Index your entire codebase, git history, PRs, tickets and discussions into a knowledge graph — then chat with it in plain English.
Every developer has felt this: you join a project (or return to your own after 6 months) and spend days just figuring out how code connects, why decisions were made, what breaks if you change something.
CodeMind fixes that. One command indexes your repo into a queryable knowledge graph — then you just ask it.
$ codemind chat
> Why does the auth middleware skip OPTIONS requests?
> Which files would break if I rename UserService?
> Who wrote the payment module and what PR introduced it?
| Feature | Description |
|---|---|
| 🗺️ Living architecture map | Auto-generated graph of your codebase — files, functions, classes and their relationships |
| 💬 Chat with your code | Natural language Q&A powered by a free local LLM (no API keys needed) |
| 🔗 Code + context linked | Connects code to the PRs, Jira tickets, and Slack threads that explain why it exists |
| 🔄 Impact predictor | Ask "what breaks if I change X?" before touching anything |
| 👥 Onboarding paths | Auto-generate "start here" guides for new team members |
| 🚨 Drift detection | Alerts when the codebase diverges from its documented architecture |
- Python 3.9+
- Docker + Docker Compose — runs Neo4j and Qdrant
- Ollama — free local LLM runtime
# 1. Clone
git clone https://github.com/vineethwilson15/codemind.git
cd codemind
# 2. Start databases (Neo4j + Qdrant)
docker compose up -d
# 3. Pull the LLM (free, runs locally — no API key needed)
ollama pull codellama
# 4. Install Python dependencies
pip install -r requirements.txt
# 5. Configure environment
cp .env.example .env # defaults work out of the box# Index any codebase (yours, an OSS project, anything)
python -m cli index --repo /path/to/your/project
# Start chatting
python -m cli chatCodeMind is pluggable — connect only what your team uses.
| Category | Supported |
|---|---|
| Git Hosts | GitHub · GitLab · Bitbucket · Azure Repos |
| Issue Trackers | Jira · Azure Boards · Linear · GitHub Issues |
| Chat / Discussions | Slack · Microsoft Teams |
| LLM | Ollama (default, free, local) · OpenAI · Anthropic |
Configure via .env — see .env.example for all options.
Your Codebase + Git History
│
▼
┌───────────────────┐
│ AST Parser │ tree-sitter — extracts functions,
│ (core/ast_parsers)│ classes, imports per language
└────────┬──────────┘
│
▼
┌───────────────┐ ┌──────────────────┐
│ Neo4j Graph │◄─────►│ Qdrant Vectors │
│ knowledge │ │ semantic search │
│ graph │ │ (local embeddings│
└───────┬───────┘ └──────────────────┘
│
▼
┌───────────────┐
│ Ollama LLM │ free, local — codellama by default
│ (core/chat) │
└───────┬───────┘
│
▼
┌──────────────────────────┐
│ CLI · API · VS Code │
└──────────────────────────┘
By default, everything runs 100% locally — no data leaves your machine. Enabling cloud LLMs (OpenAI, Anthropic) or external integrations (GitHub, Jira, Slack) will send the relevant data to those providers.
- Phase 1 — Core Engine: repo indexer, knowledge graph, local LLM chat, CLI
- Phase 2 — Integrations: GitHub/GitLab, Jira/Azure Boards, Slack/Teams
- Phase 3 — UI & Extension: VS Code extension, web UI with live architecture map
Contributions are welcome and appreciated!
# Run unit tests (no Docker needed)
python -m pytest tests/unit/
# Run integration tests (requires docker compose up -d)
python -m pytest tests/integration/MIT — see LICENSE
