Stateless FastAPI service that analyzes GitHub repos with Groq-hosted Llama 3.3 70B: security, architecture, code quality, and performance checks through one API, plus webhook-triggered analysis on every push.
PythonFastAPINext.jsTypeScriptDocker
// why this exists
Most 'AI code review' tools want a database, a queue, and persistent state before they'll look at a single file. Aegis Agent doesn't: it's a single stateless FastAPI service that pulls a repo straight from the GitHub API, walks the real directory tree instead of just the README, and hands the prioritized files to Groq's Llama 3.3 70B for analysis. Nothing to lose if the process restarts mid-request.
// how it's built
Stateless Analysis Service
- No database or Redis, every request is self-contained
- FastAPI async endpoints with auto-generated OpenAPI docs
- Model fallback chain from Llama 3.3 70B down to Llama 3.1 8B if rate-limited
Repository Walking
- Recursively explores app/, src/, components/, api/, lib/ up to 3 levels deep
- Prioritizes entry points and routes before config files
- README is always analyzed first, plus up to 9 more source files per request
Analysis Types & Integration
- Five modes: general, security, architecture, quality, performance
- GitHub webhook support for automatic analysis on push
- Accepts owner/repo shorthand or a full GitHub URL