🧠 MindPalace

MindPalace is a private, local-first AI assistant that lets you upload and chat with your documents. Inspired by Sherlock Holmes' mental filing system, it parses and stores knowledge from your files into an in-memory vector store, ready to be queried via natural language.

🚀 Features

• 🧾 Upload PDFs, text files, images (OCR), or URLs
• 🧠 Ask questions using a local LLM (via Ollama)
• 🗃️ All processing happens locally — no cloud, no leaks
• 🧩 Modular parser architecture
• 🎛️ Simple Gradio UI

🛠️ Stack

• LangChain – Orchestration & document processing
• Ollama – Run LLMs like LLaMA3, Mistral, or Gemma locally
• FAISS – In-memory vector store
• Gradio – Web UI
• Tesseract OCR – For image-to-text conversion

📦 Installation

# Clone the repo
git clone https://github.com/SherLock707/MindPalace.git
cd MindPalace

# Install dependencies
pip install -r requirements.txt

# Optional: Ensure Tesseract is installed
# Ubuntu: sudo apt install tesseract-ocr
# macOS: brew install tesseract

🧪 Run the App

Start your local LLM first:

ollama run llama3

Then start the MindPalace server:

python main.py

🧰 Folder Structure

MindPalace/
├── main.py
├── requirements.txt
├── README.md
├── helpers/
│   ├── parser_factory.py
│   ├── vectorstore.py
│   ├── llm_chain.py
│   ├── ui.py
│   └── parsers/
│       ├── pdf_parser.py
│       ├── txt_parser.py
│       ├── image_parser.py
│       └── url_parser.py

🔒 Privacy First

MindPalace runs entirely on your machine. No API calls, no telemetry, no vendor lock-in.

📌 Roadmap

• Add persistent vector DB option (Chroma, SQLite)
• Highlight citations in responses
• Export & reload sessions
• CLI interface for headless mode

🙏 Credits

• LangChain
• Ollama
• FAISS

📄 License

MIT License. Yours to hack, extend, and evolve.

---

🧑‍💻 Author

Built by an engineer who believes your documents should stay yours. Let AI read your files without sending them anywhere else.