* feat: nvidia gpu config for docker
* refactor: split NVIDIA GPU support into compose override
Address review feedback:
- Revert --engine ollama from base Dockerfile CMD (breaks non-Ollama users)
- Keep bookworm pin and OLLAMA_HOST fix in base config
- Move GPU-specific config (/proc, /sys mounts, deploy.resources.reservations)
into new docker-compose.gpu.nvidia.yml override, matching the existing ROCm
pattern (docker-compose.gpu.rocm.yml)
- Restore Ollama port to standard 11434
- Remove commented-out GPU blocks from base docker-compose.yml
- Add Ollama healthcheck with depends_on condition to base compose
- Remove run.sh from repo root
- Rewrite README Docker section to document CPU, NVIDIA, and ROCm patterns
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Jon Saad-Falcon <41205309+jonsaadfalcon@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
