Chat Agent — Interview Starter

A chat agent with tool-calling capabilities, built with FastAPI, PostgreSQL, and the OpenAI SDK.

Prerequisites

• Python 3.11+
• Docker & Docker Compose
• Node.js 24+ and pnpm
• uv (Python package manager)

Windows

Use WSL2. Run all commands from a WSL2 terminal, and clone this repo inside WSL2 (e.g. ~/bootstrap-agentic), not on your Windows drive (/mnt/c/...). Docker Desktop must have WSL2 integration enabled for your distro.

Quick Setup

./start.sh

This single command installs dependencies, prompts for your API key (provided in the email with task instructions), starts Docker, and launches the backend + frontend.

Open http://localhost:3000 to verify the chat interface loads. Press Ctrl+C to stop everything.

Run tests

uv run pytest

Note: 2 of 4 tests currently pass. This is expected — you'll address the failing tests during the interview.

Project Structure

├── backend/
│   ├── main.py              # FastAPI application
│   ├── config.py             # Settings (env vars)
│   ├── db.py                 # Database session setup
│   ├── models/               # SQLAlchemy models
│   ├── schemas/              # Pydantic request/response schemas
│   ├── routers/              # API endpoints
│   ├── agent/
│   │   ├── loop.py           # Agent loop (LLM <> tool execution)
│   │   ├── prompts.py        # System prompts
│   │   └── tools/            # Tool registry & implementations
│   └── services/
│       └── llm.py            # LLM client configuration
├── frontend/                 # React/TypeScript chat UI
├── tests/                    # Test suite
├── docker-compose.yml        # Postgres + Redis
├── start.sh                  # One-command setup & run
└── pyproject.toml            # Python dependencies

Troubleshooting

• Docker not starting? Make sure the ports in .env (POSTGRES_PORT, REDIS_PORT) are free on your machine.
• Backend won't start? Check that .env exists and docker compose ps shows healthy services.
• Frontend not loading? Make sure the backend is running on port 8000 (the frontend proxies API calls to it).
• Tests failing? 2 of 4 tests are expected to fail. If all 4 fail, check your Python environment (uv sync --all-extras).