A full-stack AI starter kit that actually finishes the plumbing
Install a CLI, answer a few prompts, and get a working FastAPI + Next.js app with agents, RAG, auth, and Docker wired up.

What it does
fastapi-fullstack is a project generator that scaffolds a complete AI application: FastAPI backend, Next.js 15 frontend, Postgres, Redis, optional vector stores (Milvus, Qdrant, pgvector, ChromaDB), Celery, and Docker Compose for dev/stage/prod. It bundles auth (JWT/OAuth), an admin panel, WebSocket streaming chat, conversation sharing, and Stripe billing pages. You pick your stack through an interactive CLI or a web configurator, and it spits out a repo you can run with make bootstrap.
The interesting bit
The generator doesn’t lock you into one agent framework — it supports six, including Pydantic AI, LangGraph, CrewAI, and the author’s own pydantic-deepagents. That’s unusual for a template; most pick a horse and ride it. The generated projects also get a Django-style CLI named after your project slug (uv run my_ai_app db migrate), which is a nice touch for long-term maintenance.
Key highlights
- Six AI agent frameworks selectable at generation time
- RAG pipeline with four vector store options
- WebSocket streaming chat with real-time tool calls and chart generation
- Dev/stage/prod Docker Compose targets with idempotent
make dev - Optional marketing site, pricing page, blog, and org multi-tenancy
Caveats
- The README is thorough but long; finding specific config options takes scrolling
- 1,323 stars is respectable but the ecosystem is young — several linked sibling repos have low activity
- Windows users need WSL2 or Git Bash; the Makefile dependency is noted but not optional
Verdict
Good fit if you want to ship an AI SaaS fast and don’t mind trading some flexibility for structure. Skip it if you already have strong opinions about your frontend framework or agent orchestration — this template makes choices for you.