Quickstart (CPU)¶

Last Updated: 2026-06-22

Get Codex running locally in under five minutes — no GPU required.

Prerequisites¶

python --version   # 3.12+
pip install -e .   # install from repo root

1 — Start the API¶

# Option A: directly
python -m codex.cli serve

# Option B: Docker Compose (recommended for isolation)
docker compose up api

The API listens on http://localhost:8000 by default.

2 — Run inference¶

curl -X POST http://localhost:8000/infer \
  -H "Content-Type: application/json" \
  -d '{"prompt": "def fibonacci(n):", "max_tokens": 64}'

Or open the demo notebook:

jupyter notebook examples/notebooks/demo_infer.ipynb

3 — Verify outputs¶

The response JSON contains completion, tokens_used, and model_version.
No GPU is required — the symbolic pipeline runs fully on CPU.

Next steps¶