fix(tui): stop streaming markdown/code after message completes

[mocksoul]

Summary

• TextPart passes streaming={true} unconditionally — even for completed messages
• This causes opentui to always skip the last table row (treated as potentially incomplete)
• Derive streaming state from message.time.completed so finished messages render all rows

Fixes #13855

Note

Full fix requires companion change in opentui (streaming setter must rebuild tables on true→false transition). See anomalyco/opentui#696

[github-actions]

Thanks for your contribution!

This PR doesn't have a linked issue. All PRs must reference an existing issue.

Please:

1. Open an issue describing the bug/feature (if one doesn't exist)
2. Add Fixes #<number> or Closes #<number> to this PR description

See CONTRIBUTING.md for details.

[github-actions]

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

Potential Duplicate Found:

• PR fix(tui): stop truncating last table row after streaming completes #13674: "fix(tui): stop truncating last table row after streaming completes"
  • fix(tui): stop truncating last table row after streaming completes #13674

Why related: This PR addresses the same issue — preventing the last table row from being truncated/skipped after streaming completes. Both PRs are fixing the same behavior where incomplete messages cause the last row to be dropped, and both likely involve adjusting the streaming state logic for completed messages.

[mapix-etnc]

Adding a +1 from a real-world user — this bug hits every table response when using OPENCODE_EXPERIMENTAL_MARKDOWN=true. The needs:issue label might be blocking — the underlying issue is #13855 (and the original report is #13539).

This PR and #13674 cover the same root fix (streaming derived from message.time.completed instead of hardcoded true). Either one would unblock users. Is there a specific reason neither has been merged yet?

[mocksoul]

Adding a +1 from a real-world user — this bug hits every table response when using OPENCODE_EXPERIMENTAL_MARKDOWN=true. The needs:issue label might be blocking — the underlying issue is #13855 (and the original report is #13539).

This PR and #13674 cover the same root fix (streaming derived from message.time.completed instead of hardcoded true). Either one would unblock users. Is there a specific reason neither has been merged yet?

I dunno. Fix in opentui part already merged in main.

[trry-hub]

Thanks for working on this fix! I was also investigating the same issue (Bug #13855) and started implementing the same solution with createMemo(() => !props.message.time.completed). Great to see you already have a PR for this!

Looking forward to this being merged! 👍