convex-performance-audit

Run npx convex insights --details to identify the hot function, then replace scan-and-filter patterns with withIndex or withSearchIndex queries. For hot list pages, consider digest tables that read smaller document shapes instead of full source documents.

OCC conflicts come from write contention on shared documents or broad read sets. Narrow reads with indexes, split hot documents like counters into shards, skip no-op writes, and move non-critical bookkeeping to scheduled functions.

No. The Convex .filter() method performs the same as JavaScript filtering after a scan; you still pay for every document read. Only .withIndex() and .withSearchIndex() actually reduce the documents scanned at the storage layer.

Use point-in-time reads when the flow is high-read, the data changes less often than users need to see, and explicit refresh is acceptable. Keep reactive subscriptions for collaborative editing, live dashboards, and presence-heavy views.

Date.now() defeats Convex's query cache because results change constantly, forcing frequent re-evaluation even when underlying data is unchanged. Use a boolean field updated by a scheduled function, or pass a coarsely rounded timestamp as an argument.

Avoid structural changes when scale is small, traffic is modest, or signals are weak. A simple scan on a small table is often acceptable, and migration-heavy rollouts are only justified by measured signals or clearly unbounded hot paths.