Show one progress indicator per V2 orchestration phase

[Paul Lizer (paullizer)]

Summary

• A running plan shows one progress indicator. Once a plan is approved (manually, on its timer or automatically), the V2 chat no longer draws a "Thinking" bubble above the plan card, which was already showing the run. The card is the only indicator until the answer arrives. Run notices, such as a saved-memory warning, stay under the finished answer's reasoning steps.
• The card says what the run is doing. Its status line used to show a bare role such as "Reasoning". It now shows "Starting", then the running step's kind of work and title (for example "Gathering: Read quarterly reports"), then "Preparing the answer" once every step has settled. "Waiting for results" is unchanged.
• Planning says Planning. While the planner works, the bubble reads "Planning" instead of "Thinking", so the plan card that replaces it reads as its result.
• Impact: V2 chat orchestration users only. Ordinary chat still shows "Thinking", including a chat reply sent while a plan waits in the same conversation. Stop and the composer lock are unchanged. There are no server, API or payload changes.

Before: after Approve, a "Thinking" bubble sat above the plan card, which showed the plan's intent, 0/1, Review and a status of "Reasoning". After: only the card shows, with a status such as "Reasoning: Draft candidate names", until the answer arrives.

Root cause

A run borrows the chat store's shared streaming flag, which routes Stop to it, locks the composer and collects the run's thoughts for the answer. executeSavedPlan took it through beginOrchestrationTurn, and MessageList drew StreamingBubble whenever streaming was true. The store recorded only that something was streaming, not that it was the run, so the bubble couldn't step aside for the card.

The bubble had nothing to add anyway. The server never streams a run's answer: it sends orchestration_step frames and the occasional notice thought, then the whole answer in one content frame just before done (_finalize in functions_orchestration_execution.py). For the whole run, the bubble could only say "Thinking". The v0.261.204 fix removed the orchestration lane's progress card for the same reason, but left this placeholder.

Approach

• chatStore.ts records who owns the streaming flag. New state orchestrationSurface is 'planning', 'running', or null for a chat stream. orchestrationSurfaces goes from a Set to a Map of conversation to phase, so leaving and reopening a conversation, and a server re-key, keep the phase. beginOrchestrationTurn takes an optional trailing phase that defaults to planning, so existing callers are unchanged. sendMessage, resumeChatStream, retryMessage and editMessage set it to null.
• Why not infer it from the orchestration store. The composer only blocks on streaming, so a chat reply can be sent while a run waits for a durable result or is being reconciled. That run is still in flight on its card, so hiding the bubble whenever a run was in flight would also hide the reply's "Thinking". A new test drives exactly that case through the real controller.
• The bubble hides only while the card is drawing the run. executeSavedPlan claims the surface as running, and the new selectActiveTurnRunInFlight is true exactly when ActiveOrchestrationCard draws its running row. StreamingBubble steps aside only when both hold, and shows as usual otherwise, so a run never ends up with no indicator.
• describeRunProgress in orchestrationPlan.ts writes the status line. That module imports only types, so the new Node test runs the shipped function directly.

Notes for review

• Version renumbers and V2 merges. V2 moved twice while this PR was open, and each time this fix moved one above it. Both renumbers change only version strings (20 in the same 12 files), and both merges follow Phase 7a follow-up: re-prove a hand-off report's lineage before reading it (0.261.252) #1647's merge of Phase 6b-2: show chat-started workflow runs and their posted results in V2 #1639:
  • Show a workflow's reply once in the conversation it created (0.261.253) #1637 (0.261.253). Commit 2 renumbers this fix from 0.261.253 to 0.261.254, and commit 3 merges V2 ea0d359fc. Two files conflicted: config.py kept the version one above V2, and the release notes put this PR's section directly above V2's v0.261.253 section, so removing either section reproduces the other side's file byte for byte. MessageList.tsx merged cleanly.
  • V2 shared conversations: mention pills, agent activity lines, generated documents and media (0.261.255) #1652 (0.261.255). Commit 4 renumbers this fix to 0.261.256, and commit 5 merges V2 6154f9599. config.py keeps VERSION = "0.261.256", one above V2. In MessageList.tsx both sides changed the top of StreamingBubble. The resolution keeps this PR's store read and adds V2 shared conversations: mention pills, agent activity lines, generated documents and media (0.261.255) #1652's collaborative selector, so every hook still runs before either early return, and both early returns stay. They can't apply at once, because orchestration routes only accept personal conversations. chatStore.ts merged cleanly, and V2 shared conversations: mention pills, agent activity lines, generated documents and media (0.261.255) #1652 adds no new streaming writes, which the ownership test confirms. The release notes didn't conflict.
  • For both merges, every file only one side changed matches that side, and each file both sides changed contains exactly both changes. No earlier recorded resolution was applied.
• test_v2_reasoning_controls.py gets a one-line locator fix. Its run-correction check read the card's only role="status" element. The running card now always has a status line ("Starting"), so the locator is scoped to the reasoning notice.
• Branch location. The branch is on paullizer/simplechat, like the other recent PRs into this base.

Commits

1. 011b17640 Show one progress indicator per orchestration phase: the fix, its tests and its docs.
2. e6c21aaf6 Renumber the progress indicator fix to 0.261.254.
3. 639a5cb2a Merge V2 ea0d359 (Show a workflow's reply once in the conversation it created (0.261.253) #1637) into the orchestration progress indicator.
4. fcde4b070 Renumber the progress indicator fix to 0.261.256.
5. 054909864 Merge V2 6154f95 (V2 shared conversations: mention pills, agent activity lines, generated documents and media (0.261.255) #1652) into the orchestration progress indicator.

Linked issue

N/A: this came from direct feedback on the V2 orchestration UI, and no issue was filed.

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. It is 0.261.256, one above V2's 0.261.255. Commit 1 used 0.261.253; Show a workflow's reply once in the conversation it created (0.261.253) #1637 took that number first and V2 shared conversations: mention pills, agent activity lines, generated documents and media (0.261.255) #1652 then took 0.261.255, so commits 2 and 4 renumber this fix (see Notes for review).
• deployers/version.txt bumped, or not needed because deployers/ was not changed. Not needed: deployers/ is unchanged.

Testing / validation

At the head, 054909864, after merging #1652. These ran on the merged tree just before it was committed, and the commit contains exactly that tree.

• npm --prefix .\application\v2_ui run typecheck: passed.
• npm --prefix .\application\v2_ui run build -- --outDir ..\..\ui_tests\artifacts\orchestration-plan-editor and npm --prefix .\application\v2_ui run build: both passed.
• node .\functional_tests\test_v2_orchestration_run_progress_status.mjs (new): passed, 8/8.
• node .\functional_tests\test_v2_orchestration_streaming_surface.mjs (new): passed, 6/6, including the check that every write taking streaming names its owner.
• node .\functional_tests\test_v2_orchestration_progress_lane.mjs: passed, 6/6.
• node .\functional_tests\test_v2_collaboration_ai_activity_logic.mjs, node .\functional_tests\test_v2_collaboration_mention_pills_logic.mjs and node .\functional_tests\test_v2_drawer_media_logic.mjs (V2 shared conversations: mention pills, agent activity lines, generated documents and media (0.261.255) #1652's): all passed.
• python .\functional_tests\test_v2_stream_reconnect.py: passed, 10/10.
• python .\functional_tests\test_v2_new_chat_reset.py: passed, 6/6.
• python .\functional_tests\test_v2_stream_leave_without_cancel.py: passed, 15/15.
• python .\functional_tests\test_v2_prompt_composer_card.py: passed, 11/11, including its TypeScript checks, which call beginOrchestrationTurn with positional arguments.
• python .\functional_tests\test_v2_prompts_workbench.py: passed, 22/22.
• python .\functional_tests\test_collaboration_ai_activity_events.py and python .\functional_tests\test_collaboration_generated_documents.py (V2 shared conversations: mention pills, agent activity lines, generated documents and media (0.261.255) #1652's): passed, 6 and 5 checks.
• python .\functional_tests\test_v2_prompt_attachment_persistence.py: 84 passed.
• python .\functional_tests\test_workflow_reply_replaces_agent_posted_message.py (Show a workflow's reply once in the conversation it created (0.261.253) #1637's): passed, 10/10.
• python .\functional_tests\test_docs_site_quality.py: passed, 6/6.
• python .\functional_tests\test_docs_app_surface_coverage.py: passed, 7/7.
• python -m pytest .\ui_tests\test_v2_orchestration_streaming_bubble.py .\ui_tests\test_v2_new_chat_reset.py .\ui_tests\test_v2_orchestration_planning_retry.py -q: 27 passed.
• python -m pytest .\ui_tests\test_v2_reasoning_controls.py -q: 33 passed.
• python .\ui_tests\test_v2_orchestration_plan_card.py: passed, 7/7.
• python -m pytest .\ui_tests\test_v2_collaboration_ux.py -q (V2 shared conversations: mention pills, agent activity lines, generated documents and media (0.261.255) #1652's, which covers the shared-conversation side of StreamingBubble): 4 passed.
• python .\scripts\check_xss_sinks.py --base-sha 6154f9599 --head-sha 054909864 (git diff --name-only 6154f9599 054909864 -- application): passed for all 7 files.

The new browser tests catch the old behaviour: with only application/v2_ui/src reverted, test_v2_orchestration_streaming_bubble.py fails 4 of 5, and only the unchanged tabular test passes.

At 011b17640, before the V2 merges:

• Run one file at a time with python -m pytest <file> -q, all passed: test_v2_orchestration_approval_persistence.py 23, test_v2_orchestration_auto_open.py 5, test_v2_orchestration_conversation_context.py 5, test_v2_orchestration_model_picker.py 11, test_v2_orchestration_workflow_run_links.py 29, test_v2_workflow_run_card.py 41, test_admin_orchestration_actions.py 8, test_admin_orchestration_workflow_runs.py 9, test_admin_orchestration_workflow_proposals.py 7, test_v2_elicitation_composer.py 18, test_v2_prompt_composer_experience.py 56, test_chat_streaming_thinking_placeholder.py 1.
• Run with python <file>, the way these runner-style files are written, all passed: orchestration composer 6/6, drawer views 3/3, elicitation 6/6, pending approval resume 5/5, plan editing 5/5, run hydration 7/7.
• Of the 20 functional_tests/test_v2_* files matching orchestration, stream or tabular parity, 19 passed; test_v2_tabular_parity.py is below.

Not passing or not run, all unrelated to this change:

• python .\functional_tests\test_v2_tabular_parity.py: 13/14, still at 054909864. The failing check ("Could not locate the composer submit handler") reads only Composer.tsx, which this PR doesn't change, so it fails the same way on V2.
• These couldn't be collected in this environment: test_chat_saved_analysis.py, test_chat_three_document_smoke.py and test_chat_workflow_results.py (module 'lib' has no attribute 'GEN_EMAIL'), and test_v2_orchestration_plan_editor_backend.py and test_v2_orchestration_recovery_backend.py (No module named 'content_screening'). test_chat_saved_analysis.py calls beginOrchestrationTurn positionally; the new parameter is optional and last, and the TypeScript checks above cover the same call shapes.
• test_docs_release_notes_integrity.py and test_docs_link_integrity.py already fail on V2: the generated release-notes pages lag the source by more than 180 releases, and 11 relative links and 10 site links are broken in pages this PR doesn't touch.
• Running many orchestration UI files in one pytest process fails with "Playwright Sync API inside the asyncio loop", where the shared page fixture meets the files' own browsers. Each file passes on its own, as above.

Documentation

• Release notes updated, or not needed. A new ### **(v0.261.256)** section has one Bug Fixes entry and one User Interface Enhancements entry. It sits directly above V2's v0.261.253 section (V2 shared conversations: mention pills, agent activity lines, generated documents and media (0.261.255) #1652 added no release-notes section) and is a pure 16-line insertion against V2 6154f9599.
• Feature documentation updated, or not needed. In CHAT_ORCHESTRATION.md, "Stream events" now describes the Planning label, the run phase and the card's status line. The version header and the test table are updated too: the streaming bubble row, plus rows for the two new Node tests.
• Fix documentation updated, or not needed. New docs/explanation/fixes/ORCHESTRATION_RUN_DUPLICATE_THINKING_INDICATOR_FIX.md covers the issue, root cause, files, tests, and before and after. ORCHESTRATION_DUPLICATE_PROGRESS_CARD_FIX.md (v0.261.204) links to it.

Security checklist

• New Flask routes include @swagger_route(security=get_auth_security()). N/A: no routes are added or changed.
• Settings sent to non-admin frontends use sanitize_settings_for_user(). N/A: no settings are sent.
• Browser JavaScript is served from local SimpleChat static assets only; no CDN-hosted JS. The change is in the bundled V2 React source, and no dependencies or script sources are added.
• No secrets, keys, connection strings, or local-only artifacts are included. Build outputs (static/v2, ui_tests/artifacts) are gitignored and not committed.