Repository navigation
perf(d1): omit unused award board reads - #622
Conversation
harlan-zw
added
the
harlan-agent-review
|
Agent verification, using read-only queries against production D1: The old full-board path and the named-only path returned identical 60 award sightings, including board, rank, and Skill identity. Both used Unix time 1791684954 for ranking windows.
Every statement reported zero rows written. These timings cover one execution per path, not latency percentiles. Background jobs remained active. The comparison called the actual loaders through a read-only D1 adapter. Raw SQL, bound parameters, metadata, and sightings are retained in the session scratchpad. |
harlan-github-agent
Bot
added
harlan-agent-review-required
🤖 MERGED
GitHub merged this pull request. No material findings were recorded. 01d3ec97-5832-4184-bb44-6c8c7fbe0fe9 |
harlan-github-agent
Bot
added
harlan-agent-ready
![harlan-github-agent[bot]](https://avatars.githubusercontent.com/in/4579275?s=60&v=4){kind=small}
There was a problem hiding this comment.
Harlan Agent Kit posted this automated review. AI open source policy.
Automated Review: READY.
harlan-github-agent
Bot
removed
the
harlan-agent-ready
1 participant
🔗 Linked issue
Follow-up to #620.
❓ Type of change
📚 Description
Hourly awards load Repository boards and fallback Skills, then discard them. Awards now use a shared named-Skill loader that preserves board ranking, windows, and top-starred demotion.
A read-only production comparison returned the same 60 award sightings. D1 reads fell from 99,272 to 31,634 rows, and queries fell from 24 to 12. No indexes or migrations change.