[bot] Cut a semsql release including the semanticsql.berkeleybop.io default

[kltm]

[bot] Opened by a Claude Code agent on behalf of @kltm.

The builder.py download default was repointed to https://semanticsql.berkeleybop.io (with a non-urllib User-Agent, since the CDN returns 403 to Python-urllib) — closing #110 — but that change is on master and not yet released: PyPI semsql latest is still 0.4.0. Until a release is cut, semsql download and anything depending on semsql still hit the raw S3 bucket.

Ask

Cut a new semsql release to PyPI containing the CDN default so downstream consumers pick it up. This unblocks the migration tracked in #112.

Related: #110, #112, INCATools/ontology-access-kit#880.

— Posted by Claude Code agent on behalf of @kltm.

[github-actions]

Summary

This is a straightforward "cut a release" ask, already assigned to @caufieldjh. Context on the linked issues, plus a couple of things worth flagging for the release itself:

Status of linked issues

• Provide sqlite via a different URL #110 (Provide sqlite via a different URL) — closed, completed via PR Download SQLite databases from the semanticsql CDN instead of S3 #113, which repointed the semsql download default to https://semanticsql.berkeleybop.io and added a non-urllib User-Agent (the CDN 403s on the default Python-urllib UA).
• Eliminate S3 bucket access #112 (Eliminate S3 bucket access) — open, tracks turning off public access to the bbop-sqlite S3 bucket once enough traffic has migrated to the CDN. This release is a prerequisite: semsql can't drive traffic off S3 until the CDN default actually ships to PyPI.
• Reduce S3 egress costs for bbop-sqlite database downloads ontology-access-kit#880 (Reduce S3 egress costs for bbop-sqlite database downloads) — open, OAK-side companion effort. Its "Phase 1" (decouple the download URL into a configurable base) is conceptually the same move Provide sqlite via a different URL #110 just made in semsql. Worth keeping OAK's phase-1 rollout in sync with this release so both libraries agree on the CDN as default.

Verified current PyPI state: semsql on PyPI is indeed still 0.4.0, uploaded 2025‑02‑05 — confirming the CDN-default commit (merged via #113, currently on master at 8350307) has not gone out yet.

How the release mechanism works here (from .github/workflows/pypi-publish.yaml): publishing is triggered by creating a GitHub Release (on: release: types: [created]); the workflow then derives the package version from git describe --tags --abbrev=0 and overwrites pyproject.toml's version before building/publishing. So cutting the release is just: tag master (e.g. v0.4.1) and publish a GitHub Release from it — no manual version bump needed in the repo itself.

One thing to consider bundling: #97 (Add graph column to main schema) was explicitly deferred with "put it into the schema for the next release in the minor series" — if that's still desired, it may be worth checking whether it should land before/alongside this cut rather than in a follow-up release.

Since this requires pushing a tag / creating a GitHub Release (repo-admin/PyPI-trust action), I'm leaving that step to a maintainer rather than acting on it here.

[kltm]

[bot] Posted by a Claude Code agent on behalf of @kltm.

Update: with raw access retired 2026-08-28 (#112), this release request became a bug fix — semsql==0.4.0's built-in download URL now returns AccessDenied, so semsql download is broken for all PyPI installs until a release ships the CDN default.

— Posted by Claude Code agent on behalf of @kltm.