fix(opencode): reject empty compaction summaries

[vladislav-miroshnikov]

Issue for this PR

Closes #41571

Type of change

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

What does this PR do?

A compaction model can finish with reasoning but no text summary. V1 currently treats that as success, which can hide the previous history behind an empty summary boundary.

This marks the summary as errored and stops before tail, replay, auto-continue, or the compacted event — so a fresh failed attempt cannot filter away earlier messages. It's limited to the safety invariant; #41663 explores the larger structured-input alternative.

How did you verify your code works?

• cd packages/opencode && bun test test/session/compaction.test.ts — 54 pass, 1 skip, 0 fail
• cd packages/opencode && bun typecheck
• Changed-file Prettier and git diff --check

The regression drives a reasoning-only processor response and verifies the error state, stopped result, absent compacted event, and retained history.

Checklist

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

[github-actions]

The following comment was made by an LLM, it may be inaccurate:

Based on my search, I found one potentially related PR:

Related PR:

• fix(session): reconcile compaction summary with preserved tail (PR fix(session): reconcile compaction summary with preserved tail #28067)
  • This PR also deals with compaction summary handling and seems to address interactions between compaction summaries and preserved tail messages, which is related to the history preservation aspect mentioned in PR fix(opencode): reject empty compaction summaries #42063.

All other results either returned the current PR itself (#42063) or are unrelated to compaction summary handling.

[OldKrab]

Really miss that fix!

[josephtingiris]

missed many i have fixed all these in #41663 ... entropy is fast ... conflicts already heading down the wrong path

[vladislav-miroshnikov]

@josephtingiris
#41663 closes #38801, not #41571 I validated that summaryText() is unchanged there, so the empty/no-text summary guard isn't in place. The processor change does stop Compacted from firing for a reasoning-only finish:"stop", but history is dropped by filterCompacted based on the summary message being error-free with a finish set, not on the event, so the head is still hidden, and finish:"length" is not covered either. The two PRs are complementary on different issues so if #41663 lands first I will rebase and adjust the guard.

[josephtingiris]

thx. it closes many others too just haven't had the time to link it all up here. hoping for help from the community. deterministic is more reliable in the long run.

[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.