A VS Code extension that performs automated code review using a 9-agent multi-provider AI pipeline with RAG (Retrieval-Augmented Generation).
Features
- 9 specialized agents: Security, Quality, Error Handling, Style, Complexity, Duplication, Documentation, Tests, Dependencies
- Multi-provider: Groq, Gemini, NVIDIA NIM, Cerebras, OpenRouter — with automatic fallback
- RAG pipeline: Vector store (LanceDB) for pattern matching and history-aware reviews
- Two Agents Debate: Opposing LLM agents argue grey-zone scores (4–7)
- Diff-Aware Re-Review: Compares each review against the previous snapshot
- Multi-language: TypeScript, JavaScript, Python, Java, Go
- Quiet Save:
Ctrl+Shift+Ssaves without triggering a review
Requirements
Configure your API keys in VS Code settings (Ctrl+, → search "AI Reviewer"):
| Setting | Provider | Purpose |
|---|---|---|
aiReviewer.groqApiKey |
Groq | Security agent |
aiReviewer.geminiApiKey |
Google Gemini | Quality agent |
aiReviewer.nvidiaApiKey |
NVIDIA NIM | Error, Style, Complexity, Docs, Tests |
aiReviewer.cerebrasApiKey |
Cerebras | Debate agent |
aiReviewer.openrouterApiKey |
OpenRouter | Debate + embeddings fallback |
aiReviewer.githubToken |
GitHub | PR context integration |
Usage
- Open a TypeScript, JavaScript, Python, Java, or Go file
- Save the file — the review panel opens automatically
- Or run AI Reviewer: Analyze Current File from the Command Palette
Supported Languages
- TypeScript / JavaScript / TSX / JSX
- Python 3
- Java
- Go
PFE Project
Developed as a Final Year Project (PFE) at ESPRIT — Mohamed Farouk BOUABID.