Skip to main content

Performance And Cache Policy

Known Hotspots​

Current heavy areas include:

  • run detail assembly
  • enriched result reconstruction
  • export generation
  • report generation
  • large run execution state

Results Reader Model​

The backend reads normalized result tables and reconstructs API views on demand through ResultsReader.

This is better than blob storage, but it can still be heavy for large runs.

Cache Policy​

The app currently applies aggressive no-cache headers through NoCacheMiddleware in app/main.py.

That is safe for correctness, but expensive for immutable completed-run reads.

Current Safe Reuse​

The backend already reuses:

  • report.html
  • export.zip

Those are the correct first artifact-level reuse points.

Next Performance Direction​

The next likely wins are:

  • optimize completed run detail/enriched reconstruction
  • reduce large repeated payload assembly
  • add selective caching only for terminal-state data if product rules allow it