Purpose

FastAPI backend to support:

• AI/LLM integration (architecture TBD - may be separate service)
• Galaxy tool execution
• NCBI cross-linking (see Link from NCBI to BRC Analytics #783)
• SRA metadata database integration

Initial Directory Structure

/backend/
├── app/
│ ├── api/v1/ # Feature routers (health, version, cache, etc.)
│ ├── core/ # Config, cache, dependencies, auth
│ └── main.py
├── Dockerfile
├── docker-compose.yml
├── nginx.conf # Reverse proxy for backend
└── pyproject.toml
**Rationale: Start simple with single service. Can refactor to /backend/api-service/ structure later if we add more services.

Prod Deployment Architecture

Frontend: Static Next.js + CloudFront (existing, unchanged)
Backend: FastAPI + nginx

Tech Stack

FastAPI + uvicorn (async Python)
nginx - Backend reverse proxy
Redis - Caching (job status, API responses)
Docker + Docker Compose
Ruff - Linting

Key Features

Async operations throughout
Shared auth/caching across features
Single URL local dev
E2E test support
Version syncing

API Endpoints (Initial)

/health - Service monitoring
/version - API metadata
/cache/* - Cache management

Open Questions

AI service architecture: Router in this service vs. separate service?

• Depends on: desire for portability, scaling needs, deployment complexity
• Decision deferred until requirements clearer

Next.js server requirement: Do we need a Next.js server for faster builds, image optimization, and NextAuth (Google login)?

• Consider: Are there alternative solutions?
• Trade-offs: Static site simplicity vs. server runtime complexity/cost

Success Criteria

FastAPI + nginx + Redis running in Docker
Health/version endpoints functional
E2E tests passing
Local dev environment working