flutter_rag_chatbot #

A Flutter package designed to help developers easily build and integrate RAG-based chatbots into mobile apps. This package leverages Retrieval-Augmented Generation to retrieve relevant documents from a local vector store and then augment queries to a Language Model (LLM) like xAI's Grok API.

Features #

• Retrieval-Augmented Generation (RAG): Combines information retrieval with text generation.
• Local Vector Store: Basic in-memory vector store for document retrieval using cosine similarity.
• LLM Integration: Designed to integrate with LLMs (e.g., xAI Grok API).
• Chat UI Widget: A pre-built Flutter widget for displaying chat conversations.
• Offline Fallback: (Planned) Cache responses for offline access.
• Multi-turn History: Maintains conversation context.
• Markdown Parsing: Displays LLM responses formatted with Markdown.
• Extensibility: Hooks for custom retrievers and LLMs.
• Future-Proofing: Includes placeholders for image inputs, query logging, and hybrid edge/cloud modes.
• Multi-Content Support: Support for URLs, PDFs, images, and strings for embedding and retrieval.

Installation #

Add the following to your pubspec.yaml file:

dependencies:
  flutter_rag_chatbot:
    git:
      url: https://github.com/your-username/flutter_rag_chatbot.git # Replace with your repository URL
      ref: main # Or the specific branch/tag
  # Or, if published to pub.flutter-io.cn:
  # flutter_rag_chatbot: ^0.1.0

Then run flutter pub get.

Usage #

1. Initialize the RagChatbot #

First, initialize the RagChatbot with your API key and a knowledge base (list of documents).

import 'package:flutter/material.dart';
import 'package:flutter_rag_chatbot/flutter_rag_chatbot.dart';

void main() {
  runApp(const MyApp());
}

class MyApp extends StatefulWidget {
  const MyApp({super.key});

  @override
  State<MyApp> createState() => _MyAppState();
}

class _MyAppState extends State<MyApp> {
  late final RagChatbot _chatbot;
  final List<String> _knowledgeBase = [
    "The capital of France is Paris.",
    "Mount Everest is the highest mountain in the world.",
    "The Earth revolves around the Sun.",
    "Flutter is a UI toolkit for building natively compiled applications for mobile, web, and desktop from a single codebase.",
  ];

  @override
  void initState() {
    super.initState();
    _chatbot = RagChatbot();
    _initializeChatbot();
  }

  Future<void> _initializeChatbot() async {
    await _chatbot.initialize(
      apiKey: 'YOUR_AI_API_KEY', // Replace with your actual API key
      aiProvider: AIProvider.grok, // Choose from: grok, chatgpt, claude, gemini
      knowledgeBase: _knowledgeBase,
      urls: ['https://example.com'], // Optional: URLs to fetch and include
      pdfPaths: ['/path/to/document.pdf'], // Optional: PDF files to include
      imagePaths: ['/path/to/image.jpg'], // Optional: Images to include
    );
    setState(() {}); // Rebuild to show the chat widget after initialization
  }

  @override
  Widget build(BuildContext context) {
    return MaterialApp(
      title: 'RAG Chatbot Demo',
      theme: ThemeData(
        primarySwatch: Colors.blue,
      ),
      home: Scaffold(
        appBar: AppBar(
          title: const Text('RAG Chatbot Demo'),
        ),
        body: _chatbot == null
            ? const Center(child: CircularProgressIndicator())
            : RagChatWidget(chatbot: _chatbot),
      ),
    );
  }
}

2. Use the RagChatWidget #

Integrate the RagChatWidget into your Flutter application:

// See the example above for full integration.
// In your widget tree:
RagChatWidget(chatbot: _chatbot)

AI Provider Selection #

The package now supports multiple AI providers for text generation:

Supported AI Providers: #

• Grok (xAI): AIProvider.grok
• ChatGPT (OpenAI): AIProvider.chatgpt
• Claude (Anthropic): AIProvider.claude
• Gemini (Google): AIProvider.gemini

Usage Example: #

await chatbot.initialize(
  apiKey: 'your-api-key',
  aiProvider: AIProvider.chatgpt, // Choose your preferred AI provider
  knowledgeBase: ['Your knowledge documents'],
  // ... other content types
);

API Key Requirements: #

• Grok: xAI API key
• ChatGPT: OpenAI API key
• Claude: Anthropic API key
• Gemini: Google API key (may require different authentication)

Multi-Content Type Support #

The package now supports multiple content types for embedding and retrieval:

Supported Content Types: #

1. Strings: Direct text documents
2. URLs: Web pages (fetches and extracts text content)
3. PDFs: PDF documents (placeholder for text extraction)
4. Images: Image files (placeholder for image analysis/description)

Usage Example: #

await chatbot.initialize(
  apiKey: 'your-api-key',
  knowledgeBase: [
    'String document 1',
    'String document 2',
  ],
  urls: [
    'https://example.com',
    'https://flutter.cn',
  ],
  pdfPaths: [
    '/path/to/document.pdf',
    '/path/to/another.pdf',
  ],
  imagePaths: [
    '/path/to/image.jpg',
    '/path/to/photo.png',
  ],
);

Implementation Notes: #

• URLs: Automatically fetches web content and extracts text from HTML
• PDFs: Currently placeholder - requires PDF parsing library integration
• Images: Currently placeholder - requires image recognition integration
• Resource Management: Remember to call chatbot.dispose() when done to release resources

Contributing #

Contributions are welcome! Please feel free to open an issue or submit a pull request.

License #

This project is licensed under the MIT License - see the LICENSE file for details.