[DeliciousBuding]

Summary

Closes #2405. Supersedes #2406 using equivalent source changes with Conventional Commit history.

Make Account-authenticated Web conversations use one real Hub data mode and a shared assistant-ui conversation surface. Loading, empty results and failures must not switch to fixture conversations. Preserve server-owned messages while adding older history and expose only supported conversation actions.

Scope / Files / Ownership

Web entry and data adapters in app/web; shared assistant-ui runtime and components in app/shared; conversation shell in app/workbench; their tests and frontend architecture documentation. One integration branch owns this frontend change. Hub and Edge APIs, database schemas and Account OIDC credentials remain unchanged by this slice.

Interfaces / Invariants

• Existing Hub REST/WS and platform ports remain authoritative for identity, messages and execution.
• assistant-ui ExternalStoreRuntime projects server snapshots and owns chat interaction; no cloud history service is introduced.
• Human authors, agent messages, tools, approvals and artifacts retain their identity.
• A persisted message does not imply successful agent dispatch or execution.
• Unsupported mutations are unavailable; pending/failed writes cannot fabricate success.

Acceptance

• Account entry and real-data regressions; conversation creation/selection/actions; earlier history, refresh and switching isolation.
• Shared, workbench and Web unit tests/typecheck/lint; Desktop consumer checks and applicable CI.
• Browser behavior and light/dark viewport evidence for composer, history, errors and controls.
• Web production build, then application-only deployment with a verified rollback image and post-deployment checks.

Negative Constraints / Dependencies

No Account token acceptance as Hub bearer; no mock evidence relabeled as real login/model execution; no changes to daily-user data for acceptance. Depends on the existing Account-authenticated Hub API. Single-agent execution and product-session revocation are separate follow-up slices under the same delivery objective.

Validation so far

• Shared: 154 files / 2774 tests passed.
• Workbench: 163 files / 1594 tests passed.
• Web: 34 files / 283 tests passed.
• Three package typechecks passed; package lint passed with existing shared/workbench warnings.
• Desktop App integration: 14/14 passed after shared component integration.
• Browser contract: 10/11 passed, then the corrected split-host test passed independently; one earlier run also hit ERR_NETWORK_CHANGED and was rerun without retries.
• Web chat six light/dark/goal/viewport captures and Desktop chat two captures passed; Web and Desktop shell captures and geometry assertions passed. Product screenshots reviewed; no horizontal overflow.
• Build and document SSOT passed. Full CI and deployment remain pending.

Component acceptance

Dimension	Evidence
Visual	Existing theme token bridge; Web/Desktop light+dark chat and shell captures
Interaction	Real assistant-ui runtime tests for send, failed draft, stop, deduplication, thread selection/actions; browser Hub dispatch and reply hydration
Keyboard	Native semantic buttons and official thread primitives; input, search, rename Enter/Escape covered by component paths
Accessibility	Named composer/actions, alert/status feedback, code/table semantic markup; visual gates use visible composer rather than hidden measurement textarea
Responsive	1440/1024/390/320 composer browser assertions, Web768/Desktop800 shell geometry, no horizontal page overflow

Raw captures are local ignored artifacts and CI artifacts, not source. No new real Account/model or packaged Desktop claim is made by fixture verification.

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. 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

• Source mode: Auto
• Repository: TokenDanceLab/AgentHub
• Branch: refactor/assistant-ui-delivery
• File: .coderabbit.yaml
• Commit: 150d293
• Layers: Repository YAML
• Review profile: CHILL
• Plan: Advanced
• Run ID: 3aa30acb-064c-46b1-bd8d-01474b2ac2fa

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

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

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.}

[DeliciousBuding]

All seven required checks passed on 150d293 (Actions run 40833093963). The remaining workbench composer selectors now match the accessible execution-device control; focused composer behavior tests passed 9/9 with task payload and unavailable-target assertions intact. Web E2E, Web/Desktop visual QA, native Windows frontend, CSS and vulnerability gates passed. This is frontend/fixture CI evidence, not real Account/Edge execution or packaged Desktop evidence. Web-only deployment with immutable image and rollback verification follows merge.