repowiserepowise
Sign in
rakaarwaky/hipfire
OverviewDocsArchitectureKnowledge GraphFilesCode HealthRefactoring

People & History

CommitsContributorsDecisions
ChatPro
Stats
repowiserepowise
ExplorePricingDocs
Sign inIndex repoIndex your repo free
repowiserakaarwaky/hipfire
HI
HI

rakaarwaky / hipfire

rustperf/gfx10303d6c5e6404.8K linessynced 3h ago

Code health

5.9out of 10Fair

This codebase scores 5.9 out of 10 on defect risk, which we rate fair. It also scores maintainability 6.1 and static performance risk 9.7 out of 10. The three are scored separately and never blended into one number. The files you change most are the weak spot: 183 of 1,441 files are git hotspots, and they average 3.0.

Full health report →
Documentation1,866pages0 with model-written prose, 1,866 built from the indexScore validation19/204.03× baselineOf the 20 lowest-health files, 19 were bug-fixed in the last 6 months — 4.03× the 24% baselineDead code60exportsUnused exports that nothing in the graph reaches
Lines of code
404.8K
Files
1,441
Symbols
21,957
Modules
44
Languages
9

Recent activity

All commits →

Commits

No commits indexed yet.

Decisions

  • Add parity gate for lowered forward vs legacy hand-rolled forwardactive · 3h ago
  • Default GPU calibration (Phase 2) with forward_prefill_batch token capactive · 3h ago
  • Empirical TV threshold for distribution sanity checkactive · 3h ago
  • Override cargo fmt to prevent full-workspace formattingactive · 3h ago
  • refactor(dispatch): ship 5.2 FINAL — qwen35 residual GEMM + HFP4 qkv/qkvza -> dispatch familiesproposed · 3h ago
  • fix(dispatch): restore MoE CPU-top-K generic fallback for qwen35proposed · 3h ago

Needs attention

69 open
  • high severity. Stale decisionAdd parity gate for lowered forward vs legacy hand-rolled forwardActive decision drifting from the code it governs
  • high severity. Stale decisionEmpirical TV threshold for distribution sanity checkActive decision drifting from the code it governs
  • high severity. Stale decisionOverride cargo fmt to prevent full-workspace formattingActive decision drifting from the code it governs
  • medium severity. Proposedrefactor(dispatch): ship 5.2 FINAL — qwen35 residual GEMM + HFP4 qkv/qkvza -> dispatch familiesAuto-proposed decision awaiting review
  • medium severity. Proposedfix(dispatch): restore MoE CPU-top-K generic fallback for qwen35Auto-proposed decision awaiting review

Where the risk concentrates

All 183 hotspots →

Ranked by prior bug fixes and change frequency, mined from full git history rather than from the code alone.

FileChurnPrior fixesMaintainersCommits 90d
dispatch.rscrates/rdna-compute/src99.9th56bug magnet16200
qwen35.rscrates/hipfire-arch-qwen35/src99.8th58bug magnet9259
daemon.rscrates/hipfire-runtime/examples99.7th80bug magnet9212
kernels.rscrates/rdna-compute/src99.6th30bug magnet12283
main.rscrates/hipfire-quantize/src99.5th18bug magnet1056

Composition

Open the graph →
rust 43%markdown 27%python 13%shell 9%json 4%Other 4%

Explore this codebase

  • Docs1,866 pages across 44 modules→
  • ChatAsk this codebase a question and get an answer with its sources→
  • Files1,441 files with per-file docs, health and history→
  • ArchitectureDependency graph, layers, and 39 entry points→
  • Code healthPer-file scores, 12,935 open findings, coverage and refactoring targets→
  • Knowledge graphEntities, communities, and the paths between them→
  • Change couplingFiles that keep changing together, mined from commit history→
  • CommitsChange-risk ranked history with agent provenance→
  • ContributorsBus factor, per-file maintainers, and the human/agent split→
  • StatsSize class, origin, lifetime churn, rhythm and records→
  • CostsWhat indexing this snapshot cost, by model and by run→