backend

Backend

FastAPI service for the Mentee bot.

Stack

• Python 3.14
• FastAPI 0.135 + Uvicorn 0.44
• Pydantic 2.13 + pydantic-settings
• Managed with uv

Setup

uv sync
cp .env.example .env   # tweak if needed

Running

uv run uvicorn app.main:app --reload --port 8001

The app will be available at http://localhost:8001. Interactive docs at /docs.

Linting

uv run ruff check .

Layout

app/
├── main.py              # FastAPI app, CORS, router wiring
├── core/config.py       # Settings (pydantic-settings)
├── domain/              # Message, Thread, User, MessageRole
├── agents/
│   ├── base.py          # AgentPort abstract interface
│   └── mock/agent.py    # MockAgent (deterministic replies, no LLM)
├── services/
│   ├── thread_store.py  # In-memory store (swap later for DB)
│   └── message_service.py  # Orchestrates user msg → agent → assistant msg
└── api/
    ├── deps.py          # Session + service injection helpers
    └── routes/
        ├── chat.py      # POST /api/chat/messages, GET /api/chat/thread
        ├── auth.py      # Stub OAuth: /login, /callback, /me, /logout
        └── health.py    # GET /health

Extension seams

• Replace MockAgent with pydantic-ai / OpenAI / Perplexity — implement AgentPort, swap in app/api/deps.py.
• Replace ThreadStore with Postgres / Mongo — keep the same method signatures.
• Replace stub auth in app/api/routes/auth.py with real MenteeGlobal OAuth — frontend contract stays identical.