Fix V2 New chat carrying over a busy conversation's state

[Paul Lizer (paullizer)]

Summary

In V2, clicking New chat while chat orchestration was planning or running did not give a clean conversation. The new chat kept the old turn's Thinking bubble and a Stop button that did nothing. It never showed its empty state and wouldn't send until the page was reloaded. V1 always starts a fresh conversation.

• Stuck streaming state: an orchestration turn sets the chat store's streaming flag but has no chat-stream controller, so detachActiveStream() never clears it. settleOrchestrationTurn doesn't clear it either once the conversation is off screen. chatStore now tracks which conversations have an orchestration turn showing as in progress. New chat clears streaming and messagesLoading. Opening a conversation shows Thinking and Stop again only if its own turn is still running in this tab.
• First-message race: a counter that advances whenever New chat starts or a conversation opens now guards the step where a first message takes over the new chat. The guard applies in sendMessage, generateImageFromReference and orchestration ensureConversation. Clicking New chat while a first message is still creating its conversation keeps the new chat empty. The message is still sent and saved in its own conversation, which appears in the rail.
• Hand-offs: workspace Chat on selected documents or a tag, and Start chatting on the home page, now start a brand-new conversation instead of reopening the last one. The documents or tag arrive as context chips.
• Unchanged on purpose:
  • Leaving a busy conversation still detaches rather than cancels, and only Stop cancels. The previous answer keeps generating and is saved.
  • Unsent composer text and chips survive New chat.
  • The rail's Chats link still returns to a reply that is streaming.

Worth a careful look: if you reopen a conversation whose orchestration turn is still running, Thinking and Stop come back and later reasoning steps appear as usual. Steps that arrived while you were away are not replayed into the live bubble. The plan card and step progress are unaffected.

Linked issue

Fixes #1617

Release Notes & Latest Features

• New Feature
• Bug Fix
• UI Enhancement
• Breaking Change
• Internal only

Is this visible to end users?

• Yes
• No

Is this admin-facing (Admin Settings, governance, deployment, config)?

• Yes
• No

Should this become a Latest Feature card?

• Yes
• No
• Already added

Screenshot needed for the card?

• Yes
• No
• Attached

Version bump

• application/single_app/config.py VERSION third segment bumped, or not needed because this is docs-only (0.261.225 -> 0.261.226)
• deployers/version.txt bumped, or not needed because deployers/ was not changed

The latest paullizer-react-v2-ui (through #1612) has been merged in. The base had already used 0.261.218 through 0.261.225, so this fix was renumbered to 0.261.226.

• config.py: set to the next free version.
• release_notes.md: every base section is kept as-is, including the base's own v0.261.218 entry, and this fix's entry sits in a new v0.261.226 section at the top.
• Test headers, the minimum version in test_v2_new_chat_reset.py, the fix doc, the feature doc and the guide now say 0.261.226.

Testing / validation

On the merged head:

• npm run typecheck in application/v2_ui: passed.
• npm --prefix .\application\v2_ui run build -- --outDir ..\..\ui_tests\artifacts\orchestration-plan-editor and npm --prefix .\application\v2_ui run build: passed.
• One combined python -m pytest run over 13 UI test files: 212 passed, 112 errors.
  • The 212 passed across 10 files: test_v2_new_chat_reset.py, test_v2_orchestration_planning_retry.py, test_v2_chat_context_selection.py, test_v2_orchestration_streaming_bubble.py, the base's new test_v2_inline_maps.py, test_v2_orchestration_plan_editor.py, test_v2_orchestration_recovery.py, test_v2_orchestration_dependency_plans.py, test_v2_orchestration_workflow_proposal_card.py and test_v2_elicitation_composer.py.
  • The 112 errors were every test in the 3 remaining suites, which check for the default SPA build. They error only when collected in one process with the other Playwright suites, as they also did before the merge. Each passes in its own process:
    • python -m pytest .\ui_tests\test_v2_reasoning_controls.py -q: 33 passed.
    • python -m pytest .\ui_tests\test_v2_orchestration_approval_persistence.py -q: 23 passed.
    • python -m pytest .\ui_tests\test_v2_prompt_composer_experience.py -q: 56 passed.
  • In total, 324 browser tests pass.
• python -m pytest test_v2_new_chat_reset.py test_v2_stream_leave_without_cancel.py test_v2_new_chat_scoping.py test_v2_chat_context_picker.py test_v2_conversation_deep_link.py test_v2_inline_maps.py test_v2_inline_media_players.py test_v2_rich_rendering.py test_v2_enhanced_citations.py -q (run in functional_tests): 79 passed, 2 failed.
  • Both failures were already failing before this PR and are unrelated: test_v2_chat_context_picker.py::test_the_composer_no_longer_carries_a_write_only_selection and test_v2_conversation_deep_link.py::test_incoming_link_is_captured_before_any_effect_runs. They cover Composer.tsx and ChatPage.tsx, which neither this PR nor the merged base touches.
• python .\functional_tests\test_v2_new_chat_reset.py: 6/6 passed.
• python .\functional_tests\test_docs_app_surface_coverage.py: 7/7 passed. python .\functional_tests\test_docs_site_quality.py: 6/6 passed.
• python .\functional_tests\test_docs_release_notes_integrity.py: 0/1, failed. This was already failing before this PR because the generated release-notes pages are stale. The base also added its v0.261.218 to v0.261.224 sections without regenerating them, so about 131 earlier versions are now stale.

Before the merge:

• With only the application/v2_ui/src changes stashed, python -m pytest .\ui_tests\test_v2_new_chat_reset.py -q gave 6 failed, 1 passed. The plain chat-stream case already worked and is kept as a regression guard. The 3 updated tests in test_v2_orchestration_planning_retry.py and test_v2_chat_context_selection.py also fail with the source changes stashed.
• Broad V2 UI regression sweep: 446 passed across 22 files with python -m pytest .\ui_tests\<file> -q. A further 39/39 passed across 7 suites run directly with python .\ui_tests\<file>, as those files' headers direct. Per-file results are below.
• Not run because of local environment problems:
  • ui_tests/test_chat_saved_analysis.py: pyOpenSSL/cryptography import error during test collection.
  • ui_tests/test_v2_orchestration_plan_editor_backend.py and ui_tests/test_v2_orchestration_recovery_backend.py: ModuleNotFoundError: No module named 'content_screening'.

The pytest runs also passed -p no:cacheprovider and a -W ignore warning filter.

Pre-merge UI regression sweep per file

python -m pytest .\ui_tests\<file> -q, all passed:

File	Passed
test_v2_orchestration_actions.py	16
test_v2_orchestration_approval_persistence.py	23
test_v2_orchestration_auto_open.py	5
test_v2_orchestration_conversation_context.py	5
test_v2_orchestration_dependency_plans.py	32
test_v2_orchestration_generated_images.py	3
test_v2_orchestration_model_picker.py	11
test_v2_orchestration_outputs.py	68
test_v2_orchestration_plan_editor.py	35
test_v2_orchestration_planning_retry.py	15
test_v2_orchestration_recovery.py	32
test_v2_orchestration_streaming_bubble.py	3
test_v2_orchestration_workflow_proposal_card.py	22
test_v2_orchestration_workflow_results_input.py	5
test_v2_orchestration_workflow_run_floor.py	9
test_v2_orchestration_workflow_run_links.py	29
test_v2_group_participant_invites.py	7
test_v2_elicitation_composer.py	18
test_v2_reasoning_controls.py	33
test_v2_prompt_composer_experience.py	56
test_v2_chat_image_uploads.py	15
test_v2_image_reference_generation.py	4

python .\ui_tests\<file>, all passed:

File	Result
test_v2_orchestration_composer.py	6/6
test_v2_orchestration_drawer_views.py	3/3
test_v2_orchestration_elicitation.py	6/6
test_v2_orchestration_pending_approval_resume.py	5/5
test_v2_orchestration_plan_card.py	7/7
test_v2_orchestration_plan_editing.py	5/5
test_v2_orchestration_run_hydration.py	7/7

Documentation

• Release notes updated, or not needed (Bug Fixes entry under v0.261.226)
• Feature documentation updated, or not needed (chat hand-off section of V2_GROUP_DOCUMENT_BROWSING.md, and docs/guides/manage-group-workspaces.md)
• Fix documentation updated, or not needed (docs/explanation/fixes/V2_NEW_CHAT_RESET_FIX.md)

Security checklist

• New Flask routes include @swagger_route(security=get_auth_security()) (N/A: no routes added or changed)
• Settings sent to non-admin frontends use sanitize_settings_for_user() (N/A: no settings payloads changed)
• Browser JavaScript is served from local SimpleChat static assets only; no CDN-hosted JS (no new browser assets)
• No secrets, keys, connection strings, or local-only artifacts are included (build outputs and the test harness bundle are git-ignored)