Skip to content

New env for embeddings clustering with fast_hdbscan - #111

Merged
julenmendieta merged 4 commits into
mainfrom
julenmendieta/MILAB-7065_embeedingsClusteringEnv
Oct 8, 2026
Merged

julenmendieta merged 4 commits into
mainfrom
julenmendieta/MILAB-7065_embeedingsClusteringEnv

Conversation

@julenmendieta

@julenmendieta julenmendieta commented Oct 2, 2026 •

Copy link
Copy Markdown
Contributor

RetriggerConfidence Score: 4/5

The PR appears safe to merge, though the Linux ARM64 wheel build should be made reproducible.

Fix All in Claude CodeFindings

  1. P2 Unpinned ARM64 build dependencies ▶
Fix with agent prompt
### Issue 1
python-3.12.10-embedding-clustering/config.json:51
The Linux ARM64 `hdbscan` wheel is compiled after the builder installs unpinned versions of NumPy, Cython, setuptools, and wheel. A later rebuild can therefore use different build inputs and produce a different wheel, undermining the new environment’s reproducibility claim. Please pin these build requirements, including a NumPy version compatible with the shipped `numpy==2.2.6`.

Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!

---

For each issue above, determine whether it is valid and should be fixed. If so, fix it directly.

Summary

The PR adds a Python 3.12.10 embedding-clustering run environment, registers it in the catalogue, and builds it for four platforms.

  • Run environment: The packaged Python runtime and wheel directory used by the block; this PR adds a dedicated embedding-clustering variant.
  • Dependency closure: The wheels supplied for offline installation; this PR explicitly pins the clustering libraries and their listed transitive dependencies and downloads them with noDeps.
  • Platform root: The packaged output for one OS and architecture; this PR adds Linux x64/ARM64, macOS ARM64, and Windows x64 roots, omitting macOS Intel.
  • Source-built wheel: A wheel compiled during the build; this PR builds hdbscan on Linux ARM64, but leaves its build requirements unpinned.
Diagram
%%{init: {'theme': 'neutral'}}%%
flowchart LR
  A[CI platform job] --> B[Build Python 3.12.10]
  B --> C[Download pinned wheels]
  C --> D{Linux ARM64?}
  D -- Yes --> E[Compile hdbscan wheel]
  D -- No --> F[Download hdbscan wheel]
  E --> G[Package platform root]
  F --> G
  G --> H[Catalogue entrypoint]
Loading

Reviews (1) · Last reviewed commit: "New env for embeddings clustering with f..."

Comment thread python-3.12.10-embedding-clustering/config.json Outdated
@julenmendieta
julenmendieta merged commit 55f7a73 into main Oct 8, 2026
152 of 154 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants