The old flow fired OpenAI on wizard open with just ticket + chatter,
so the AI summary was just a paraphrase of what the user originally
reported — your engineering analysis (scope, limitations, recommended
approach) never made it to the owner. Restructure to a two-step flow:
1. Open wizard → empty findings + empty summary, NO OpenAI call
2. You write findings: scope / effort / approach / risk
3. Click 'Generate Summary from Findings' → OpenAI runs with
ticket + chatter + findings, where the prompt explicitly tells
the model to weight findings MORE THAN the original report
4. Review/edit, then Send
Bulk wizard mirrors the flow per line: each row gets its own
findings + summary, one 'Generate All Summaries' button fans out
parallel OpenAI calls using each line's own findings.
Updated SUMMARY_PROMPT to:
- Tell the model the support engineer's findings are authoritative
- Emit a bullet structure that leads with the recommendation, not
the user's restated ask
- Side with findings over the original report when they conflict
New tests cover:
- default_get does NOT fire OpenAI (regression guard for auto-AI)
- Findings text actually reaches the OpenAI prompt
- Send works with a manually-typed summary (no AI in the loop)
- Existing bulk + validation paths still pass with the new shape
Also folds in the deferred code-review #7: ThreadPoolExecutor now
explicitly cancels pending futures on timeout via
shutdown(wait=False, cancel_futures=True) so a slow OpenAI day can't
hold the wizard open for ceil(N/workers)*15s.
Bumps fusion_helpdesk_central to 19.0.2.3.0.
Smoke-tested live on nexa: opening the wizard makes zero OpenAI calls;
clicking Generate with findings='My findings: scope is XL, ~8h' makes
exactly one call and the findings text is verifiably in the prompt
body received by call_openai_chat.
Ships the design spec at docs/superpowers/specs/2026-05-27-owner-approval-flow-design.md.
What's new on central (fusion_helpdesk_central 19.0.1.2.0 -> 19.0.2.0.0):
- Engagement model: 8 new fields on helpdesk.ticket (state, snapshotted
owner email/name, single-use UUID4 token, sent/reminded/decided
timestamps, AI summary, stored-computed turnaround hours).
- Wizard: single + bulk modes on one fusion.helpdesk.engagement.wizard
TransientModel with a child wizard.line for per-ticket bulk summaries.
default_get pulls the OpenAI summary on open; AI fan-out for bulk is
parallel via ThreadPoolExecutor (max 5 workers, 30s overall cap).
- OpenAI client in utils.py — stdlib urllib, 15s per-call timeout, every
failure collapses to '' so the wizard's manual-summary fallback fires.
- Public portal: /fusion_helpdesk/engagement/<token>/<decision> GET +
POST, four branded standalone QWeb pages (confirm/done/invalid/error).
Token is single-use, cleared on confirm. Decision posts a public
comment attributed to the resolved owner partner; chatter propagates
to the employee's My Tickets thread per the "fully visible" UX choice.
- Mail templates (single + bulk) with magic-link buttons. Bulk template
renders one card per ticket, each with its own approve/reject URL.
- Reminder cron: daily, single-shot per engagement, configurable via
fusion_helpdesk_central.engagement_reminder_days ICP (default 3, 0
disables).
- Reporting dashboard: pivot/graph/list/kanban over helpdesk.ticket
filtered to engaged ones, with avg-turnaround measure. Menu lives
under Helpdesk > Reporting > Owner Engagements.
- Client_key extended with owner_email/owner_name fields; ticket.create
upserts them from the client-side piggyback (no new sync endpoint).
- 100% coverage on utils + integration tests on wizard, controllers,
re-engagement, cron, computed turnaround. OpenAI mocked in CI.
What's new on client (fusion_helpdesk 19.0.1.7.1 -> 19.0.2.0.0):
- Two new ICP settings: fusion_helpdesk.owner_email / .owner_name with
a new "Owner Approval" block in Settings > Fusion Helpdesk.
- controllers/main.py::submit piggybacks both keys on every ticket
payload so central keeps client_key.owner_email/name fresh
automatically.
Verified live end-to-end on entech -> nexa: payload upsert, wizard with
mocked AI, action_send, portal GET/POST/GET-again cycle, second click
hits the friendly invalid-token page. Token entropy = 122 bits (UUID4).
