Installation¶

Requirements¶

Python ≥ 3.12
An LLM API key (Anthropic, OpenAI, Google, Azure, or any LiteLLM provider)

Install¶

uv (recommended)pippipx (global CLI)From source

uv add retrai

pip install retrai

pipx install retrai

git clone https://github.com/bmsuisse/retrAI
cd retrAI
uv sync

Configure Your LLM¶

retrAI uses LiteLLM under the hood, so it works with any provider — just set the right env var:

Provider	Env Variable	Example Model
Anthropic	`ANTHROPIC_API_KEY`	`claude-sonnet-4-6`
OpenAI	`OPENAI_API_KEY`	`gpt-4o`
Google	`GEMINI_API_KEY`	`gemini/gemini-2.0-flash`
Azure OpenAI	`AZURE_API_KEY` + `AZURE_API_BASE`	`azure/gpt-4o`
Ollama	(no key needed)	`ollama/llama3`
GitHub Copilot	(auto via OAuth)	`copilot/gpt-4o`

Set it via environment variable or .env file (auto-loaded):

# Pick your provider
export ANTHROPIC_API_KEY="sk-ant-..."

# Or create a .env file (retrAI loads it automatically)
echo 'ANTHROPIC_API_KEY=sk-ant-...' > .env

First-time setup

Run retrai init for an interactive wizard that helps you pick a provider, model, and API key.

Verify Installation¶

retrai --help

You should see the available commands: run, serve, tui, init, generate-eval.

Frontend (optional)¶

The web dashboard is served automatically by retrai serve. For frontend development:

cd frontend
bun install
bun run dev    # dev server on :5173, proxies /api → :8000