Skip to content

v0.16.0

Latest
Latest

Choose a tag to compare

View all tags
@jay-m-dev jay-m-dev released this 03 Nov 23:32
Immutable release. Only release title and notes can be modified.
v0.16.0
d71cb3e
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

What's changed

  • Upgrade Motoro from v0.13.0 to released v0.15.0 and delegate model discovery, connection checks, inference connections, and tool-support metadata to its provider API.
  • Scope model metadata caches to each user and credential revision, coordinate refreshes, and retain successful listings during transient provider outages.
  • Preserve provider sampling metadata, defaults, and provenance through model selection, factorial configurations, previews, publication, and execution. Unknown support permits explicit values; unadvertised defaults remain provider-controlled.
  • Freeze sampling requests and runtime provenance in new published experiment versions, preventing later discovery changes from altering published runs.
  • Show actionable run-preparation validation errors and safer credential-configuration errors.
  • Clarify the develop integration / main release workflow and required CI checks.

Upgrade notes

  • Rebuild the API and both workers from this release so all processes use the locked Motoro v0.15.0 dependency. Run the normal backup and migration sequence; this ASAREE comparison adds no product schema migrations, but Motoro's migration chain must still run.
  • Older published experiment versions retain their original sampling semantics and require their original runtime for replay. To run with the new sampling policy, resolve any saved capability override prompts and republish the experiment draft. Historical publications are not rewritten.
  • Saved draft capability rules without recorded provenance require choosing Use automatic metadata or Keep saved override in the Model inspector before publishing.
  • Model metadata caches use a new namespace; no manual Redis cache deletion or database backfill is required.
  • Ollama's Local provider base URL uses its OpenAI-compatible endpoint, such as http://HOST:11434/v1.

Validation

  • The release-candidate backend and frontend CI checks passed.
  • Both required checks also passed on the merged main release commit d71cb3e04f0d3229633ade5af27ea54dace9e5f8: release CI. Release promotion is PR #29; the Motoro tag update is PR #41.
  • 243 focused provider, connection, discovery, model-configuration, and sampling-snapshot tests passed locally, including the Local-model sampling payload regression and exact installed Motoro release provenance.
  • The maintainer reported the manual smoke-test checklist passed before authorizing release promotion.
  • Candidate CI reports 1,866 regression tests passed / 3 skipped, 14 statistical validation tests passed, and 569 frontend tests passed.
  • Nonblocking findings remain: mypy reports 640 errors in 52 files under the existing warn-only policy; frontend lint reports 15 warnings and no errors; npm audit reports 23 vulnerabilities (1 low, 10 moderate, 11 high, 1 critical). Review dependency audit findings before production deployment. uv also reports the unchanged, yanked mako==1.4.0 dependency.

Full changelog: v0.15.0...v0.16.0

Assets 3
Loading