MARC-AI — Multi-Agent Review & Consolidation

Cloud-native FastAPI application that orchestrates LangGraph agents to run static linters & scanners (Ruff, ESLint, Bandit, Semgrep, Radon) and consolidates findings into explainable markdown reports using Azure OpenAI.

Features

• Multi-agent orchestration with LangGraph
• Static analysis only — no code execution
• Parallel agent execution for style, security, and performance checks
• AI-powered explanations via Azure OpenAI (GPT-4)
• Async REST API with FastAPI
• Containerized with multi-stage Docker builds

Quick Start

Prerequisites

• Python 3.11+
• Docker & Docker Compose
• Node.js & npm (for ESLint)

Local Development

1. Install dependencies:

pip install -r requirements.txt

2. Copy environment template:

cp .env.example .env
# Edit .env with your Azure OpenAI credentials

3. Run the application:

uvicorn app.main:app --reload

4. Access API docs at http://localhost:8000/docs

Docker Development

Build and run the development environment:

# Build and start in foreground
docker-compose -f docker-compose.dev.yml up --build

# Or run in background
docker-compose -f docker-compose.dev.yml up -d --build

# View logs
docker-compose -f docker-compose.dev.yml logs -f

# Stop
docker-compose -f docker-compose.dev.yml down

Docker Production

Build and run the production environment:

# Build and start in background
docker-compose -f docker-compose.prod.yml up -d --build

# View logs
docker-compose -f docker-compose.prod.yml logs -f

# Stop
docker-compose -f docker-compose.prod.yml down

Build Specific Stage Only

# Build dev stage
docker build --target dev -t marc-ai:dev .

# Build prod stage
docker build --target prod -t marc-ai:prod .

# Run manually (without docker-compose)
docker run -p 8000:8000 --env-file .env marc-ai:prod

API Endpoints

• GET /api/v1/health/ — Health check
• POST /api/v1/review/analyze — Trigger code analysis
• GET /api/v1/review/status/{run_id} — Check analysis status
• GET /api/v1/review/report/{run_id} — Retrieve analysis report

Project Structure

marc-ai/
├── app/
│   ├── agents/          # Analysis agents (style, security, performance)
│   ├── core/            # Config, logging, dependencies
│   ├── models/          # Pydantic schemas
│   ├── routers/         # API endpoints
│   ├── services/        # Business logic
│   ├── utils/           # Utilities (subprocess, parsers)
│   └── workflows/       # LangGraph workflows
├── tests/               # Test suite
├── Dockerfile           # Multi-stage container build
└── requirements.txt     # Python dependencies

Security

• All static analysis tools run via safe subprocess calls (shell=False)
• Timeouts and resource limits enforced
• No repository code execution
• Isolated working directories per analysis run

Development Status

Current Phase: Initialization — scaffolding complete, business logic pending.

See INIT_PROJECT.md for detailed architecture and implementation guidance.

License

MIT