FE-1793: Keep experiment runs and results in their originating conversation

[kostandinang]

🌟 What is the purpose of this PR?

Keep an experiment proposal, its execution state and its result follow-up in the originating conversation. This is a complete experiment feature extracted from #9829, not a change to the simulation engine.

🔗 Related links

• FE-1793 (internal)
• Original UI PR #9829
• Conversation permissions/reset #9856
• Voice mediation #9836

🚫 Blocked by

• Permissions/reset #9856, branch ka/fe-1793-conversation-permissions.

Stack: main → #9829 → #9856 → this PR → #9836. Voice steering #9826 is independent; mediation follows as a separate comparison.

🔍 What does this change?

• Keep the host draft widget, session draft state, shared execution card and lifecycle tests together.
• Own shared experiment draft state without depending on the voice-steering module changes.
• Replace a running proposal in place with validation/progress and final metrics. Retain dismissed/cancelled records and support retry after failure.
• Retain the original blue simulation and purple optimization cards for both the built-in assistant and Brunch. Keep one shared appearance; the shared card only gains an optional Retry run action. Do not add candidate-progress plumbing or touch petrinaut-core.
• Send a completed local result as a separate turn when its originating conversation is ready. Preserve unsent text, do not automatically resume a stopped response, and offer retry if result submission fails.
• Register the follow-up control with the host and report host-owned experiment activity through the composer context.
• Keep the user guide and library patch changeset with this behavior.

Pre-Merge Checklist 🚀

🚢 Has this modified a publishable library?

• Adds a patch changeset for @hashintel/petrinaut execution cards and host activity reporting.

📜 Does this require a change to the docs?

• Lifecycle, retry and result-follow-up behavior are documented in the assistant guide. Screenshot replacements remain pending where needed.

🕸️ Does this require a change to the Turbo Graph?

• No execution-graph changes.

⚠️ Known issues

• See the current PR checks for CI on the main-based head. Earlier website and Playwright integration runs failed before tests because Docker could not pull the MinIO images (unauthorized: access to the requested resource is not authorized); this is not a result for the current head.
• No provider or microphone testing was performed.
• The permissions layer is a convenient linear review base, not a fundamental experiment-engine dependency.

🐾 Next steps

Review this feature independently from the mediation layer above it. Live voice/provider verification remains separate. This PR remains draft.

🛡 What tests cover this?

• Main-based experiments tree: all 18 Turborepo build, type-check and lint tasks passed. Both complete unit suites passed separately.
• Petrinaut: 173 files / 1,515 tests. Website: 63 files / 895 tests.
• Draft/execution tests cover proposal persistence, run/cancel/retry, result delivery, originating-conversation checks, and preservation of unsent text.
• Before the main-based separation, browser assertions passed for running simulation, running optimization and finished cards. Cancellation removes progress; the original title placement, 5px progress bars and unboxed metrics are retained. Desktop and 390px captures were inspected. The rendered component sources remain unchanged; these were not new browser runs on the current head.
• Formatting and git diff --check passed.

❓ How to test this?

1. Run turbo run build test:unit lint:tsc lint:eslint --filter @hashintel/petrinaut --filter @apps/petrinaut-website.
2. Use the deterministic draft integration test and draft-widget tests to exercise Run, Cancel, completion, failure and result retry without a provider.
3. Open Storybook's RunningSimulation, RunningExperiment and FinishedExperiment fixtures. Check the original blue/purple styling, progress, cancellation and unboxed finished metrics.
4. Verify a local result follows up only in the matching ready conversation, preserves the composer draft and does not automatically resume a stopped response.

📹 Demo

Controlled Storybook recordings using deterministic data, not a real optimizer or provider.

Running

pr-9857-running-experiment.mp4

Finished

pr-9857-finished-experiment.mp4

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
hash	Ready	Preview	Oct 1, 2026 6:29am UTC
petrinaut	Ready	Preview	Oct 1, 2026 6:29am UTC
petrinaut-docs	Ready	Preview	Oct 1, 2026 6:29am UTC

1 Skipped Deployment

Project	Deployment	Actions	Updated
hashdotdesign-tokens	Ignored	Preview	Oct 1, 2026 6:29am UTC

[cursor]

Stale Bugbot comment from a previous run.

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 2abc566. Configure here.}

[lunelson]

Thanks — all threads addressed.