https://github.com/shivammishra1603/rag-bot

RAG Bot is a Retrieval-Augmented Generation (RAG) chatbot built with Streamlit, FAISS, HuggingFace embeddings, and Google Gemini. It lets you upload PDFs, process them into a vector store, and interact with your documents in a natural chat interface with conversational memory and real-time processing.
https://github.com/shivammishra1603/rag-bot

chatbot faiss google-gemini huggingface langchain rag streamlit vector-database

Last synced: 2 months ago
JSON representation

RAG Bot is a Retrieval-Augmented Generation (RAG) chatbot built with Streamlit, FAISS, HuggingFace embeddings, and Google Gemini. It lets you upload PDFs, process them into a vector store, and interact with your documents in a natural chat interface with conversational memory and real-time processing.

• Host: GitHub
• URL: https://github.com/shivammishra1603/rag-bot
• Owner: ShivamMishra1603
• Created: 2025-08-26T01:17:08.000Z (10 months ago)
• Default Branch: main
• Last Pushed: 2025-08-27T00:41:11.000Z (10 months ago)
• Last Synced: 2025-08-27T09:21:13.202Z (10 months ago)
• Topics: chatbot, faiss, google-gemini, huggingface, langchain, rag, streamlit, vector-database
• Language: Python
• Homepage:
• Size: 920 KB
• Stars: 0
• Watchers: 0
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# RAG Bot

Retrieval-Augmented Generation (RAG) chatbot that lets you upload PDFs, index them locally with FAISS, and chat using Google’s Gemini models — all wrapped in a clean Streamlit UI.

## Demo

![Demo](public/videos/Demo.gif)

## Features

- **Document Upload & Processing**: Upload multiple PDF files for AI analysis

- **RAG-Powered Responses**: AI answers questions based on uploaded document content

- **Conversational Memory**: Maintains chat history for contextual conversations

- **Vector Store Persistence**: Saves processed documents for future sessions

- **Modern UI**: Clean, WhatsApp-inspired chat interface

- **Real-time Processing**: Live document processing with progress indicators

## Installation

### Prerequisites

- Python 3.8+

- Google API key for Gemini model

### Setup Steps

1. **Clone the repository**

   ```bash

   git clone https://github.com/ShivamMishra1603/rag-bot

   cd rag-bot

   ```

2. **Install dependencies**

   ```bash

   pip install -r requirements.txt

   ```

3. **Environment Configuration**

   Create a `.env` file in the root directory:

   ```env

   GOOGLE_API_KEY=your_google_api_key_here

   ```

4. **Run the application**

   ```bash

   streamlit run app.py

## Usage

### Getting Started

1. **Launch the Application**

   ```bash

   streamlit run app.py

   ```

2. **Upload Documents**

   - Use the sidebar to upload one or more PDF files

   - Click "Process Documents" to create the vector store

3. **Start Chatting**

   - Ask questions about your uploaded documents

   - The AI will provide answers based on the document content

### System Status Indicators

The sidebar shows real-time system status:

- **Vector Store**: Shows if documents are loaded and ready

- **RAG Chain**: Indicates if the conversational system is active

- **Controls**: Clear chat history and reset memory

## File Structure

```

rag-bot/

├── app.py                 # Main Streamlit application

├── requirements.txt       # Python dependencies

├── .env                   # Environment variables (create this)

├── src/                   # Core application modules

│   ├── chain.py           # RAG chain implementation

│   ├── embeddings.py      # Embedding management

│   ├── loaders.py         # Document processing

│   └── vectorstore.py     # Vector database operations

├── vectorstore/           # Persistent vector storage (auto-created)

│   ├── faiss_index.faiss  # FAISS index file (auto-generated)

│   ├── faiss_index.pkl    # FAISS metadata file (auto-generated)

└── README.md              # This file

```

### Core Components

#### Component Details

#### `app.py` - Main Application

The Streamlit-based web interface that orchestrates all components:

- **Chat Interface**: WhatsApp-style messaging UI with timestamps

- **Document Upload**: PDF file upload and processing workflow

- **RAG System Setup**: Initializes and manages the RAG pipeline

- **Session Management**: Handles chat history and system state

#### `src/embeddings.py` - Embedding Management

Handles text vectorization using HuggingFace embeddings:

- **Model**: `BAAI/bge-base-en-v1.5` (base English model)

- **Configuration**: CPU-based processing with normalized embeddings

- **Caching**: Streamlit resource caching for performance

#### `src/loaders.py` - Document Processing

Manages PDF document loading and text chunking:

- **PDF Processing**: Uses PyPDFLoader for document extraction

- **Text Splitting**: Recursive character-based splitting

- **Chunk Configuration**:

  - Default chunk size: 1000 characters

  - Overlap: 200 characters

  - Smart separators: `["\n\n", "\n", " ", ""]`

- **Metadata Preservation**: Maintains source file information

#### `src/chain.py` - Conversational RAG Chain

Implements the core conversational retrieval system:

- **ConversationalRAGChain**: Main class managing the RAG pipeline

- **Google Gemini Integration**: Uses `Gemini` model for responses

- **Memory Management**: Conversation buffer with configurable window size

- **Custom Prompting**: Tailored prompts for document-based Q&A

#### `src/vectorstore.py` - Vector Database

Manages FAISS vector store operations:

- **FAISS Integration**: Facebook AI Similarity Search for vector storage

- **Persistence**: Save/load vector stores to/from disk

- **Document Management**: Add new documents to existing stores

- **Retrieval**: Configurable similarity search (default: top-4 results)

## Dependencies

### Core Libraries

- **streamlit**: Web application framework

- **langchain**: LLM application framework

- **langchain-community**: Community integrations

- **langchain-google-genai**: Google Generative AI integration

- **langchain-huggingface**: HuggingFace embeddings integration

### Processing Libraries

- **sentence-transformers**: Sentence embedding models

- **pypdf**: PDF document processing

- **faiss-cpu**: Vector similarity search

- **python-dotenv**: Environment variable management

## Configuration

### Model Settings

- **LLM Model**: `gemini-1.5-flash` (configurable in `chain.py`)

- **Temperature**: 0.7 (controls response creativity)

- **Memory Window**: 10 messages (conversation history)

### Embedding Settings

- **Model**: `BAAI/bge-base-en-v1.5`

- **Device**: CPU (configurable for GPU)

- **Normalization**: Enabled for better similarity matching

### Document Processing

- **Chunk Size**: 1000 characters

- **Chunk Overlap**: 200 characters

- **Supported Formats**: PDF files only

### Vector Store

- **Backend**: FAISS (Facebook AI Similarity Search)

- **Persistence**: Local disk storage in `vectorstore/` directory

- **Retrieval**: Top-4 similar documents per query

## Error Handling

The application includes comprehensive error handling:

- **Document Processing**: Graceful handling of corrupt PDFs

- **API Failures**: Fallback messages for API issues

- **Vector Store**: Error recovery for storage operations

- **Memory Management**: Automatic cleanup and reset options

## Troubleshooting

### Common Issues

1. **"GOOGLE_API_KEY not found"**

   - Ensure `.env` file exists with valid API key

2. **"No vector store available"**

   - Upload and process documents first

3. **"Error processing PDF"**

   - Check if PDF is not corrupted or password-protected

4. **Memory issues with large documents**

   - Reduce chunk size or process fewer documents at once