## TODO

- Author module document summarizing implemented features, deployment, and test coverage; link to `project-queues/active/parallama`.
- Validate and document NVidia multi-GPU scheduling modes (round-robin, memory-aware, hybrid) with test evidence; AMD supported but not assumed.
- Implement hardware probing and drive model placement from hardware scan (avoid static model configuration); document discovery logic.
- Publish packaging guidance: container images, explicit tags, and compatibility matrices across environments.