MLExpert AI App Starter

Monorepo boilerplate for AI SaaS apps. See PRD.md for the full spec and TODO.md for progress.

Prerequisites

uv (installs Python 3.12 automatically)
Node.js 20+ and npm
Git

Install

# Backend deps (creates .venv and installs dev group)
uv sync --project backend

# Frontend deps
npm --prefix frontend install

# Install git hooks (one-time)
uv run --project backend pre-commit install

Run

Backend

uv run --project backend uvicorn app.main:app --reload --app-dir backend

Runs at http://localhost:8000.

Backend API

Method	Endpoint	Description
GET	`/healthz`	Liveness check
POST	`/api/v1/documents`	Upload a `.txt`, `.md`, or `.pdf` (≤ 5 MB)
GET	`/api/v1/documents`	List uploaded documents (metadata only)
DELETE	`/api/v1/documents/{id}`	Remove a document
POST	`/api/v1/threads`	Create a thread (optionally attach documents)
GET	`/api/v1/threads`	List threads
GET	`/api/v1/threads/{id}`	Get thread with messages
DELETE	`/api/v1/threads/{id}`	Delete a thread
POST	`/api/v1/chat/stream`	Stream chat response (Vercel AI SDK data-stream format)

See client.py for a CLI tool to upload, list, and delete documents:

uv run --project backend python client.py upload file.txt
uv run --project backend python client.py list
uv run --project backend python client.py delete <doc-id>

Frontend

npm --prefix frontend run dev

Visit http://localhost:3000.

Test & lint

Run every check (FE + BE)

uv run --project backend pre-commit run --all-files && npm --prefix frontend run typecheck

This covers: YAML/JSON validation, ruff check + ruff format, ESLint, backend pytest, and frontend TypeScript typecheck (tsc isn't in pre-commit because it's too slow to run on every commit).

Individual checks

# Backend
uv run --project backend pytest
uv run --project backend ruff check backend
uv run --project backend ruff format backend

# Frontend
npm --prefix frontend run lint
npm --prefix frontend run typecheck

Pre-commit runs automatically on git commit. If a hook fails, fix the issue, re-stage, and commit again.

Environment variables

Copy .env.example to .env and fill in your keys. AI_PROVIDER + MODEL_NAME control which LLM is used (ollama, openai, or anthropic).

API types for the frontend

The backend exposes an OpenAPI schema at http://localhost:8000/openapi.json. The frontend can generate typed clients from it using openapi-typescript.

Layout

.
├── PRD.md                    # Product requirements
├── TODO.md                   # Progress tracker
├── README.md                 # You are here
├── .env.example              # Environment variables template
├── .pre-commit-config.yaml   # yaml/json/ruff/eslint/pytest hooks
├── backend/
│   ├── pyproject.toml        # uv-managed deps, pytest config
│   ├── client.py             # CLI tool to upload/list/delete documents
│   ├── app/
│   │   ├── main.py           # FastAPI app, CORS, router mounts
│   │   ├── schemas.py        # Pydantic request/response models
│   │   ├── state.py          # In-memory dataclasses (Document, Thread, etc.)
│   │   ├── core/
│   │   │   └── config.py     # Env-driven settings
│   │   ├── routers/
│   │   │   ├── health.py     # GET /healthz
│   │   │   ├── documents.py  # Upload, list, delete documents
│   │   │   ├── threads.py    # Create, list, get, delete threads
│   │   │   └── chat.py       # Streaming chat endpoint
│   │   └── services/
│   │       ├── parse.py      # Text extraction (txt/md/pdf)
│   │       ├── llm.py        # get_llm() provider factory
│   │       ├── graph.py      # LangGraph single-node workflow
│   │       └── sse.py        # Vercel AI SDK data-stream encoder
│   └── tests/                # pytest (asyncio_mode = "auto")
└── frontend/
    ├── package.json
    ├── src/app/              # Next.js App Router
    └── ...

ai-project-template

Description

README