,15 @@ var OllamaProvider = class {
           ...messages.map((m) => ({ role: m.role, content: m.content }))
         ]
       }, { responseType: "stream" }));
+      let lineBuffer = "";
       try {
         for (var iter = __forAwait(response.data), more, temp, error; more = !(temp = yield new __await(iter.next())).done; more = false) {
           const chunk = temp.value;
-          const lines = chunk.toString().split("\n").filter(Boolean);
+          lineBuffer += chunk.toString();
+          const lines = lineBuffer.split("\n");
+          lineBuffer = lines.pop() || "";
           for (const line of lines) {
+            if (!line.trim()) continue;
             try {
               const json = JSON.parse(line);
               if ((_e = json.message) == null ? void 0 : _e.content) {
@@ -2216,6 +2315,7 @@ var OllamaProvider = class {
               }
               if (json.done) return;
             } catch (e) {
+              console.warn("[OllamaProvider] Failed to parse streaming line:", line);
             }
           }
         }
@@ -2229,6 +2329,13 @@ var OllamaProvider = class {
             throw error[0];
         }
       }
+      if (lineBuffer.trim()) {
+        try {
+          const json = JSON.parse(lineBuffer);
+          if ((_f = json.message) == null ? void 0 : _f.content) yield json.message.content;
+        } catch (e) {
+        }
+      }
     });
   }
   buildSystemPrompt(context) {
@@ -2374,6 +2481,49 @@ ${context}`;
     });
     return (_f = response.text) != null ? _f : "";
   }
+  chatStream(messages, context, options) {
+    return __asyncGenerator(this, null, function* () {
+      var _a, _b, _c, _d, _e;
+      const systemPrompt = (_a = this.llmConfig.systemPrompt) != null ? _a : `You are a helpful assistant. Answer questions based on the provided context.
+
+Context:
+${context}`;
+      const systemInstruction = systemPrompt.includes("{{context}}") ? systemPrompt.replace("{{context}}", context) : `${systemPrompt}
+
+Context:
+${context}`;
+      const geminiMessages = messages.map((m) => ({
+        role: m.role === "assistant" ? "model" : "user",
+        parts: [{ text: m.content }]
+      }));
+      const result = yield new __await(this.client.models.generateContentStream({
+        model: this.llmConfig.model,
+        contents: geminiMessages,
+        config: {
+          systemInstruction,
+          temperature: (_c = (_b = options == null ? void 0 : options.temperature) != null ? _b : this.llmConfig.temperature) != null ? _c : 0.7,
+          maxOutputTokens: (_e = (_d = options == null ? void 0 : options.maxTokens) != null ? _d : this.llmConfig.maxTokens) != null ? _e : 1024,
+          stopSequences: options == null ? void 0 : options.stop
+        }
+      }));
+      try {
+        for (var iter = __forAwait(result.stream), more, temp, error; more = !(temp = yield new __await(iter.next())).done; more = false) {
+          const chunk = temp.value;
+          const text = chunk.text;
+          if (text) yield text;
+        }
+      } 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) {
     var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j, _k;
     const model = this.sanitizeModel(

package/dist/server.mjs

@@ -40,7 +40,7 @@ import {
   sseFrame,
   sseMetaFrame,
   sseTextFrame
-} from "./chunk-G2LVK46T.mjs";
+} from "./chunk-JZ4H7EP6.mjs";
 import "./chunk-YLTMFW4M.mjs";
 import {
   PineconeProvider
@@ -56,7 +56,7 @@ import {
 } from "./chunk-U55XRW3U.mjs";
 import {
   QdrantProvider
-} from "./chunk-65S4BQL2.mjs";
+} from "./chunk-63HITIWC.mjs";
 import {
   BaseVectorProvider
 } from "./chunk-IMP6FUCY.mjs";

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@retrivora-ai/rag-engine",
-  "version": "1.1.0",
+  "version": "1.1.1",
   "description": "Retrivora AI is a plug-and-play AI engine for RAG chat experiences — generic vector DB + LLM provider, embeddable or standalone.",
   "author": "Abhinav Alkuchi",
   "license": "MIT",

package/src/llm/providers/AnthropicProvider.ts

@@ -98,6 +98,35 @@ export class AnthropicProvider implements ILLMProvider {
     return block.type === 'text' ? block.text : '';
   }
 
+  async *chatStream(messages: ChatMessage[], context: string, options?: ChatOptions): AsyncIterable<string> {
+    const systemPrompt =
+      this.llmConfig.systemPrompt ??
+      `You are a helpful assistant. Use the context below to answer the user's question accurately.\n\nContext:\n${context}`;
+
+    const system = systemPrompt.includes('{{context}}')
+      ? systemPrompt.replace('{{context}}', context)
+      : `${systemPrompt}\n\nContext:\n${context}`;
+
+    const anthropicMessages: Anthropic.MessageParam[] = messages.map((m) => ({
+      role: m.role === 'assistant' ? 'assistant' : 'user',
+      content: m.content,
+    }));
+
+    const stream = await this.client.messages.create({
+      model: this.llmConfig.model,
+      max_tokens: options?.maxTokens ?? this.llmConfig.maxTokens ?? 1024,
+      system,
+      messages: anthropicMessages,
+      stream: true,
+    });
+
+    for await (const chunk of stream) {
+      if (chunk.type === 'content_block_delta' && chunk.delta.type === 'text_delta') {
+        yield chunk.delta.text;
+      }
+    }
+  }
+
   async embed(text: string, options?: EmbedOptions): Promise<number[]> {
     void text;
     void options;

package/src/llm/providers/GeminiProvider.ts

@@ -130,6 +130,37 @@ export class GeminiProvider implements ILLMProvider {
     return response.text ?? '';
   }
 
+  async *chatStream(messages: ChatMessage[], context: string, options?: ChatOptions): AsyncIterable<string> {
+    const systemPrompt =
+      this.llmConfig.systemPrompt ??
+      `You are a helpful assistant. Answer questions based on the provided context.\n\nContext:\n${context}`;
+
+    const systemInstruction = systemPrompt.includes('{{context}}')
+      ? systemPrompt.replace('{{context}}', context)
+      : `${systemPrompt}\n\nContext:\n${context}`;
+
+    const geminiMessages = messages.map((m) => ({
+      role: m.role === 'assistant' ? 'model' : 'user',
+      parts: [{ text: m.content }],
+    }));
+
+    const result = await this.client.models.generateContentStream({
+      model: this.llmConfig.model,
+      contents: geminiMessages,
+      config: {
+        systemInstruction,
+        temperature: options?.temperature ?? this.llmConfig.temperature ?? 0.7,
+        maxOutputTokens: options?.maxTokens ?? this.llmConfig.maxTokens ?? 1024,
+        stopSequences: options?.stop,
+      },
+    });
+
+    for await (const chunk of result.stream) {
+      const text = chunk.text;
+      if (text) yield text;
+    }
+  }
+
   async embed(text: string, options?: EmbedOptions): Promise<number[]> {
     const model = this.sanitizeModel(
       options?.model ??

package/src/llm/providers/OllamaProvider.ts

@@ -118,9 +118,16 @@ export class OllamaProvider implements ILLMProvider {
       ],
     }, { responseType: 'stream' });
 
+    let lineBuffer = '';
     for await (const chunk of response.data) {
-      const lines = chunk.toString().split('\n').filter(Boolean);
+      lineBuffer += chunk.toString();
+      const lines = lineBuffer.split('\n');
+      
+      // Keep the last partial line in the buffer
+      lineBuffer = lines.pop() || '';
+
       for (const line of lines) {
+        if (!line.trim()) continue;
         try {
           const json = JSON.parse(line);
           if (json.message?.content) {
@@ -128,10 +135,22 @@ export class OllamaProvider implements ILLMProvider {
           }
           if (json.done) return;
         } catch {
-          // Ignore parsing errors for partial lines
+          // If parsing fails, it might be a malformed line or something else
+          // But with line buffering, this should be rare
+          console.warn('[OllamaProvider] Failed to parse streaming line:', line);
         }
       }
     }
+
+    // Process anything left in the buffer at the end
+    if (lineBuffer.trim()) {
+      try {
+        const json = JSON.parse(lineBuffer);
+        if (json.message?.content) yield json.message.content;
+      } catch {
+        // Final cleanup
+      }
+    }
   }
 
   private buildSystemPrompt(context: string): string {

package/src/llm/providers/OpenAIProvider.ts

@@ -108,6 +108,41 @@ export class OpenAIProvider implements ILLMProvider {
 
     return completion.choices[0]?.message?.content ?? '';
   }
+  
+  async *chatStream(messages: ChatMessage[], context: string, options?: ChatOptions): AsyncIterable<string> {
+    const systemContent =
+      this.llmConfig.systemPrompt ??
+      `You are a helpful assistant. Answer questions based on the provided context.\n\nContext:\n${context}`;
+
+    const systemMessage: OpenAI.Chat.ChatCompletionSystemMessageParam = {
+      role: 'system',
+      content: systemContent.includes('{{context}}')
+        ? systemContent.replace('{{context}}', context)
+        : `${systemContent}\n\nContext:\n${context}`,
+    };
+
+    const formattedMessages: OpenAI.Chat.ChatCompletionMessageParam[] = [
+      systemMessage,
+      ...messages.map((m) => ({
+        role: m.role as 'user' | 'assistant',
+        content: m.content,
+      })),
+    ];
+
+    const stream = await this.client.chat.completions.create({
+      model: this.llmConfig.model,
+      messages: formattedMessages,
+      max_tokens: options?.maxTokens ?? this.llmConfig.maxTokens ?? 1024,
+      temperature: options?.temperature ?? this.llmConfig.temperature ?? 0.7,
+      stop: options?.stop,
+      stream: true,
+    });
+
+    for await (const chunk of stream) {
+      const content = chunk.choices[0]?.delta?.content || '';
+      if (content) yield content;
+    }
+  }
 
   async embed(text: string, options?: EmbedOptions): Promise<number[]> {
     const results = await this.batchEmbed([text], options);

package/src/providers/vectordb/QdrantProvider.ts

@@ -13,6 +13,7 @@ export class QdrantProvider extends BaseVectorProvider {
   private contentField: string;
   private metadataField: string;
   private isFlatPayload: boolean;
+  private schemaDiscovered = false;
 
   constructor(config: VectorDBConfig) {
     super(config);
@@ -26,6 +27,7 @@ export class QdrantProvider extends BaseVectorProvider {
 
     this.http = axios.create({
       baseURL: baseUrl,
+      timeout: 15000, // 15s timeout for vector DB operations
       headers: {
         'Content-Type': 'application/json',
         ...(opts.apiKey ? { 'api-key': opts.apiKey as string } : {}),
@@ -34,6 +36,8 @@ export class QdrantProvider extends BaseVectorProvider {
   }
 
   async initialize(): Promise<void> {
+    if (this.schemaDiscovered) return;
+    
     await this.ping();
     await this.ensureCollection();
 
@@ -49,13 +53,18 @@ export class QdrantProvider extends BaseVectorProvider {
 
     if (allFields.length > 0) {
       console.log(`[QdrantProvider] 🛠 Ensuring indexes for ${allFields.length} discovered fields...`);
-      for (const fieldInfo of allFields) {
-        const [fieldName, schemaType] = fieldInfo.split(':');
-        await this.ensureIndex(fieldName, (schemaType as any) || 'keyword');
-      }
+      // Parallelize index creation for much faster startup
+      await Promise.all(
+        allFields.map(async (fieldInfo) => {
+          const [fieldName, schemaType] = fieldInfo.split(':');
+          return this.ensureIndex(fieldName, (schemaType as any) || 'keyword');
+        })
+      );
     } else {
       console.log(`[QdrantProvider] ℹ️ No fields discovered for indexing.`);
     }
+
+    this.schemaDiscovered = true;
   }
 
   /**