🚀 LuminaDoc: Your Offline Document Intelligence Hub 🔍📚

LuminaDoc is a powerful, privacy-focused document analysis tool that brings RAG (Retrieval-Augmented Generation) capabilities to your local environment. Process, analyze, and interact with your documents using advanced LLMs—all without requiring an internet connection.

---

📚 Key Features

• 100% Offline Processing: Your documents never leave your machine.
• Built-in RAG Pipeline: Advanced retrieval and generation workflows.
• Local Vector Storage: Supports ChromaDB for efficient retrieval.
• Format Support: PDF.
• Intelligent Chunking: Semantic search with precise document splitting.
• Conversational Interface: Ask questions, get contextual answers.
• Custom Knowledge Base: Build and manage your knowledge repositories.

---

💡 Perfect For:

• 📊 Researchers: Analyze private datasets securely.
• ⚖️ Legal Professionals: Handle confidential legal documents locally.
• 💼 Business Analysts: Process sensitive data without cloud dependencies.
• 🧑‍💻 Developers: Build offline-first AI-powered applications.
• 🔐 Privacy Advocates: Ensure your data stays secure and private.

---

🛠️ Tech Stack

• Backend: Python 3.9+
• Frontend: Streamlit
• Database: Local Vector DB (ChromaDB)
• Embedding Model: Ollama (nomic-embed-text)
• LLM Support: Local LLMs (e.g., llama3.2:latest)
• Cross-Encoder: Sentence Transformers (ms-marco-MiniLM-L-6-v2)

---

🔒 Privacy First

• No Cloud Services: All processing happens locally.
• No API Keys Required: Run seamlessly without third-party dependencies.
• No Data Sharing: Your documents remain secure on your machine.

---

📦 Installation Guide

Follow these steps to install and run LuminaDoc locally:

✅ Step 1: Clone the Repository

Open your terminal and clone the LuminaDoc repository:

git clone https://github.com/yourusername/luminadoc.git
cd luminadoc

---

✅ Step 2: Create a Virtual Environment

Create an isolated Python environment:

python3 -m venv venv
source venv/bin/activate  # On Linux/MacOS
venv\Scripts\activate     # On Windows

---

✅ Step 3: Install Dependencies

Install the required Python packages:

pip install --upgrade pip
pip install -r requirements.txt

Ensure Required Dependencies Include:

• streamlit
• langchain-ollama
• chromadb
• sentence-transformers
• PyMuPDF
• httpx
• langchain-community
• langchain
• ollama

---

✅ Step 4: Install Ollama and Pull Models

If Ollama is not installed, run the following commands:

For macOS:

brew install ollama

For Linux:

curl -fsSL https://ollama.com/install.sh | sh

Verify the installation:

ollama --version

Pull Required Models:

Make sure to download the required models:

ollama pull nomic-embed-text
ollama pull llama3.2:latest

Start the Ollama server:

ollama serve

Verify the models:

curl http://localhost:11434/api/tags

Ensure the nomic-embed-text and llama3.2 models are listed.

---

✅ Step 5: Start LuminaDoc

Run the Streamlit app locally:

streamlit run app.py

The app will start, and you’ll see a URL in your terminal, usually:

Local URL: http://localhost:8501

Open this in your browser.

---

🚀 Using LuminaDoc

1. Upload a Document:

   • In the sidebar, upload a PDF file.
   • Click on the Process button.

2. Ask Questions:

   • Enter your question in the text area.
   • Click Ask to retrieve relevant information.

3. View Results:

   • The system will show the retrieved context and the AI-generated answer.
   • Explore retrieved documents and relevant text via expandable sections.

---

✅ Environment Variables (Optional)

If you want to customize configurations:

Create a .env file:

OLLAMA_SERVER_URL=http://localhost:11434
VECTOR_DB_PATH=./rag-chroma

Update your code to load .env using dotenv if needed.

---

✅ Troubleshooting

1. Port Conflicts: Ensure Ollama is running on http://localhost:11434.

2. Dependencies Issues: Run pip install -r requirements.txt again.

3. Streamlit Errors: Clear the cache:

   streamlit cache clear

4. Logs: Check server logs for errors:

   ollama serve --verbose

---

🧠 How It Works

1. Document Upload & Processing:

   • Documents are uploaded and split into smaller semantic chunks.
   • Chunks are embedded using nomic-embed-text and stored in ChromaDB.

2. User Query:

   • Your query searches the vector database for relevant chunks.
   • Results are re-ranked using CrossEncoder.

3. LLM Response:

   • The context is passed to LLM (e.g., llama3.2) for final generation.
   • The answer is displayed in the Streamlit interface.

---

🔗 Next Steps

• Explore advanced features like custom pipelines.
• Experiment with different local LLM models.
• Contribute to the project on GitHub!

---

Illuminate your documents with AI-powered insights—locally and securely. 🚀📚