LocalRAG

A lightweight, local Retrieval-Augmented Generation (RAG) web application that allows you to chat with your documents using local LLMs. Built with Flask, LangChain, ChromaDB, and Ollama.

🚀 Features

• Local & Private: Your documents and queries never leave your machine.
• Multi-Session Support: Organise different topics into separate chat sessions.
• Document Management: Upload and index various file formats including:
  • PDF (.pdf)
  • Text (.txt)
  • CSV (.csv)
  • Markdown (.md)
• Flexible Configuration:
  • Memory: Adjust how many past interactions are included in the context.
  • Collections: Group documents into different vector database collections.
  • Model Selection: Switch between different local models available on Ollama.
• Smart Context: Automatically retrieves relevant chunks from your documents to answer questions.
• Interactive UI: Clean, responsive interface with Markdown support and source citations.
• Fun Empty State: Displays random interesting Wikipedia articles while you wait or before you start.

🛠️ Prerequisites

1. Python 3.8+
2. Ollama: Download and install from ollama.com.
3. Required Models:
   • For answering: ollama pull deepseek-r1:1.5b (or your preferred model)
   • For embeddings: ollama pull nomic-embed-text

📦 Installation

1. Clone the repository:

   git clone <repository-url>
   cd RAG-web

2. Create a virtual environment:

   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies:

   pip install -r requirements.txt

🚦 Getting Started

1. Start Ollama: Ensure the Ollama service is running on your machine.
2. Run the application:

   python app.py

   Or use the provided batch file (Windows):

   run.bat

3. Open in Browser: Navigate to http://127.0.0.1:5000.

📖 How to Use

1. Create a Session: Click the + button in the sidebar to start a new chat.
2. Upload Documents: Click the paperclip icon, select your files, and they will be indexed into the current collection.
3. Configure:
   • Context Memory: Set how many previous message pairs to remember.
   • Collection: Name your vector store (e.g., "LegalDocs", "Manuals").
   • Model: Specify the Ollama model name you want to use.
4. Ask Questions: Type your query in the input bar. The assistant will search your documents and provide an answer with citations.

📂 Project Structure

RAG-web/
├── app.py              # Main Flask application & API routes
├── utils.py            # RAG logic, document loaders, & embedding helpers
├── requirements.txt    # Python dependencies
├── sessions.json       # Persisted chat sessions (generated)
├── chroma/             # Local vector database storage (generated)
├── static/
│   ├── app.js          # Frontend logic (Vanilla JS)
│   └── style.css       # Styling
└── templates/
    └── index.html      # Main UI template

🤝 Contributing

Feel free to fork, open issues, and submit PRs

📜 License

MIT License.