fast-agent-stack¶

FastAPI framework for production AI agents.

fast-agent-stack wires together the infrastructure you need for real AI applications — auth, database, vector search, background tasks, rate limiting, observability, and email — so you can focus on your agent logic.

Quick Start¶

mkdir myproject && cd myproject
pip install fast-agent-stack
fastagentstack new myproject --preset agent
uv pip install -r pyproject.toml
fastagentstack migrate
fastagentstack dev

Presets¶

Preset	What you get
`minimal`	FastAPI + SQLAlchemy only. Fastest start.
`standard`	+ JWT auth + SQLAdmin
`full`	+ background tasks + rate limiting + tracing + email
`agent`	+ Bedrock LLM + Qdrant + S3 + all of `full`

Key Design Choices¶

Database: SQLAlchemy 2 async, Alembic migrations, multi-DB support (Postgres, MySQL, SQLite)
Auth: JWT + session backends, pluggable via dotted-path (ADR-034)
AI: LLMBackend Protocol — Bedrock, OpenAI, Anthropic, LiteLLM
RAG: StorageProtocol + VectorStoreProtocol + EmbeddingProtocol
Tasks: Dramatiq + Redis/Valkey
Observability: OpenTelemetry + Jaeger (OTLP exporter, config-driven)