CubelaPetarandClaude Opus 4.8
b6b242196d
Follows the homelab pattern: ironicbadger.docker_compose_generator v2 renders services/<host>/NN-<stack>/compose.yml templates into ~/docker/compose.yaml on the host. - 01-vllm: chat model, fixed --gpu-memory-utilization - 02-embeddings: second vLLM instance (--task embed) rather than a separate toolchain, so SM120 support only has to be solved once - 03-openwebui: Open WebUI + pgvector (not chroma — corpus size) - 99-network: shared bridge; leading comment keeps networks: top-level - pin docker_compose_generator to 2.0.1 — galaxy tags mix v1/v2 formats - group_vars: stack config incl. LDAP placeholders still to be filled The role only writes the compose file; starting the stack stays manual. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
phy-srv-gpu01
GPU server for AI/ML workloads. Hardware is ordered/assessed; OS setup and configuration are the next step.
| IP | 192.168.66.69 (planned) |
| Ansible group | phy_srv_gpu01 |
| Status | planned — not yet configured |
Hardware
HPE ProLiant DL380 Gen12, 2× Intel Xeon 6714P (8-core, 4.0 GHz), 128 GB RAM, NVIDIA RTX PRO 6000 96 GB, 2× 960 GB NVMe SSD, redundant PSU — full BOM in HW.md.
Planning notes
- Hardware assessment (2026-07-06)
- Software assessment (2026-07-06)
- Deep dive (2026-07-07)
- Project plan for the setup (2026-07-10) — start here when the server arrives
Open pre-work items are tracked in the repo-root TODO.md.
Runbooks
- NVIDIA driver, CUDA repo & container toolkit (2026-07-14) — automated by the Ansible role
nvidia_gpu
Scripts
- share-analysis.ps1 — read-only SMB share analysis (projektplan §2.1); run on a Windows machine with read access to the shares