Uploaded image for project: 'AI Platform Core Components'
  1. AI Platform Core Components
  2. AIPCC-11630

[opendatahub-io/base-containers] Document build-time index switching pattern for pyproject.toml parity

XMLWordPrintable

    • False
    • Hide

      None

      Show
      None
    • False

      [4031434683] Upstream Reporter: Lalatendu Mohanty
      Upstream issue status: Open
      Upstream description:

      With #119, #120, and #121 resolved, base-containers images will pre-configure index URLs so that consumers use plain version pins in pyproject.toml. Document this pattern so consumers know what to do and what to avoid.

      *What to document:*

      • Which env vars the base images set (`UV_EXTRA_INDEX_URL`, `UV_INDEX_URL`, `PIP_INDEX_URL`)
      • How downstream overrides them via `--build-arg` to point at AIPCC
      • Why plain `torch==X.Y.Z` resolves to CUDA wheels in both contexts (PyTorch index in midstream, AIPCC in downstream)
      • What consumers must avoid putting in pyproject.toml:
      • No `[[tool.uv.index]]` blocks for accelerator indexes
      • No `+cu128` / `+rocm` local version suffixes
      • No `[and-cuda]` extras
      • No direct wheel URLs

      Parent: #77

              Unassigned Unassigned
              upstream-sync Upstream Sync
              Antonio's Team
              Votes:
              0 Vote for this issue
              Watchers:
              1 Start watching this issue

                Created:
                Updated: