convex-performance-audit

Diagnose slow Convex subscriptions by gathering health insights and running `npx convex insights --details` to identify high bytes or documents read. Classify the bottleneck as read amplification, subscription cost, OCC conflicts, or function budget issues, then route the specific symptom to a targeted fix playbook.

OCC conflicts in Convex occur when concurrent mutations contend on overlapping read/write sets, triggering retries. Trace the complete read/write set for the scoped user flow, audit sibling writers to ensure consistent fixes, and apply guardrails to minimize transaction contention without over-engineering migrations.

Fix Convex functions nearing execution or payload limits by auditing function budgets and tracing the full read/write set of the scoped path. Route the function budget issue to the appropriate reference playbook, apply the smallest high-impact remediation, and verify correctness alongside budget improvements.

Optimize reactive invalidation-heavy Convex UI by scoping one concrete user flow and tracing its complete read/write set. Audit sibling readers to prevent piecemeal regressions, apply consistent fixes with guardrails, and verify that subscription cost and invalidation frequency are reduced.

Avoid migration-heavy changes for Convex performance bottlenecks unless clear measured signal exists from insights and a safe rollout plan is established. The audit workflow prioritizes the smallest high-impact remediation matching the symptom to reduce risk and prevent over-engineering.

Yes, you need to run `npx convex insights --details` to gather the strongest available signals before optimizing a slow Convex app. This guided performance triage prevents guessing by mapping observed signals directly to the correct problem class for targeted remediation.