What problem does it solve?
Performance optimization in Rspack requires repeated cycles of code changes, correctness testing, CI validation, and CodSpeed benchmark interpretation, which is tedious and error-prone when done manually across many rounds.
Core Features & Use Cases
- Autonomous Optimization Rounds: Runs up to 10 rounds of scoped code changes, local correctness checks, independent diff review, push, and CI/CodSpeed validation without asking for routine confirmation.
- CI and Review Triage: Classifies GitHub check results, reruns flaky jobs, fixes related failures, and resolves Copilot and human review comments before trusting benchmark data.
- Persistent Progress Tracking: Maintains a single fixed PR comment logging every round, commit SHA, change summary, and CodSpeed delta, plus a terminal
@codspeedbot explain request when the final result improves.
- Use Case: A maintainer asks to improve a specific Rspack benchmark by 5% within one crate; the agent iterates optimization commits, validates correctness locally, waits on 10-minute CI follow-ups, and stops only when CodSpeed confirms the threshold or the round limit is reached.
Quick Start
Use $rspack-perf-codspeed-goal to optimize the specified Rspack benchmark to the requested percentage threshold within the given crate or plugin scope.
