Update roadmap status: Phase 7's 7a design, and container-only results and self-authored Run as

[Paul Lizer (paullizer)]

Summary

This PR updates docs/explanation/features/CHAT_ORCHESTRATION_WORKFLOWS_ROADMAP.md to match paullizer-react-v2-ui at 0.261.232 (e1e6949). Docs only, one file, +60/−14, in two commits. Each commit can be taken or dropped on its own.

37c4b9a: status, Phase 7 and two follow-ups

• Status line and row 7. The status line now reads 0.261.232. Row 7 names 7a's setting, enable_chat_orchestration_workflow_handoff (off by default), and points to §6 for its design.
• Phase 7 (§6). A new "7a in progress" block lists what 7a's approved plan settles. 7a's PR will record anything that changes:
  • the gate: the new setting plus Chat Orchestration, personal workflows, Propose Workflows From Chat, Run Workflows From Chat and Use Workflow Results In Chat; WorkflowUser role; the user's own private chats only
  • the workflow_handoff step: a v3 For each blueprint over up to 25 named # documents, or a keyword workspace query over scopes the turn already authorized. Its tasks run on a model or a local agent.
  • two approvals: the plan always waits (manual floor), then the hand-off card. The step isn't a deliverable; the server adds a fixed note to the reply.
  • accept creates the workflow disabled, with a deterministic ID and a one-time origin marker, and queues exactly one durable run that 6b-1 posts back. Chat's workflow_run won't start it again.
  • a rolling 24-hour hand-off limit (default 5, range 1–100) instead of Phase 4's cap of 20
  • at most 2,000 loop items (or the admin's lower loop limit), because each item uses two of a run's 5,000 execution admissions. A query that matches more pauses the run before any item is reviewed; the user can only cancel it.
  • a narrow 6a reader branch for a hand-off run's v3 report. Other v3 runs still answer workflow_result_unsupported.
• §9 row 7. Decided for 7a: the workflow is created disabled, with a one-time marker, and is never archived or deleted automatically.
• Two new repository follow-ups, found while reviewing 6b-2 and 7a:
  • Cancelling a durable run answers 500 on a runtime conflict or outage. All four cancel routes catch only LookupError and WorkflowCancellationConflictError, but the runtime raises WorkflowRuntimeConflict and RuntimeUnavailable, which subclass RuntimeError.
  • A v3 For each loop can allow up to 5,000 items, but a run admits only 5,000 executions. A one-task loop pauses with execution_budget_exceeded after about 2,500 items, after doing that work, and nothing checks this at save.
• The stale release-notes count is now 168 at 0.261.232.

c0af9a5: align with merged #1621 and #1630

The screening stack (#1628 with #1629 and #1631, then #1632) and #1630 are merged. The roadmap still said some things V2 no longer does, so this commit updates them:

• §7. Reads and deliveries re-check the chat, workflow and run that hold a result. Since 0.261.231, the documents a run's tasks read are provenance only and aren't re-checked.
• Gotcha 36. Delivery no longer withholds a result over a source document the requester can't read.
• 6a. "Lost access" in its Done when now means access to the workflow or run.
• Gotchas 4 and 15. Since 0.261.229, a revision the Run as user saved, assistant edits included, runs as them without a separate approval. A change someone else saves, or a later change by someone else to an agent or action it uses, still asks again.
• The intro paragraph names both changes.
• §5's original design text and the shipped Phase 3 notes stay as history. Ask for M365 Run as approval only when someone else authored the revision #1630 already updated the feature docs they describe.

Sources checked

• 7a's approved plan
• M365_RUN_AS_SELF_AUTHORED_APPROVAL.md
• WORKFLOW_SAVED_RESULT_SOURCE_RECHECK_FIX.md
• CHAT_WORKFLOW_RESULTS_FOLLOW_UP.md and CHAT_WORKFLOW_RESULT_DELIVERY.md as changed by Stop re-checking saved workflow results against their source documents (screening layer 3a) #1631
• the release-note headings for .229–.232
• the follow-ups were checked in code at e1e6949:
  • the cancel routes and _request_workflow_run_cancellation in route_backend_workflows.py
  • the journal admission check in functions_workflow_journal.py
  • the run limits in functions_workflow_flow.py and functions_workflow_limits.py

Linked issue

Refs #1543 #1546 #1549 #1621

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
• deployers/version.txt bumped, or not needed because deployers/ was not changed

Testing / validation

I ran each test through its __main__ at c0af9a5 with PYTHONPATH=application/single_app;functional_tests. Each result matches unmodified e1e6949:

• python functional_tests/test_docs_app_surface_coverage.py: passed, 7/7.
• python functional_tests/test_docs_site_quality.py: passed, 6/6.
• python functional_tests/test_docs_link_integrity.py: 2/5, the same as the base. The 11 broken links are on other pages. None is in this file, and the new link to M365_RUN_AS_SELF_AUTHORED_APPROVAL.md resolves.
• python functional_tests/test_docs_release_notes_integrity.py: 0/1, the same as the base. 168 releases at 0.261.232 aren't on any generated page; the follow-up line now gives that number.
• Outside table rows, no added line is longer than 118 columns. The file is still CRLF throughout.
• scripts/build_docs_inventory.py isn't needed: no application surface changed.
• I re-fetched paullizer-react-v2-ui before opening this PR. It's at e1e6949, the base of both commits.

Documentation

• Release notes updated, or not needed
• Feature documentation updated, or not needed
• Fix documentation updated, or not needed

Security checklist

• New Flask routes include @swagger_route(security=get_auth_security()): none added
• Settings sent to non-admin frontends use sanitize_settings_for_user(): not applicable
• Browser JavaScript is served from local SimpleChat static assets only; no CDN-hosted JS: not applicable
• No secrets, keys, connection strings, or local-only artifacts are included