ReproDeck needs a repeatable way to measure how repository intelligence and investigation setup behave as a repo gets large.
Build a deterministic fixture generator with a few size tiers and record at least:
- repository scan time;
- context selection/compilation time;
- SQLite write/read costs for the generated case;
- peak memory for the main flow where practical;
- whether limits fail closed instead of silently dropping evidence.
The benchmark should be runnable locally and in CI without downloading a real third-party repository. Document the fixture shape and measurement method before publishing any performance claims.
ReproDeck needs a repeatable way to measure how repository intelligence and investigation setup behave as a repo gets large.
Build a deterministic fixture generator with a few size tiers and record at least:
The benchmark should be runnable locally and in CI without downloading a real third-party repository. Document the fixture shape and measurement method before publishing any performance claims.