Skip to content

Preserve partner-recovery progress around slow queries - #105

Merged
iskandr merged 1 commit into
mainfrom
fix/104-batched-partner-recovery
Oct 2, 2026
Merged

iskandr merged 1 commit into
mainfrom
fix/104-batched-partner-recovery

Conversation

@iskandr

@iskandr iskandr commented Oct 2, 2026

Copy link
Copy Markdown
Owner

Large mate/SA recovery queries currently lose the entire partner query's progress when one slow region causes a timeout. Add optional partner batches, query-count limits and a separate partner timeout; with on_timeout="incomplete", completed batches survive and later batches continue. Failed locations remain unresolved and never become evidence of absence. Default request/cache identities and record matching are preserved.

On the first Sid PacBio audit batch, the existing combined query timed out after 300 seconds with 404,697 retained seed records. A bounded-query pilot finished in 78 seconds with 404,699 verified records and explicit incomplete status. The audit is still tuning query sizes; this PR provides bounded progress without claiming complete template recovery.

Validation: 395 tests pass; ruff and strict MkDocs build pass; wheel/sdist build and Twine checks pass. New tests cover a timeout followed by a recoverable late mate, exact multiplicity, retry, query budgets, default cache identity, and invalid policy values. Version 0.15.1.

Fixes #104. Needed by openvax/isovar#433.

@iskandr
iskandr merged commit 000cbad into main Oct 2, 2026
18 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.

Bound large partner-recovery queries and preserve progress around slow pointer regions

1 participant