[currentbits]

Part of #1531.

A Claude thread that has been idle for more than an hour, past the prompt-cache TTL, with at least 50k context tokens now shows a composer strip: "Prompt cache expired; the next send re-reads the whole context." It has a Compact first (≈N tokens) button that runs the existing native compact (handleCompact). The strip only appears when native compact is available and the thread isn't archived.

• The decision is a pure helper, coldCompactTokens in src/contextRing.ts. It uses usage.contextTokens, which already includes cache_read and cache_creation.
• Limits:
  • "Idle" is measured from thread.updatedAt, so a non-run change such as a rename restarts the hour.
  • The check runs on render, so a thread left open shows the strip on its next re-render rather than the minute it goes cold.
  • The 50k threshold is a constant.

Tests are in test/contextRing.test.ts and test/composerSlashActions.test.tsx.

🤖 Generated with Claude Code

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-09T12:14:47.452547Z	9ff32b3	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 9ff32b3ff0

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Derive the timeout from the actual cache tier

For Claude sessions whose usage reports ephemeral_5m_input_tokens, this hard-coded hour suppresses the warning for 55 minutes after the cache has already expired. Five minutes is the platform default and one hour is an optional extended tier (Anthropic prompt-caching documentation); the runner currently discards the usage.cache_creation tier breakdown when producing SessionUsage, so these sessions are indistinguishable here. Preserve the observed tier/TTL and use it in this calculation rather than assuming every Claude session receives the extended cache.

Useful? React with 👍 / 👎.