@retrivora-ai/rag-engine 0.3.1 → 0.3.2

package/.env.example

@@ -20,7 +20,7 @@ VECTOR_DB_REST_API_KEY=your-rest-api-key
 # VECTOR_DB_UPSERT_PATH=/upsert
 
 # ─── LLM ──────────────────────────────────────────────────────
-# Choose: openai | anthropic | ollama | universal_rest
+# Choose: openai | anthropic | ollama | gemini | universal_rest
 LLM_PROVIDER=openai
 LLM_MODEL=gpt-4o
 LLM_MAX_TOKENS=1024
@@ -33,6 +33,9 @@ OPENAI_API_KEY=sk-...
 # Anthropic (if provider=anthropic)
 ANTHROPIC_API_KEY=sk-ant-...
 
+# Gemini (if provider=gemini)
+GEMINI_API_KEY=AIza...
+
 # Ollama (if provider=ollama)
 LLM_BASE_URL=http://localhost:11434
 
@@ -52,7 +55,7 @@ VECTOR_BASE_URL=http://localhost:6333
 VECTOR_DB_INDEX=my-index
 
 # ─── Embedding ────────────────────────────────────────────────
-# Choose: openai | ollama | custom
+# Choose: openai | ollama | gemini | custom
 EMBEDDING_PROVIDER=openai
 EMBEDDING_MODEL=text-embedding-3-small
 EMBEDDING_DIMENSIONS=1536

package/dist/{RagConfig-D3Inaf9N.d.mts → RagConfig-Crz-l02S.d.mts}

@@ -51,7 +51,7 @@ interface VectorDBConfig {
      */
     options: Record<string, unknown>;
 }
-type LLMProvider = 'openai' | 'anthropic' | 'ollama' | 'rest' | 'universal_rest' | 'custom';
+type LLMProvider = 'openai' | 'anthropic' | 'ollama' | 'gemini' | 'rest' | 'universal_rest' | 'custom';
 interface LLMConfig {
     /** Which LLM provider to use */
     provider: LLMProvider;
@@ -77,7 +77,7 @@ interface LLMConfig {
      */
     options?: Record<string, unknown>;
 }
-type EmbeddingProvider = 'openai' | 'ollama' | 'rest' | 'universal_rest' | 'custom';
+type EmbeddingProvider = 'openai' | 'ollama' | 'gemini' | 'rest' | 'universal_rest' | 'custom';
 interface EmbeddingConfig {
     /** Which embedding provider to use */
     provider: EmbeddingProvider;

package/dist/{RagConfig-D3Inaf9N.d.ts → RagConfig-Crz-l02S.d.ts}

@@ -51,7 +51,7 @@ interface VectorDBConfig {
      */
     options: Record<string, unknown>;
 }
-type LLMProvider = 'openai' | 'anthropic' | 'ollama' | 'rest' | 'universal_rest' | 'custom';
+type LLMProvider = 'openai' | 'anthropic' | 'ollama' | 'gemini' | 'rest' | 'universal_rest' | 'custom';
 interface LLMConfig {
     /** Which LLM provider to use */
     provider: LLMProvider;
@@ -77,7 +77,7 @@ interface LLMConfig {
      */
     options?: Record<string, unknown>;
 }
-type EmbeddingProvider = 'openai' | 'ollama' | 'rest' | 'universal_rest' | 'custom';
+type EmbeddingProvider = 'openai' | 'ollama' | 'gemini' | 'rest' | 'universal_rest' | 'custom';
 interface EmbeddingConfig {
     /** Which embedding provider to use */
     provider: EmbeddingProvider;

package/dist/{chunk-GT72OIOD.mjs → chunk-7BPQDPOC.mjs}

@@ -515,6 +515,16 @@ var ConfigValidator = class {
           });
         }
         break;
+      case "gemini":
+        if (!config.apiKey) {
+          errors.push({
+            field: "llm.apiKey",
+            message: "Gemini API key is required",
+            suggestion: "Set GEMINI_API_KEY environment variable",
+            severity: "error"
+          });
+        }
+        break;
       case "ollama":
         if (!config.baseUrl) {
           errors.push({
@@ -581,6 +591,14 @@ var ConfigValidator = class {
         severity: "error"
       });
     }
+    if (config.provider === "gemini" && !config.apiKey) {
+      errors.push({
+        field: "embedding.apiKey",
+        message: "Gemini API key is required for embedding",
+        suggestion: "Set GEMINI_API_KEY environment variable",
+        severity: "error"
+      });
+    }
     if (config.provider === "ollama" && !config.baseUrl) {
       errors.push({
         field: "embedding.baseUrl",
@@ -959,6 +977,101 @@ ${context}`;
   }
 };
 
+// src/llm/providers/GeminiProvider.ts
+import { GoogleGenAI } from "@google/genai";
+var GeminiProvider = class {
+  constructor(llmConfig, embeddingConfig) {
+    if (!llmConfig.apiKey) throw new Error("[GeminiProvider] llmConfig.apiKey is required");
+    this.client = new GoogleGenAI({ apiKey: llmConfig.apiKey });
+    this.llmConfig = llmConfig;
+    this.embeddingConfig = embeddingConfig;
+  }
+  async chat(messages, context, options) {
+    var _a, _b, _c, _d, _e, _f;
+    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:
+${context}`;
+    const geminiMessages = messages.map((m) => ({
+      role: m.role === "assistant" ? "model" : "user",
+      parts: [{ text: m.content }]
+    }));
+    const response = await this.client.models.generateContent({
+      model: this.llmConfig.model,
+      contents: geminiMessages,
+      config: {
+        systemInstruction,
+        temperature: (_c = (_b = options == null ? void 0 : options.temperature) != null ? _b : this.llmConfig.temperature) != null ? _c : 0.7,
+        maxOutputTokens: (_e = (_d = options == null ? void 0 : options.maxTokens) != null ? _d : this.llmConfig.maxTokens) != null ? _e : 1024,
+        stopSequences: options == null ? void 0 : options.stop
+      }
+    });
+    return (_f = response.text) != null ? _f : "";
+  }
+  async embed(text, options) {
+    var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j;
+    const model = (_c = (_b = options == null ? void 0 : options.model) != null ? _b : (_a = this.embeddingConfig) == null ? void 0 : _a.model) != null ? _c : "text-embedding-004";
+    const apiKey = (_e = (_d = this.embeddingConfig) == null ? void 0 : _d.apiKey) != null ? _e : this.llmConfig.apiKey;
+    const client = apiKey !== this.llmConfig.apiKey ? new GoogleGenAI({ apiKey }) : this.client;
+    let content = text;
+    const queryPrefix = (_f = this.embeddingConfig) == null ? void 0 : _f.queryPrefix;
+    const docPrefix = (_g = this.embeddingConfig) == null ? void 0 : _g.documentPrefix;
+    if ((options == null ? void 0 : options.taskType) === "query" && queryPrefix) {
+      if (!content.startsWith(queryPrefix)) {
+        content = `${queryPrefix}${text}`;
+      }
+    } else if ((options == null ? void 0 : options.taskType) === "document" && docPrefix) {
+      if (!content.startsWith(docPrefix)) {
+        content = `${docPrefix}${text}`;
+      }
+    }
+    const response = await client.models.embedContent({
+      model,
+      contents: content
+    });
+    return (_j = (_i = (_h = response.embeddings) == null ? void 0 : _h[0]) == null ? void 0 : _i.values) != null ? _j : [];
+  }
+  async batchEmbed(texts, options) {
+    var _a, _b, _c, _d, _e, _f, _g, _h, _i;
+    const vectors = [];
+    const model = (_c = (_b = options == null ? void 0 : options.model) != null ? _b : (_a = this.embeddingConfig) == null ? void 0 : _a.model) != null ? _c : "text-embedding-004";
+    const apiKey = (_e = (_d = this.embeddingConfig) == null ? void 0 : _d.apiKey) != null ? _e : this.llmConfig.apiKey;
+    const client = apiKey !== this.llmConfig.apiKey ? new GoogleGenAI({ apiKey }) : this.client;
+    let contents = texts;
+    const queryPrefix = (_f = this.embeddingConfig) == null ? void 0 : _f.queryPrefix;
+    const docPrefix = (_g = this.embeddingConfig) == null ? void 0 : _g.documentPrefix;
+    if ((options == null ? void 0 : options.taskType) === "query" && queryPrefix) {
+      contents = texts.map((text) => text.startsWith(queryPrefix) ? text : `${queryPrefix}${text}`);
+    } else if ((options == null ? void 0 : options.taskType) === "document" && docPrefix) {
+      contents = texts.map((text) => text.startsWith(docPrefix) ? text : `${docPrefix}${text}`);
+    }
+    const response = await client.models.embedContent({
+      model,
+      contents
+    });
+    return (_i = (_h = response.embeddings) == null ? void 0 : _h.map((e) => {
+      var _a2;
+      return (_a2 = e.values) != null ? _a2 : [];
+    })) != null ? _i : [];
+  }
+  async ping() {
+    try {
+      await this.client.models.embedContent({
+        model: "text-embedding-004",
+        contents: "ping"
+      });
+      return true;
+    } catch (err) {
+      console.error("[GeminiProvider] Ping failed:", err);
+      return false;
+    }
+  }
+};
+
 // src/llm/providers/UniversalLLMAdapter.ts
 import axios2 from "axios";
 
@@ -1155,6 +1268,8 @@ var LLMFactory = class _LLMFactory {
         return new AnthropicProvider(llmConfig, embeddingConfig);
       case "ollama":
         return new OllamaProvider(llmConfig, embeddingConfig);
+      case "gemini":
+        return new GeminiProvider(llmConfig, embeddingConfig);
       case "rest":
       case "universal_rest":
       case "custom":
@@ -1546,7 +1661,7 @@ var EmbeddingStrategyResolver = class {
    * Check if an LLM provider natively supports embeddings
    */
   static supportsEmbedding(provider) {
-    const providersWithEmbedding = ["openai", "ollama", "rest", "universal_rest"];
+    const providersWithEmbedding = ["openai", "ollama", "gemini", "rest", "universal_rest"];
     return providersWithEmbedding.includes(provider);
   }
   /**

package/dist/handlers/index.d.mts

@@ -1,3 +1,3 @@
-import '../RagConfig-D3Inaf9N.mjs';
-export { c as createChatHandler, a as createHealthHandler, b as createIngestHandler, d as createUploadHandler } from '../index-Ymwm-_OR.mjs';
+import '../RagConfig-Crz-l02S.mjs';
+export { c as createChatHandler, a as createHealthHandler, b as createIngestHandler, d as createUploadHandler } from '../index-Cd4CdmEi.mjs';
 import 'next/server';

package/dist/handlers/index.d.ts

@@ -1,3 +1,3 @@
-import '../RagConfig-D3Inaf9N.js';
-export { c as createChatHandler, a as createHealthHandler, b as createIngestHandler, d as createUploadHandler } from '../index-BhNJQ2SS.js';
+import '../RagConfig-Crz-l02S.js';
+export { c as createChatHandler, a as createHealthHandler, b as createIngestHandler, d as createUploadHandler } from '../index-DQ1pXmlK.js';
 import 'next/server';

package/dist/handlers/index.js

@@ -1683,6 +1683,16 @@ var ConfigValidator = class {
           });
         }
         break;
+      case "gemini":
+        if (!config.apiKey) {
+          errors.push({
+            field: "llm.apiKey",
+            message: "Gemini API key is required",
+            suggestion: "Set GEMINI_API_KEY environment variable",
+            severity: "error"
+          });
+        }
+        break;
       case "ollama":
         if (!config.baseUrl) {
           errors.push({
@@ -1749,6 +1759,14 @@ var ConfigValidator = class {
         severity: "error"
       });
     }
+    if (config.provider === "gemini" && !config.apiKey) {
+      errors.push({
+        field: "embedding.apiKey",
+        message: "Gemini API key is required for embedding",
+        suggestion: "Set GEMINI_API_KEY environment variable",
+        severity: "error"
+      });
+    }
     if (config.provider === "ollama" && !config.baseUrl) {
       errors.push({
         field: "embedding.baseUrl",
@@ -2127,6 +2145,101 @@ ${context}`;
   }
 };
 
+// src/llm/providers/GeminiProvider.ts
+var import_genai = require("@google/genai");
+var GeminiProvider = class {
+  constructor(llmConfig, embeddingConfig) {
+    if (!llmConfig.apiKey) throw new Error("[GeminiProvider] llmConfig.apiKey is required");
+    this.client = new import_genai.GoogleGenAI({ apiKey: llmConfig.apiKey });
+    this.llmConfig = llmConfig;
+    this.embeddingConfig = embeddingConfig;
+  }
+  async chat(messages, context, options) {
+    var _a, _b, _c, _d, _e, _f;
+    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:
+${context}`;
+    const geminiMessages = messages.map((m) => ({
+      role: m.role === "assistant" ? "model" : "user",
+      parts: [{ text: m.content }]
+    }));
+    const response = await this.client.models.generateContent({
+      model: this.llmConfig.model,
+      contents: geminiMessages,
+      config: {
+        systemInstruction,
+        temperature: (_c = (_b = options == null ? void 0 : options.temperature) != null ? _b : this.llmConfig.temperature) != null ? _c : 0.7,
+        maxOutputTokens: (_e = (_d = options == null ? void 0 : options.maxTokens) != null ? _d : this.llmConfig.maxTokens) != null ? _e : 1024,
+        stopSequences: options == null ? void 0 : options.stop
+      }
+    });
+    return (_f = response.text) != null ? _f : "";
+  }
+  async embed(text, options) {
+    var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j;
+    const model = (_c = (_b = options == null ? void 0 : options.model) != null ? _b : (_a = this.embeddingConfig) == null ? void 0 : _a.model) != null ? _c : "text-embedding-004";
+    const apiKey = (_e = (_d = this.embeddingConfig) == null ? void 0 : _d.apiKey) != null ? _e : this.llmConfig.apiKey;
+    const client = apiKey !== this.llmConfig.apiKey ? new import_genai.GoogleGenAI({ apiKey }) : this.client;
+    let content = text;
+    const queryPrefix = (_f = this.embeddingConfig) == null ? void 0 : _f.queryPrefix;
+    const docPrefix = (_g = this.embeddingConfig) == null ? void 0 : _g.documentPrefix;
+    if ((options == null ? void 0 : options.taskType) === "query" && queryPrefix) {
+      if (!content.startsWith(queryPrefix)) {
+        content = `${queryPrefix}${text}`;
+      }
+    } else if ((options == null ? void 0 : options.taskType) === "document" && docPrefix) {
+      if (!content.startsWith(docPrefix)) {
+        content = `${docPrefix}${text}`;
+      }
+    }
+    const response = await client.models.embedContent({
+      model,
+      contents: content
+    });
+    return (_j = (_i = (_h = response.embeddings) == null ? void 0 : _h[0]) == null ? void 0 : _i.values) != null ? _j : [];
+  }
+  async batchEmbed(texts, options) {
+    var _a, _b, _c, _d, _e, _f, _g, _h, _i;
+    const vectors = [];
+    const model = (_c = (_b = options == null ? void 0 : options.model) != null ? _b : (_a = this.embeddingConfig) == null ? void 0 : _a.model) != null ? _c : "text-embedding-004";
+    const apiKey = (_e = (_d = this.embeddingConfig) == null ? void 0 : _d.apiKey) != null ? _e : this.llmConfig.apiKey;
+    const client = apiKey !== this.llmConfig.apiKey ? new import_genai.GoogleGenAI({ apiKey }) : this.client;
+    let contents = texts;
+    const queryPrefix = (_f = this.embeddingConfig) == null ? void 0 : _f.queryPrefix;
+    const docPrefix = (_g = this.embeddingConfig) == null ? void 0 : _g.documentPrefix;
+    if ((options == null ? void 0 : options.taskType) === "query" && queryPrefix) {
+      contents = texts.map((text) => text.startsWith(queryPrefix) ? text : `${queryPrefix}${text}`);
+    } else if ((options == null ? void 0 : options.taskType) === "document" && docPrefix) {
+      contents = texts.map((text) => text.startsWith(docPrefix) ? text : `${docPrefix}${text}`);
+    }
+    const response = await client.models.embedContent({
+      model,
+      contents
+    });
+    return (_i = (_h = response.embeddings) == null ? void 0 : _h.map((e) => {
+      var _a2;
+      return (_a2 = e.values) != null ? _a2 : [];
+    })) != null ? _i : [];
+  }
+  async ping() {
+    try {
+      await this.client.models.embedContent({
+        model: "text-embedding-004",
+        contents: "ping"
+      });
+      return true;
+    } catch (err) {
+      console.error("[GeminiProvider] Ping failed:", err);
+      return false;
+    }
+  }
+};
+
 // src/llm/providers/UniversalLLMAdapter.ts
 var import_axios2 = __toESM(require("axios"));
 init_templateUtils();
@@ -2285,6 +2398,8 @@ var LLMFactory = class _LLMFactory {
         return new AnthropicProvider(llmConfig, embeddingConfig);
       case "ollama":
         return new OllamaProvider(llmConfig, embeddingConfig);
+      case "gemini":
+        return new GeminiProvider(llmConfig, embeddingConfig);
       case "rest":
       case "universal_rest":
       case "custom":
@@ -2670,7 +2785,7 @@ var EmbeddingStrategyResolver = class {
    * Check if an LLM provider natively supports embeddings
    */
   static supportsEmbedding(provider) {
-    const providersWithEmbedding = ["openai", "ollama", "rest", "universal_rest"];
+    const providersWithEmbedding = ["openai", "ollama", "gemini", "rest", "universal_rest"];
     return providersWithEmbedding.includes(provider);
   }
   /**

package/dist/handlers/index.mjs

@@ -3,7 +3,7 @@ import {
   createHealthHandler,
   createIngestHandler,
   createUploadHandler
-} from "../chunk-GT72OIOD.mjs";
+} from "../chunk-7BPQDPOC.mjs";
 import "../chunk-EDLTMSNY.mjs";
 import "../chunk-FWCSY2DS.mjs";
 export {

package/dist/{index-Ymwm-_OR.d.mts → index-Cd4CdmEi.d.mts}

@@ -1,4 +1,4 @@
-import { c as VectorDBConfig, L as LLMConfig, E as EmbeddingConfig, a as RagConfig, C as ChatResponse } from './RagConfig-D3Inaf9N.mjs';
+import { c as VectorDBConfig, L as LLMConfig, E as EmbeddingConfig, a as RagConfig, C as ChatResponse } from './RagConfig-Crz-l02S.mjs';
 import { NextRequest, NextResponse } from 'next/server';
 
 /**

package/dist/{index-BhNJQ2SS.d.ts → index-DQ1pXmlK.d.ts}

@@ -1,4 +1,4 @@
-import { c as VectorDBConfig, L as LLMConfig, E as EmbeddingConfig, a as RagConfig, C as ChatResponse } from './RagConfig-D3Inaf9N.js';
+import { c as VectorDBConfig, L as LLMConfig, E as EmbeddingConfig, a as RagConfig, C as ChatResponse } from './RagConfig-Crz-l02S.js';
 import { NextRequest, NextResponse } from 'next/server';
 
 /**

package/dist/index.d.mts

@@ -2,8 +2,8 @@ import * as react_jsx_runtime from 'react/jsx-runtime';
 import React$1, { ReactNode } from 'react';
 import { C as ChatMessage } from './DocumentChunker-BICIjSuG.mjs';
 export { a as ChatOptions, b as Chunk, c as ChunkOptions, E as EmbedOptions, I as ILLMProvider } from './DocumentChunker-BICIjSuG.mjs';
-import { V as VectorMatch, U as UIConfig } from './RagConfig-D3Inaf9N.mjs';
-export { C as ChatResponse, E as EmbeddingConfig, I as IngestDocument, L as LLMConfig, R as RAGConfig, a as RagConfig, b as UpsertDocument, c as VectorDBConfig } from './RagConfig-D3Inaf9N.mjs';
+import { V as VectorMatch, U as UIConfig } from './RagConfig-Crz-l02S.mjs';
+export { C as ChatResponse, E as EmbeddingConfig, I as IngestDocument, L as LLMConfig, R as RAGConfig, a as RagConfig, b as UpsertDocument, c as VectorDBConfig } from './RagConfig-Crz-l02S.mjs';
 
 interface ChatWidgetProps {
     /** Position of the floating button. Defaults to bottom-right. */

package/dist/index.d.ts

@@ -2,8 +2,8 @@ import * as react_jsx_runtime from 'react/jsx-runtime';
 import React$1, { ReactNode } from 'react';
 import { C as ChatMessage } from './DocumentChunker-BICIjSuG.js';
 export { a as ChatOptions, b as Chunk, c as ChunkOptions, E as EmbedOptions, I as ILLMProvider } from './DocumentChunker-BICIjSuG.js';
-import { V as VectorMatch, U as UIConfig } from './RagConfig-D3Inaf9N.js';
-export { C as ChatResponse, E as EmbeddingConfig, I as IngestDocument, L as LLMConfig, R as RAGConfig, a as RagConfig, b as UpsertDocument, c as VectorDBConfig } from './RagConfig-D3Inaf9N.js';
+import { V as VectorMatch, U as UIConfig } from './RagConfig-Crz-l02S.js';
+export { C as ChatResponse, E as EmbeddingConfig, I as IngestDocument, L as LLMConfig, R as RAGConfig, a as RagConfig, b as UpsertDocument, c as VectorDBConfig } from './RagConfig-Crz-l02S.js';
 
 interface ChatWidgetProps {
     /** Position of the floating button. Defaults to bottom-right. */

package/dist/server.d.mts

@@ -1,9 +1,9 @@
-import { a as RagConfig, C as ChatResponse, I as IngestDocument, c as VectorDBConfig, b as UpsertDocument, V as VectorMatch, L as LLMConfig, E as EmbeddingConfig, d as VectorDBProvider, e as LLMProvider, f as EmbeddingProvider } from './RagConfig-D3Inaf9N.mjs';
-export { R as RAGConfig, U as UIConfig } from './RagConfig-D3Inaf9N.mjs';
+import { a as RagConfig, C as ChatResponse, I as IngestDocument, c as VectorDBConfig, b as UpsertDocument, V as VectorMatch, L as LLMConfig, E as EmbeddingConfig, d as VectorDBProvider, e as LLMProvider, f as EmbeddingProvider } from './RagConfig-Crz-l02S.mjs';
+export { R as RAGConfig, U as UIConfig } from './RagConfig-Crz-l02S.mjs';
 import { C as ChatMessage, I as ILLMProvider, a as ChatOptions, E as EmbedOptions } from './DocumentChunker-BICIjSuG.mjs';
 export { b as Chunk, c as ChunkOptions, D as DocumentChunker } from './DocumentChunker-BICIjSuG.mjs';
-import { H as HealthCheckResult } from './index-Ymwm-_OR.mjs';
-export { P as ProviderHealthCheck, c as createChatHandler, a as createHealthHandler, b as createIngestHandler, d as createUploadHandler } from './index-Ymwm-_OR.mjs';
+import { H as HealthCheckResult } from './index-Cd4CdmEi.mjs';
+export { P as ProviderHealthCheck, c as createChatHandler, a as createHealthHandler, b as createIngestHandler, d as createUploadHandler } from './index-Cd4CdmEi.mjs';
 import 'next/server';
 
 /**

package/dist/server.d.ts

@@ -1,9 +1,9 @@
-import { a as RagConfig, C as ChatResponse, I as IngestDocument, c as VectorDBConfig, b as UpsertDocument, V as VectorMatch, L as LLMConfig, E as EmbeddingConfig, d as VectorDBProvider, e as LLMProvider, f as EmbeddingProvider } from './RagConfig-D3Inaf9N.js';
-export { R as RAGConfig, U as UIConfig } from './RagConfig-D3Inaf9N.js';
+import { a as RagConfig, C as ChatResponse, I as IngestDocument, c as VectorDBConfig, b as UpsertDocument, V as VectorMatch, L as LLMConfig, E as EmbeddingConfig, d as VectorDBProvider, e as LLMProvider, f as EmbeddingProvider } from './RagConfig-Crz-l02S.js';
+export { R as RAGConfig, U as UIConfig } from './RagConfig-Crz-l02S.js';
 import { C as ChatMessage, I as ILLMProvider, a as ChatOptions, E as EmbedOptions } from './DocumentChunker-BICIjSuG.js';
 export { b as Chunk, c as ChunkOptions, D as DocumentChunker } from './DocumentChunker-BICIjSuG.js';
-import { H as HealthCheckResult } from './index-BhNJQ2SS.js';
-export { P as ProviderHealthCheck, c as createChatHandler, a as createHealthHandler, b as createIngestHandler, d as createUploadHandler } from './index-BhNJQ2SS.js';
+import { H as HealthCheckResult } from './index-DQ1pXmlK.js';
+export { P as ProviderHealthCheck, c as createChatHandler, a as createHealthHandler, b as createIngestHandler, d as createUploadHandler } from './index-DQ1pXmlK.js';
 import 'next/server';
 
 /**

package/dist/server.js

@@ -1727,6 +1727,16 @@ var ConfigValidator = class {
           });
         }
         break;
+      case "gemini":
+        if (!config.apiKey) {
+          errors.push({
+            field: "llm.apiKey",
+            message: "Gemini API key is required",
+            suggestion: "Set GEMINI_API_KEY environment variable",
+            severity: "error"
+          });
+        }
+        break;
       case "ollama":
         if (!config.baseUrl) {
           errors.push({
@@ -1793,6 +1803,14 @@ var ConfigValidator = class {
         severity: "error"
       });
     }
+    if (config.provider === "gemini" && !config.apiKey) {
+      errors.push({
+        field: "embedding.apiKey",
+        message: "Gemini API key is required for embedding",
+        suggestion: "Set GEMINI_API_KEY environment variable",
+        severity: "error"
+      });
+    }
     if (config.provider === "ollama" && !config.baseUrl) {
       errors.push({
         field: "embedding.baseUrl",
@@ -2171,6 +2189,101 @@ ${context}`;
   }
 };
 
+// src/llm/providers/GeminiProvider.ts
+var import_genai = require("@google/genai");
+var GeminiProvider = class {
+  constructor(llmConfig, embeddingConfig) {
+    if (!llmConfig.apiKey) throw new Error("[GeminiProvider] llmConfig.apiKey is required");
+    this.client = new import_genai.GoogleGenAI({ apiKey: llmConfig.apiKey });
+    this.llmConfig = llmConfig;
+    this.embeddingConfig = embeddingConfig;
+  }
+  async chat(messages, context, options) {
+    var _a, _b, _c, _d, _e, _f;
+    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:
+${context}`;
+    const geminiMessages = messages.map((m) => ({
+      role: m.role === "assistant" ? "model" : "user",
+      parts: [{ text: m.content }]
+    }));
+    const response = await this.client.models.generateContent({
+      model: this.llmConfig.model,
+      contents: geminiMessages,
+      config: {
+        systemInstruction,
+        temperature: (_c = (_b = options == null ? void 0 : options.temperature) != null ? _b : this.llmConfig.temperature) != null ? _c : 0.7,
+        maxOutputTokens: (_e = (_d = options == null ? void 0 : options.maxTokens) != null ? _d : this.llmConfig.maxTokens) != null ? _e : 1024,
+        stopSequences: options == null ? void 0 : options.stop
+      }
+    });
+    return (_f = response.text) != null ? _f : "";
+  }
+  async embed(text, options) {
+    var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j;
+    const model = (_c = (_b = options == null ? void 0 : options.model) != null ? _b : (_a = this.embeddingConfig) == null ? void 0 : _a.model) != null ? _c : "text-embedding-004";
+    const apiKey = (_e = (_d = this.embeddingConfig) == null ? void 0 : _d.apiKey) != null ? _e : this.llmConfig.apiKey;
+    const client = apiKey !== this.llmConfig.apiKey ? new import_genai.GoogleGenAI({ apiKey }) : this.client;
+    let content = text;
+    const queryPrefix = (_f = this.embeddingConfig) == null ? void 0 : _f.queryPrefix;
+    const docPrefix = (_g = this.embeddingConfig) == null ? void 0 : _g.documentPrefix;
+    if ((options == null ? void 0 : options.taskType) === "query" && queryPrefix) {
+      if (!content.startsWith(queryPrefix)) {
+        content = `${queryPrefix}${text}`;
+      }
+    } else if ((options == null ? void 0 : options.taskType) === "document" && docPrefix) {
+      if (!content.startsWith(docPrefix)) {
+        content = `${docPrefix}${text}`;
+      }
+    }
+    const response = await client.models.embedContent({
+      model,
+      contents: content
+    });
+    return (_j = (_i = (_h = response.embeddings) == null ? void 0 : _h[0]) == null ? void 0 : _i.values) != null ? _j : [];
+  }
+  async batchEmbed(texts, options) {
+    var _a, _b, _c, _d, _e, _f, _g, _h, _i;
+    const vectors = [];
+    const model = (_c = (_b = options == null ? void 0 : options.model) != null ? _b : (_a = this.embeddingConfig) == null ? void 0 : _a.model) != null ? _c : "text-embedding-004";
+    const apiKey = (_e = (_d = this.embeddingConfig) == null ? void 0 : _d.apiKey) != null ? _e : this.llmConfig.apiKey;
+    const client = apiKey !== this.llmConfig.apiKey ? new import_genai.GoogleGenAI({ apiKey }) : this.client;
+    let contents = texts;
+    const queryPrefix = (_f = this.embeddingConfig) == null ? void 0 : _f.queryPrefix;
+    const docPrefix = (_g = this.embeddingConfig) == null ? void 0 : _g.documentPrefix;
+    if ((options == null ? void 0 : options.taskType) === "query" && queryPrefix) {
+      contents = texts.map((text) => text.startsWith(queryPrefix) ? text : `${queryPrefix}${text}`);
+    } else if ((options == null ? void 0 : options.taskType) === "document" && docPrefix) {
+      contents = texts.map((text) => text.startsWith(docPrefix) ? text : `${docPrefix}${text}`);
+    }
+    const response = await client.models.embedContent({
+      model,
+      contents
+    });
+    return (_i = (_h = response.embeddings) == null ? void 0 : _h.map((e) => {
+      var _a2;
+      return (_a2 = e.values) != null ? _a2 : [];
+    })) != null ? _i : [];
+  }
+  async ping() {
+    try {
+      await this.client.models.embedContent({
+        model: "text-embedding-004",
+        contents: "ping"
+      });
+      return true;
+    } catch (err) {
+      console.error("[GeminiProvider] Ping failed:", err);
+      return false;
+    }
+  }
+};
+
 // src/llm/providers/UniversalLLMAdapter.ts
 var import_axios2 = __toESM(require("axios"));
 init_templateUtils();
@@ -2368,6 +2481,8 @@ var LLMFactory = class _LLMFactory {
         return new AnthropicProvider(llmConfig, embeddingConfig);
       case "ollama":
         return new OllamaProvider(llmConfig, embeddingConfig);
+      case "gemini":
+        return new GeminiProvider(llmConfig, embeddingConfig);
       case "rest":
       case "universal_rest":
       case "custom":
@@ -2759,7 +2874,7 @@ var EmbeddingStrategyResolver = class {
    * Check if an LLM provider natively supports embeddings
    */
   static supportsEmbedding(provider) {
-    const providersWithEmbedding = ["openai", "ollama", "rest", "universal_rest"];
+    const providersWithEmbedding = ["openai", "ollama", "gemini", "rest", "universal_rest"];
     return providersWithEmbedding.includes(provider);
   }
   /**

package/dist/server.mjs

@@ -34,7 +34,7 @@ import {
   createIngestHandler,
   createUploadHandler,
   getRagConfig
-} from "./chunk-GT72OIOD.mjs";
+} from "./chunk-7BPQDPOC.mjs";
 import "./chunk-EDLTMSNY.mjs";
 import {
   PineconeProvider

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@retrivora-ai/rag-engine",
-  "version": "0.3.1",
+  "version": "0.3.2",
   "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",
@@ -85,6 +85,7 @@
   },
   "dependencies": {
     "@anthropic-ai/sdk": "^0.90.0",
+    "@google/genai": "^1.50.1",
     "@pinecone-database/pinecone": "^7.2.0",
     "axios": "^1.15.0",
     "lucide-react": "^1.8.0",

package/src/config/EmbeddingStrategy.ts

@@ -125,7 +125,7 @@ export class EmbeddingStrategyResolver {
    * Check if an LLM provider natively supports embeddings
    */
   private static supportsEmbedding(provider: LLMProvider | string): boolean {
-    const providersWithEmbedding = ['openai', 'ollama', 'rest', 'universal_rest'];
+    const providersWithEmbedding = ['openai', 'ollama', 'gemini', 'rest', 'universal_rest'];
     return providersWithEmbedding.includes(provider);
   }
 

package/src/config/RagConfig.ts

@@ -40,7 +40,7 @@ export interface VectorDBConfig {
 // LLM Provider
 // ---------------------------------------------------------------------------
 
-export type LLMProvider = 'openai' | 'anthropic' | 'ollama' | 'rest' | 'universal_rest' | 'custom';
+export type LLMProvider = 'openai' | 'anthropic' | 'ollama' | 'gemini' | 'rest' | 'universal_rest' | 'custom';
 
 export interface LLMConfig {
   /** Which LLM provider to use */
@@ -72,7 +72,7 @@ export interface LLMConfig {
 // Embedding
 // ---------------------------------------------------------------------------
 
-export type EmbeddingProvider = 'openai' | 'ollama' | 'rest' | 'universal_rest' | 'custom';
+export type EmbeddingProvider = 'openai' | 'ollama' | 'gemini' | 'rest' | 'universal_rest' | 'custom';
 
 export interface EmbeddingConfig {
   /** Which embedding provider to use */

package/src/core/ConfigValidator.ts

@@ -380,6 +380,16 @@ export class ConfigValidator {
           });
         }
         break;
+      case 'gemini':
+        if (!config.apiKey) {
+          errors.push({
+            field: 'llm.apiKey',
+            message: 'Gemini API key is required',
+            suggestion: 'Set GEMINI_API_KEY environment variable',
+            severity: 'error',
+          });
+        }
+        break;
       case 'ollama':
         if (!config.baseUrl) {
           errors.push({
@@ -454,6 +464,15 @@ export class ConfigValidator {
       });
     }
 
+    if (config.provider === 'gemini' && !config.apiKey) {
+      errors.push({
+        field: 'embedding.apiKey',
+        message: 'Gemini API key is required for embedding',
+        suggestion: 'Set GEMINI_API_KEY environment variable',
+        severity: 'error',
+      });
+    }
+
     if (config.provider === 'ollama' && !config.baseUrl) {
       errors.push({
         field: 'embedding.baseUrl',

package/src/llm/LLMFactory.ts

@@ -10,6 +10,7 @@ import { LLMConfig, EmbeddingConfig, LLMProvider } from '../config/RagConfig';
 import { OpenAIProvider } from './providers/OpenAIProvider';
 import { AnthropicProvider } from './providers/AnthropicProvider';
 import { OllamaProvider } from './providers/OllamaProvider';
+import { GeminiProvider } from './providers/GeminiProvider';
 import { UniversalLLMAdapter } from './providers/UniversalLLMAdapter';
 
 export class LLMFactory {
@@ -21,6 +22,8 @@ export class LLMFactory {
         return new AnthropicProvider(llmConfig, embeddingConfig);
       case 'ollama':
         return new OllamaProvider(llmConfig, embeddingConfig);
+      case 'gemini':
+        return new GeminiProvider(llmConfig, embeddingConfig);
       case 'rest':
       case 'universal_rest':
       case 'custom':

package/src/llm/providers/GeminiProvider.ts

@@ -0,0 +1,147 @@
+/**
+ * Gemini LLM Provider (Google Gen AI)
+ *
+ * Handles both chat completion (gemini-2.5-flash, gemini-2.5-pro, etc.) and
+ * embedding (text-embedding-004).
+ * Uses the modern @google/genai SDK.
+ *
+ * Required LLMConfig fields:
+ *   - apiKey: string  – Gemini API key
+ *   - model:  string  – e.g. "gemini-2.5-flash"
+ *
+ * Required EmbeddingConfig fields (when used for embedding):
+ *   - apiKey: string  – same or separate key
+ *   - model:  string  – e.g. "text-embedding-004"
+ */
+
+import { GoogleGenAI } from '@google/genai';
+import { ILLMProvider, ChatMessage, ChatOptions, EmbedOptions } from '../ILLMProvider';
+import { LLMConfig, EmbeddingConfig } from '../../config/RagConfig';
+
+export class GeminiProvider implements ILLMProvider {
+  private readonly client: GoogleGenAI;
+  private readonly llmConfig: LLMConfig;
+  private readonly embeddingConfig?: EmbeddingConfig;
+
+  constructor(llmConfig: LLMConfig, embeddingConfig?: EmbeddingConfig) {
+    if (!llmConfig.apiKey) throw new Error('[GeminiProvider] llmConfig.apiKey is required');
+    this.client = new GoogleGenAI({ apiKey: llmConfig.apiKey });
+    this.llmConfig = llmConfig;
+    this.embeddingConfig = embeddingConfig;
+  }
+
+  async chat(messages: ChatMessage[], context: string, options?: ChatOptions): Promise<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${context}`;
+
+    // Gemini requires alternating user/model messages, starting with user.
+    // If the first message is not user, or there are consecutive same-role messages, we need to adapt.
+    // To simplify and comply with typical usage, we'll map 'assistant' to 'model'.
+    const geminiMessages = messages.map((m) => ({
+      role: m.role === 'assistant' ? 'model' : 'user',
+      parts: [{ text: m.content }],
+    }));
+
+    const response = await this.client.models.generateContent({
+      model: this.llmConfig.model,
+      contents: geminiMessages,
+      config: {
+        systemInstruction,
+        temperature: options?.temperature ?? this.llmConfig.temperature ?? 0.7,
+        maxOutputTokens: options?.maxTokens ?? this.llmConfig.maxTokens ?? 1024,
+        stopSequences: options?.stop,
+      },
+    });
+
+    return response.text ?? '';
+  }
+
+  async embed(text: string, options?: EmbedOptions): Promise<number[]> {
+    const model =
+      options?.model ??
+      this.embeddingConfig?.model ??
+      'text-embedding-004';
+
+    const apiKey = this.embeddingConfig?.apiKey ?? this.llmConfig.apiKey;
+    const client = apiKey !== this.llmConfig.apiKey
+      ? new GoogleGenAI({ apiKey })
+      : this.client;
+
+    let content = text;
+    
+    // Dynamically apply prefixes from configuration if they exist
+    const queryPrefix = this.embeddingConfig?.queryPrefix;
+    const docPrefix = this.embeddingConfig?.documentPrefix;
+    
+    if (options?.taskType === 'query' && queryPrefix) {
+      if (!content.startsWith(queryPrefix)) {
+        content = `${queryPrefix}${text}`;
+      }
+    } else if (options?.taskType === 'document' && docPrefix) {
+      if (!content.startsWith(docPrefix)) {
+        content = `${docPrefix}${text}`;
+      }
+    }
+
+    const response = await client.models.embedContent({
+      model,
+      contents: content,
+    });
+
+    return response.embeddings?.[0]?.values ?? [];
+  }
+
+  async batchEmbed(texts: string[], options?: EmbedOptions): Promise<number[][]> {
+    const vectors: number[][] = [];
+    // Sequential fallback for Gemini (API doesn't have a direct batch endpoint in the same way, though you can pass multiple contents.
+    // For simplicity and to match the other providers' structure, we'll iterate or pass array.
+    // The new SDK supports an array of contents.
+    const model =
+      options?.model ??
+      this.embeddingConfig?.model ??
+      'text-embedding-004';
+
+    const apiKey = this.embeddingConfig?.apiKey ?? this.llmConfig.apiKey;
+    const client = apiKey !== this.llmConfig.apiKey
+      ? new GoogleGenAI({ apiKey })
+      : this.client;
+
+    let contents = texts;
+    
+    // Apply prefixes if needed
+    const queryPrefix = this.embeddingConfig?.queryPrefix;
+    const docPrefix = this.embeddingConfig?.documentPrefix;
+    
+    if (options?.taskType === 'query' && queryPrefix) {
+      contents = texts.map(text => text.startsWith(queryPrefix) ? text : `${queryPrefix}${text}`);
+    } else if (options?.taskType === 'document' && docPrefix) {
+      contents = texts.map(text => text.startsWith(docPrefix) ? text : `${docPrefix}${text}`);
+    }
+
+    const response = await client.models.embedContent({
+      model,
+      contents,
+    });
+
+    return response.embeddings?.map(e => e.values ?? []) ?? [];
+  }
+
+  async ping(): Promise<boolean> {
+    try {
+      // Perform a lightweight embed call to verify connectivity
+      await this.client.models.embedContent({
+        model: 'text-embedding-004',
+        contents: 'ping',
+      });
+      return true;
+    } catch (err) {
+      console.error('[GeminiProvider] Ping failed:', err);
+      return false;
+    }
+  }
+}