🧠 CodeSense

AI-powered semantic code search over your own codebase using Tree-sitter + CodeBERT + FAISS + Streamlit

---

🚀 Overview

CodeSense lets you perform natural language searches over source code — like "connect to database" or "handle login request" — and retrieve the most semantically relevant functions, not just keyword matches.

Built with:

• Tree-sitter for accurate code parsing
• CodeBERT for semantic embeddings
• FAISS for fast vector search
• Streamlit for interactive UI

---

📦 Features

✅ Parse source code into meaningful chunks
✅ Embed code semantically using CodeBERT
✅ Perform fast vector similarity search
✅ Interactive Streamlit UI for natural language queries
✅ Plug-and-play: works on any Python repo

---

📂 Project Structure

codesense/
├── app/ # Streamlit UI
├── backend/ # Parser, Embedding, Search engine
├── data/ # Raw code, processed chunks, embeddings
├── scripts/ # Manual scripts for testing modules
└── README.md

---

🛠️ Getting Started

1. Clone & Setup

git clone https://github.com/NoxiousTab/codesense.git
cd codesense
python -m venv venv
venv\Scripts\activate    # or source venv/bin/activate on Unix
(Only If Required) pip install tree_sitter-0.20.4-cp310-cp310-win_amd64.whl
pip install -r requirements.txt
python backend/parser/build_language.py # to build my-languages.so

2. Clone Repos to Analyze

python scripts/test_clone.py

3. Parse Code and extract chunks

python scripts/test_extract_chunks.py

4. Embed Chunks

python scripts/test_embed_chunks.py

5. Launch UI by streamlit

streamlit run app/ui.py