|
| 1 | +# 🧭 CodeCompass - GitHub Repo Chat |
| 2 | + |
| 3 | + |
| 4 | + |
| 5 | +CodeCompass turns a GitHub repository into an interactive, searchable knowledge base. Add a public repo, index it locally with embeddings (Chroma), and chat with an LLM grounded on real files from that repo. |
| 6 | + |
| 7 | +🔒 Privacy‑first • 🎯 Repo‑scoped answers • 💾 Local vector store • 🔖 Clear citations |
| 8 | + |
| 9 | +--- |
| 10 | + |
| 11 | +## ✨ Highlights |
| 12 | + |
| 13 | +- Repo‑scoped conversational search - answers stay focused on the repo you select |
| 14 | +- Source grounding - responses reference retrieved passages and files |
| 15 | +- Persistent index - once indexed, the repo remains searchable |
| 16 | +- Simple ingestion - accepts `owner/repo` or full GitHub URLs; normalization prevents duplicates |
| 17 | + |
| 18 | +--- |
| 19 | + |
| 20 | +## 📸 Screenshots |
| 21 | + |
| 22 | + |
| 23 | + |
| 24 | + |
| 25 | + |
| 26 | +--- |
| 27 | + |
| 28 | +## 🏗️ Architecture |
| 29 | + |
| 30 | +``` |
| 31 | +React (Vite + Tailwind) |
| 32 | + ↕ |
| 33 | +Flask REST API |
| 34 | + ↕ |
| 35 | +Embedchain (ingest + retrieve) |
| 36 | + ↕ ↘ |
| 37 | + Chroma (vectors) Groq LLM |
| 38 | +``` |
| 39 | + |
| 40 | +--- |
| 41 | + |
| 42 | +## 🛠️ Technical notes |
| 43 | + |
| 44 | +- Embeddings: sentence‑transformers `all‑MiniLM‑L6‑v2` (fast and compact) |
| 45 | +- Vector DB: Chroma persisted under `backend/db` for local, fast retrieval |
| 46 | +- LLM: Groq (configurable). Prompt encourages concise, source‑grounded answers |
| 47 | +- Metadata: each chunk stores `github_repo`, `file_path`, `chunk_index` for precise citations |
| 48 | + |
| 49 | +--- |
| 50 | + |
| 51 | +## 🧰 Troubleshooting |
| 52 | + |
| 53 | +- 401/403 from LLM → check `GROQ_API_KEY` in `backend/.env`, restart Flask |
| 54 | +- Cannot add repo → ensure it’s a public GitHub repository or provide a `GITHUB_TOKEN` |
| 55 | + |
| 56 | +--- |
| 57 | + |
| 58 | +## 🗺️ Roadmap |
| 59 | + |
| 60 | +- File/line citations with clickable open‑in‑editor behavior |
| 61 | +- Partial/continuous indexing for very large repos |
| 62 | +- Answer streaming UI for progressive responses |
| 63 | +- Docker images and a simple hosted deployment option |
0 commit comments