Your magical assistant for developer documentation
CodeSage leverages RAG (Retrieval-Augmented Generation) and LLM technology to help developers query their documentation and get precise, context-aware answers. Upload your docs and embeddings, and CodeSage becomes your personal copilot for faster coding and learning.
- Chat-style interface using Streamlit
- Smart document retrieval with FAISS
- Answers generated via Mistral
- Handles Markdown and text documentation
- Easy deployment with Docker
### 1. Clone the repository
git clone https://github.com/jhanikita/codesage.git
cd codesage
### 2. Create virtual environment
python3 -m venv .venv
source .venv/bin/activate
### 3. Install dependencies
pip install --upgrade pip
pip install -r requirements.txt
pip install streamlit uvicorn
### 4. Run backend
uvicorn src.api:app --host 0.0.0.0 --port 8000
### 5. Run frontend (in another terminal)
streamlit run src/frontend.py --server.port 8501 --server.headless true
### Note -
1. Make sure FAISS index exists: embeddings/faiss_index.pkl
2. For Ollama Mistral, pull model locally using - ollama pull mistral
3. You can also add more documentation in the data/ folder to make the assistant smarter and more accurate.
4. You can use docker as well