fix(ai-orchestration): observe agentStream output before draining

[tombeckenham]

🎯 Changes

Targets #979's branch, not main.

executeAgent awaited drainAgentStream() before result.output. When the stream fails, drain throws and the already-in-flight output promise is never observed — leaving a rejected promise unhandled. Under Node's default --unhandled-rejections=throw that terminates the worker, so an ordinary model error took down the process in addition to correctly failing the step.

Reproduced against the real workflow-core runtime before fixing:

unhandled rejections: 1 [ 'Error: output never observed' ]
events: [ 'RUN_STARTED', 'STEP_STARTED', 'STEP_FAILED', 'RUN_ERRORED' ]

The fix attaches a no-op catch to output as soon as the result is recognized, then awaits it after draining. Behaviour on the success path is unchanged.

Also adds the first test coverage for agentStream() — it was exported from index.ts but had no unit or E2E exercise, which is why this path went unnoticed. The test asserts both halves: the step fails with AgentStreamError, and no unhandled rejection escapes. Verified it fails without the source change.

✅ Checklist

• I have followed the steps in the Contributing guide.
• I have tested this code locally with pnpm run test:pr.

🚀 Release Impact

• This change is docs/CI/dev-only (no release).

@tanstack/ai-orchestration is still unreleased (0.0.0) and #979 already carries a minor changeset introducing the package. This fixes code from that same unreleased changeset, so a second changeset would only add noise. Happy to add one if you'd rather it be itemised.

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: a86acdaf-962f-44d0-84d8-5e16ad8f6dc2

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}