fix(tui): bound completed shell output frames

[lurenjia534]

Related Issue

Resolve #2320

Problem

Completed foreground shell commands passed their entire captured output to the terminal renderer. A command that emitted many short lines could therefore create a very large frame, causing unnecessary wrapping work and memory pressure even though only the tail is useful in the interactive view.

What changed

• Bound each completed stdout/stderr presentation tail by both line count and character count.
• Apply the same bounded presentation to live completion and session replay.
• Preserve the complete formatted output in the transcript entry so session history and exports remain lossless.
• Keep completed replay components timer-free and recalculate theme styling when the active theme changes.
• Cover large short-line output, a single long line, stdout/stderr, theme changes, and replay retention with regression tests.

Validation

• pnpm --filter @moonshot-ai/kimi-code exec vitest run test/tui/components/messages/shell-run.test.ts test/tui/message-replay.test.ts
• pnpm --filter @moonshot-ai/kimi-code exec vitest run — 2,462 passed, 5 skipped
• pnpm --filter @moonshot-ai/kimi-code typecheck
• pnpm --filter @moonshot-ai/kimi-code build
• pnpm exec oxlint --type-aware --quiet
• git diff --check
• Controlled before/after memory probe against upstream
  691ec4679ea19d6be8ac18f359088384ed3e446d and fixed
  6fb6bac22889bc0a1c2816aef437c76c3b397be5, using the real
  ShellRunComponent → pi-tui Text render path while retaining the complete
  formatBashOutputForDisplay transcript. With 50,000 high-entropy short
  lines (2,750,000 characters), five isolated node --expose-gc processes
  produced these forced-GC medians:
  • upstream: 50,000 rendered rows, 15,679,944 B retained-heap delta,
    9,850,880 B RSS delta
  • fixed: 6 rendered rows, 2,784,936 B retained-heap delta,
    2,772,992 B RSS delta
  • reduction: 82.24% retained heap and 71.85% RSS
• The same probe at 300,000 lines retained the complete transcript and reduced
  the forced-GC median by 82.60% for heap and 68.83% for RSS.
  These are post-GC deltas for the affected render path on this run, not a
  cross-platform claim about whole-process peak CLI memory.

Checklist

• I have read the CONTRIBUTING document.
• I have linked a related issue, or explained the problem above.
• I have added tests that prove my feature works.
• Ran gen-changesets skill, or this PR needs no changeset.
• Ran gen-docs skill, or this PR needs no doc update.

[changeset-bot]

🦋 Changeset detected

Latest commit: 6fb6bac

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 1 package

Name	Type
@moonshot-ai/kimi-code	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[pkg-pr-new]

pnpm dlx https://pkg.pr.new/@moonshot-ai/kimi-code@6fb6bac

npx https://pkg.pr.new/@moonshot-ai/kimi-code@6fb6bac

commit: 6fb6bac

[lurenjia534]

@codex review

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. Bravo.

Reviewed commit: 6fb6bac228

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".