Greptile Summary

This PR fixes a display bug in formatTokensCompact() where the cache hit rate denominator used totalTokens (which includes output tokens and can be undersized for legacy rows), producing impossible percentages like "120% cached". The fix prioritises inputTokens + cacheRead + cacheWrite as the denominator when inputTokens is present — this exactly equals the prompt-side token count in Anthropic's API model. For legacy sessions that predate inputTokens tracking, it falls back to Math.max(totalTokens, cacheRead + cacheWrite), ensuring the denominator never drops below the known cached-prompt tokens.

• Logic is correct: inputTokens (non-cached prompt tokens) + cacheRead (cache hits) + cacheWrite (new cache creations) = total prompt tokens, so the denominator is properly scoped to prompt-side usage only.
• No division-by-zero risk: the cache section is guarded by cacheRead > 0, so promptTokensFromParts (and both fallback branches) are always > 0.
• Good defensive validation: each field is individually checked for Number.isFinite and >= 0 before use.
• Tests added: two targeted regression tests cover both the legacy clamping path and the new inputTokens-present path with concrete expected percentages.
• No behaviour change outside the cache hit rate display line.

Confidence Score: 5/5

Safe to merge — the fix is narrowly scoped to a display-only calculation, is mathematically sound, and is covered by regression tests.

No P0 or P1 findings. The denominator logic is correct (prompt-side tokens only), edge cases (zero inputTokens, legacy undefined, NaN/Infinity) are all guarded, division-by-zero is impossible inside the existing cacheRead > 0 guard, and two new regression tests validate both code paths.

No files require special attention.

_{Reviews (1): Last reviewed commit: "fix(status): cap cache hit rate at 100% ..." | Re-trigger Greptile}

Merged via squash.

• Merge SHA: acf8470
• formatTokensCompact() now uses prompt-side token fields (inputTokens + cacheRead + cacheWrite) as denominator
• Legacy fallback: max(totalTokens, cacheRead + cacheWrite) prevents >100%
• CHANGELOG entry added
• 64 status tests passing, typecheck clean

🤖 We're reviewing this PR with Aisle

We're running a security check on the changes in this PR now. This usually takes a few minutes. ⌛
We'll post the results here as soon as they're ready.

Progress:

• Analysis
• Finalization

Latest run failed. Keeping previous successful results. Trace ID: 019d3c7792f0af6f0c5cb559c709b01f.

_{Last updated on: 2026-03-30T09:47:27Z}