Metadata-Version: 2.5
Name: embodied-first-exam
Version: 0.1.0
Summary: Embodied First Exam on Harbor: `harbor run --core|--all|--privileged|--standard|--standard-realtime`, or `efe run`
Project-URL: Homepage, https://www.embodied-first-exam.ai
Project-URL: Repository, https://github.com/mll-lab-nu/embodied-first-exam
License-Expression: MIT
License-File: LICENSE
Requires-Python: <3.15,>=3.12
Requires-Dist: harbor==0.23.0
Description-Content-Type: text/markdown

# Embodied First Exam

Code agents program robots in simulation: [embodied-first-exam.ai](https://www.embodied-first-exam.ai),
[github.com/mll-lab-nu/embodied-first-exam](https://github.com/mll-lab-nu/embodied-first-exam).

```bash
# needs git and Docker with Compose; GPU tasks also need an NVIDIA GPU and the NVIDIA Container Toolkit
uv tool install embodied-first-exam --with-executables-from harbor
# your API keys: in the shell, or in a .env file in the working directory
# the core set; the tasks are cloned on first use (a clone you run from, or $EFE_ROOT, is used instead)
harbor run -d embodied-first-exam -a codex -m openai/gpt-6-astra
```

This installs Harbor 0.23.0 with the benchmark's flags on `harbor run` and the `efe` command. The tasks are not in
the package: see the repository's README.
