Skip to content

fix(ui): include cache tokens in contextPercent calculation#70530

Closed
vmanuelort-sys wants to merge 1 commit into
openclaw:mainfrom
vmanuelort-sys:fix/context-percent-cache-tokens
Closed

fix(ui): include cache tokens in contextPercent calculation#70530
vmanuelort-sys wants to merge 1 commit into
openclaw:mainfrom
vmanuelort-sys:fix/context-percent-cache-tokens

Conversation

@vmanuelort-sys

@vmanuelort-sys vmanuelort-sys commented Apr 23, 2026

Copy link
Copy Markdown

Summary

extractGroupMeta in ui/src/ui/chat/grouped-render.ts computed contextPercent from fresh input tokens only, ignoring cacheRead + cacheWrite. With Anthropic prompt caching, most of the context window arrives as cached tokens, so input is routinely in the single digits while cached tokens make up the real load — the UI renders "0% ctx" even on near-full sessions.

Fix

Use input + cacheRead + cacheWrite as the numerator. Matches the intent of the ctx % indicator (how full the context window is), and the values are already aggregated in the same function.

const totalInput = input + cacheRead + cacheWrite;
const contextPercent =
  contextWindow && totalInput > 0
    ? Math.min(Math.round((totalInput / contextWindow) * 100), 100)
    : null;

Evidence

Live session (agent:main:main, 2026-04-23):

  • inputTokens: 3
  • cacheWrite: 93,056
  • contextTokens: 200,000
  • Before: renders 0%
  • After: renders 47%

Gates

  • pnpm check:changed (typecheck core + core-test, lint, import-cycles, webhook/pairing guards, 171 ui tests)
  • pnpm build
  • Committed via scripts/committer

@greptile-apps

greptile-apps Bot commented Apr 23, 2026

Copy link
Copy Markdown
Contributor

Greptile Summary

Fixes contextPercent in extractGroupMeta to include cacheRead and cacheWrite tokens alongside fresh input tokens, so the "ctx %" indicator accurately reflects context window usage when Anthropic prompt caching is active. The change is minimal and correct.

Confidence Score: 5/5

Safe to merge — single-line arithmetic fix with no side effects.

No logic, syntax, or security issues. The fix correctly accounts for all token categories in the context percentage calculation, and the zero-guard is preserved.

No files require special attention.

Reviews (1): Last reviewed commit: "fix(ui): include cache tokens in context..." | Re-trigger Greptile

@prtags

prtags Bot commented Apr 23, 2026

Copy link
Copy Markdown

Related work from PRtags group liked-piglet-oe6w

Title: Control UI context percent ignores cached prompt tokens

Number Title
#70530* fix(ui): include cache tokens in contextPercent calculation
#70532 fix(ui): include cached tokens in context usage
#70592 fix(ui): include cacheRead in contextPercent calculation

* This PR

This was referenced Apr 23, 2026
Merged
Closed
@steipete steipete added duplicate This issue or pull request already exists close:duplicate Closed as duplicate dedupe:child Duplicate issue/PR child in dedupe cluster labels Apr 23, 2026
@steipete

steipete commented Apr 23, 2026

Copy link
Copy Markdown
Contributor

Thanks for the fix. This is now covered by the landed #70532 / commit 6415e35.

#70532 fixes the same Control UI footer context-percent bug in ui/src/ui/chat/grouped-render.ts, using input + cacheRead + cacheWrite and adding regression coverage for both cached prompt tokens and output-token exclusion. Closing this PR as a duplicate.

@steipete steipete closed this Apr 23, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

app: web-ui App: web-ui close:duplicate Closed as duplicate dedupe:child Duplicate issue/PR child in dedupe cluster duplicate This issue or pull request already exists size: XS

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@vmanuelort-sys @steipete