Keep model prereleases out of package distribution

.github/workflows/distribute.yml

@@ -21,6 +21,8 @@ on:
 
 jobs:
   resolve:
+    # Model-only prereleases must not reach PyPI, Homebrew, Spaces, or ComfyUI.
+    if: ${{ github.event_name != 'release' || (startsWith(github.event.release.tag_name, 'v') && !github.event.release.prerelease) }}
     runs-on: ubuntu-latest
     # Thirty attempts may spend up to 20 seconds each in the JSON request,
     # simple-index request, and inter-attempt delay. Leave bounded headroom.

.github/workflows/publish.yml

@@ -10,6 +10,8 @@ permissions:
 
 jobs:
   publish:
+    # Model-only prereleases are not Python package releases.
+    if: ${{ startsWith(github.event.release.tag_name, 'v') && !github.event.release.prerelease }}
     runs-on: ubuntu-latest
     environment: pypi
     steps:
@@ -38,4 +40,4 @@ jobs:
       # is needed. uv's uploader also accepts Metadata-Version 2.5 -- the permanent
       # fix that lets the hatchling pin be relaxed (see pyproject [build-system]).
       - name: Publish to PyPI
-        run: uv publish
+        run: uv publish

docs/release-and-distribution.md

@@ -102,7 +102,15 @@ catalog, paths, hashes, embeddings, and feature caches are outside this flow.
 The workflow is manual and requires the model-freeze release tag plus an
 explicit `publish` confirmation. Keep model-freeze GitHub releases marked as
 pre-releases so a model artifact cannot replace the latest package release in
-GitHub's release UI.
+GitHub's release UI. Package publication and distribution jobs run only for
+non-prerelease `v*` tags, so model-only prereleases cannot enter PyPI or the
+package fan-out.
+
+The newer `wiltodelta/openai-google-image-source-classifier` uses the separate
+`publish-image-source-hf.yml` workflow. Its release asset is only the
+hash-verified derivative `openai-google-source-v1.npz`; the workflow uploads
+that artifact plus `docs/image-source-hf/README.md` and `metrics.json`. The
+training corpus does not enter the library checkout or this release.
 
 If a distribution job fails because a repository or Hugging Face credential is
 invalid, rotate the corresponding GitHub secret and rerun the failed job. A