fix(opencode): make text/reasoning delta accumulation O(N) instead of O(N²)

[hardes11]

Issue for this PR

Fixes #30067

Type of change

• Bug fix
• New feature
• Refactor / code improvement
• Documentation

What does this PR do?

Resubmission of #30058 by meitalbensinai (auto-closed for missing PR template; commit cherry-picked unchanged, authorship preserved).

Replaces text += delta accumulation in the reasoning-delta and text-delta handlers of session/processor.ts with a lazy chunked-text getter. Once anything reads .text between writes (UI render loop, NDJSON serializer, bus event broadcaster), the runtime flattens the rope string and the next += re-copies the whole buffer — O(N²) in cumulative part length. On thinking-mode models emitting 1500+ small reasoning tokens per turn, this grows per-step latency from ~6s to 100s+ by turn 50-80 (perf evidence in #30067: memmove at ~30% of CPU, GC pinned). Same shape as the fix that landed in vercel/ai for processUIMessageStream.

The getter is non-enumerable, so it does not leak through JSON.stringify/structuredClone; .text reassignment still works via the setter.

How did you verify your code works?

• Cherry-picked the original commit onto current dev and resolved the conflict in the import region (dev had since dropped an unused log variable; the incoming function is the only addition).
• Diff is one file, +48/-2; confirmed no += accumulation remains in either handler and no orphaned references were introduced.
• Ran a local harness (Bun 1.3.14) executing the exact installChunkedText implementation from this diff:
  • Functional: after 80k appended deltas, .text is byte-identical to naive += accumulation; the getter caches the joined string after first read; the setter resets the buffer and subsequent pushes accumulate correctly; _chunks is absent from JSON.stringify and object spread (non-enumerable, no leakage).
  • Performance: with flatten-forcing reads (JSON.stringify every 100 deltas), the chunked path is ~2.9x faster at 80k deltas / 1.6MB accumulated text, and the naive path re-copies the whole buffer on every append once a read forces flatten, so the gap widens with length.
• Did not build the app or run the full test suite; the harness exercises the changed mechanism in isolation, not the whole binary.

Screenshots / recordings

n/a

Checklist

• I have tested my changes locally
• I have not included unrelated changes in this PR

[github-actions]

Thanks for updating your PR! It now meets our contributing guidelines. 👍

[github-actions]

Automated PR Cleanup

Thank you for contributing to opencode.

Due to the high volume of PRs from users and AI agents, we periodically close older PRs using automated criteria so maintainers can focus review time on the most active and community-supported contributions.

This PR was closed because it matched the following cleanup criteria:

• The PR was created more than 1 month ago
• The PR had fewer than 2 positive reactions
• Positive reactions are counted as thumbs-up, heart, celebration, or rocket reactions on the PR

PRs created within the last month are not affected by this cleanup.

If you believe this PR was closed incorrectly, or if you are still actively working on it, please leave a comment explaining why it should be reopened. A maintainer can review and reopen it if appropriate.

Thanks again for taking the time to contribute.