🧠 Local Mind

---

📸 Demo

Click to expand

---

🧑‍💻 Requirements

• Tavily Search API Key:
  Local Mind uses Tavily Search to fetch live web results for its research agent.

  • Get your free Tavily API key here: https://app.tavily.com/home
  • Add your key to your .env file as:

    TAVILY_API_KEY=your-key-here
    

  • This is required for web research to function!

• Local LLM Model:
  Local Mind runs a quantized Jan-nano model on your machine, so all your data stays private.

  • What is Jan-nano?
    Jan-nano is a highly efficient, open-source LLM by Menlo, specifically designed for running on local CPUs and resource-limited hardware, making it perfect for private, local AI.

    • Trained on high-quality English datasets.
    • Optimized for speed and context length.
    • Well-suited for chat, question answering, and code.

  • Quantized Model Downloads:
    Local Mind supports quantized GGUF versions for best performance on your system.

    • Choose a quantized file (*.gguf) from this list.
    • Download the version that matches your hardware and put it in your server/model/ or the main server directory.

    Model Variant	RAM Required	File Size	Download Link
    Q4, Q5, Q6	Varies	~1-2GB	Choose here

---

🚀 Quick Start

You need both a Tavily API key and a quantized Jan-nano GGUF model file!

1. Get your Tavily API Key
   and add it to server/.env:

TAVILY\_API\_KEY=your-key-here

2. Download a quantized Jan-nano model:
   Pick from Jan-nano GGUF releases
   and put your chosen .gguf file in server/model/.

3. Follow previous setup for server and client...

---

🧠 About Jan-nano

• Jan-nano is a state-of-the-art, efficient LLM built for privacy, low resource usage, and versatility.
• Model card & benchmarks
• Quantized downloads & sizes
• Pick Q4 for low RAM, Q5/Q6 for best quality if you have more resources.

---

🕸️ About Tavily Search

• Tavily is a fast, privacy-friendly web search API for AI research agents.
• Get your API key at https://app.tavily.com/home.
• Enables Local Mind to find and cite up-to-date answers beyond your files.

---

See the rest of the README above for setup, API, and use cases!

---

🏗️ Project Structure

local-mind/
├── client/       # Next.js React frontend (chat UI, file upload, etc)
├── server/       # FastAPI backend, LLM, RAG, and web search agents
│   ├── app/
│   │   ├── config.py
│   │   ├── utils.py
│   │   ├── rag.py
│   │   ├── research.py
│   │   └── main.py
│   ├── data/           # Your uploaded/managed documents
│   ├── store\_rag/      # Local vector store index
│   └── requirements.txt
├── README.md

---

✨ What is Local Mind?

Local Mind is your own personal, private, full-stack AI workspace.

• Upload files, chat with your own knowledge base
• Research with live web search (and get real, cited answers)
• All locally, all private — powered by FastAPI and Next.js

---

🖥️ Client (Next.js)

• Beautiful chat interface
• File upload and management
• Live streaming answers from LLM
• Web search results shown in real time

The client talks to the FastAPI backend via simple HTTP endpoints.

---

⚡️ Server (FastAPI, LlamaIndex, LangChain, LLM)

• Runs your local LLM (Llama.cpp)
• Retrieval-Augmented Generation on your uploaded files
• Web research agent for up-to-date answers
• Automatic file watching and index updating
• API endpoints for chat, file management, research

---

🛠️ Use Cases

• Personal Knowledge Base: Chat with your notes, manuals, or code docs.
• Research Copilot: Get the best web and local insights, cited and summarized.
• Secure Team Docs: Host on LAN, share within your org—no data ever leaves.
• Developer Assistant: Index your codebase docs, get instant answers.
• Academic Summaries: Ask questions to your PDFs or web research.

---

🚀 Quick Start

1️⃣ Clone the repo

git clone https://github.com/yourusername/local-mind.git
cd local-mind

2️⃣ Setup the Server

cd server
python -m venv .venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate
pip install -r requirements.txt
# Place your Llama.cpp GGUF model in this folder (e.g. jan-nano-128k-Q5_K_M.gguf)
python -m app.main
# or
uvicorn app.main:app --reload

• The server runs at http://localhost:8000
• Interactive API: http://localhost:8000/docs

3️⃣ Setup the Client

cd ../client
npm install
npm run dev

• The client runs at http://localhost:3000

---

🌐 How it Works

1. Upload your files (PDF, TXT, etc) via the web UI.
2. Chat with your knowledge base. Get answers instantly.
3. Ask web questions: The agent pulls the latest info and cites real URLs.
4. All private: Your files and chats never leave your device.

---

📖 API Reference

The FastAPI backend exposes endpoints such as:

• POST /rag/upload/ — Upload documents
• GET /rag/files/ — List uploaded files
• DELETE /rag/delete/{filename} — Delete documents
• GET /rag/stream — Chat with your files (RAG)
• GET /research/stream — Research agent (web search)
• GET /health — Health check

See http://localhost:8000/docs for full details.

---

💡 Why Local Mind?

• 🛡️ Privacy First: 100% local, no data leaves your computer.
• 🧠 Multi-source AI: Mixes your files and the web.
• ⚡ Lightning Fast: No cloud lag, no rate limits.
• 🧩 Composable: Easy to extend or add your own tools.

---

🤝 Contributing

Pull requests, issues, and suggestions are always welcome!

• Want to add more LLM models? New RAG features? Better UI? Jump in!

---