LLM Code Review Bot
Automated pull-request reviews powered by GPT-4 — zero-config GitHub integration
What is this project?
LLM Code Review Bot wires into your GitHub repository via Webhooks. Every time a pull request is opened or updated, the service fetches the diff, chunks it into context windows, and sends each chunk to GPT-4 with a project-specific system prompt.
The results are de-duplicated, ranked by severity (error → warning → suggestion), and posted as inline review comments directly on the PR. A summary card is added as a PR-level comment with an overall quality score.
The system is stateless per review — each job is queued via Celery, results are cached in Redis to avoid re-reviewing unchanged hunks, and all review history is persisted in PostgreSQL for analytics. Teams report a 40% reduction in human review turnaround and near-zero security antipatterns slipping through to main branches.
Interface Gallery
System Diagrams
Build Timeline
What's Included
- Full FastAPI + Celery + Redis source code
- Prompt templates with configurable severity rules
- PostgreSQL schema & Alembic migrations
- Docker Compose + Kubernetes Helm chart
- GitHub App setup guide & deployment docs
- 90-day free email support