ADR 0001 — Use FastAPI for the project API

Status

Accepted

Context

The project needs a modern, fast, and developer-friendly HTTP API framework to expose endpoints (chat, embeddings, admin, health) and integrate model/logic layers. The team values quick iteration, type safety, automatic OpenAPI docs, and async I/O for concurrency.

Decision

We will use FastAPI as the primary web framework for the project’s API layer.

Consequences

• Pros:
  • Excellent async support and performance for I/O-bound workloads.
  • Automatic generation of OpenAPI and interactive docs (Redoc/Swagger).
  • First-class Pydantic integration for request/response validation and typed models.
  • Large community and ecosystem, many extensions and deployment examples.
• Cons / Trade-offs:
  • Slight learning curve for developers unfamiliar with async/await in Python.
  • If parts of the codebase need heavy CPU-bound work, additional worker processes or offloading strategies will be required.

Alternatives considered

• Flask (with Flask-RESTX): simpler but lacks built-in async, slower to get typed request/response validation and OpenAPI generation.
• Django REST Framework: heavier, more batteries-included — too large for a lightweight service primarily serving model/AI endpoints.

References

• https://fastapi.tiangolo.com/
• Pydantic docs