feat(task): report task cache statistics

[jdx]

Summary

• add --task-cache-stats for an opt-in per-run cache summary
• report artifact lookup hit rate, uncompressed restored bytes, and recorded execution time saved
• persist compatible metrics in cache manifests while defaulting older entries to zero
• exclude freshness skips from cache lookup totals and document the behavior

Why

Cache hits were individually visible, but users could not assess the aggregate benefit of caching across a run.

User impact

Users can run mise run --task-cache-stats <task> to receive a summary such as Task cache: 1/1 hits (100%), 10 B restored, 6.4ms saved. Existing cache entries remain readable; older entries contribute zero bytes and time until rewritten.

Validation

• cargo test stats -- --nocapture
• mise run test:e2e e2e/tasks/test_task_artifact_cache
• mise run render
• mise run lint
• cargo clippy --workspace --all-features --all-targets -- -D warnings

AI-assisted — Tool: Codex; model: unavailable/unavailable; version: unavailable.

---

Note

Low Risk
Additive CLI and manifest fields with backward-compatible defaults; behavior is observability-only and does not change cache keying or restore logic beyond recording metrics.

Overview
Adds --task-cache-stats to mise run for an opt-in end-of-run summary of experimental task output cache lookups: hit rate, uncompressed bytes restored, and execution time saved from manifest metadata.

Cache manifests now record restored_bytes (archive + captured log bytes) and execution_duration_ns when entries are written; restores surface those fields on hits and older entries deserialize with zeros. TaskExecutor aggregates per-run hits/misses and metrics; freshness skips that never hit the artifact cache are excluded from totals (Task cache: no lookups).

The flag conflicts with --dry-run; docs, usage spec, man pages, parity tracker, and e2e coverage are updated.

^{Reviewed by Cursor Bugbot for commit ac6b27b. Bugbot is set up for automated code reviews on this repo. Configure here.}

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Central YAML (base), Organization UI (inherited)

Review profile: CHILL

Plan: Pro Plus

Run ID: 2a0e3817-7ffa-4fc7-90fe-4084e7571635

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit ede2f60. Configure here.}

[greptile-apps]

Greptile Summary

This PR adds opt-in aggregate task-cache statistics and persists compatible byte and execution-duration metadata in cache manifests.

• Tracks cache hits and misses across task execution.
• Reports restored bytes and recorded execution time saved.
• Keeps older cache manifests readable through zero-valued defaults.
• Updates generated CLI documentation and end-to-end coverage.

Confidence Score: 5/5

The PR appears safe to merge.

No blocking failure remains.

Important Files Changed

Filename	Overview
src/cli/run.rs	Adds the CLI flag and prints aggregate cache statistics after task result processing, including on failed runs.
src/task/task_cache.rs	Persists backward-compatible restored-byte and execution-duration metadata and returns it with cache hits.
src/task/task_executor.rs	Accumulates cache lookup statistics and records command execution duration when storing successful results.
e2e/tasks/test_task_artifact_cache	Covers hits, misses, freshness skips, failed tasks, and the dry-run flag conflict.

_{Reviews (2): Last reviewed commit: "feat(task): report task cache statistics" | Re-trigger Greptile}

[github-actions]

Instruction counts

benchmark	trend	instructions	Δ	wall (min)	Δ
env	▁▁▅▅▄▄▂▃▄▄▂█▃▄	84,768,161 → 84,817,059	+0.06%	18.30 → 18.24ms	-0.35%
hook-env	▃▁▆▇▅▅▂▄▃▄▆█▄▇	87,107,664 → 87,190,487	+0.10%	27.66 → 19.55ms	-29.33%
ls	▁▂▆▅▆▆▄▅▄▅▅█▆▆	76,738,032 → 76,754,461	+0.02%	15.51 → 16.49ms	+6.31%
registry	▅▃▁▁█▇▇████▄▂▅	50,296,101 → 50,369,232	+0.15%	10.91 → 11.78ms	+8.05%
startup	▄▄▄▄▄▄▄▄▄▄▄▇▁█	16,054,361 → 16,171,578	+0.73%	7.81 → 8.61ms	+10.29%

No instruction-count regression above 1%.

_{Only instruction counts gate. Wall clock is shown for context — on identical hardware it moves 4-20% run to run.}

_{Measured by tak — instruction-counted CLI benchmarks, stored in this repository's git notes.}

_{ac6b27b187f8 vs aec80d623194 · measured on the runner, not pushed to the history.}