Repro: open the engagement wizard on a ticket, write findings, click
'Generate Summary from Findings'. Notification: "Ticket N no longer
exists" and the whole dialog closes — even though the ticket clearly
exists in the DB.
Root cause was two compounding bugs:
1. action_generate_summary returned an act_window dict with
res_id=self.id to "stay open after writing the summary field". The
web client honoured that by opening a NEW act_window — and the new
action's context inherited active_id=<wizard_id> (because that's
the res_id of the action being opened). Wizard ids are not ticket
ids, but our default_get didn't know the difference.
2. default_get read ctx.get('active_id') unconditionally, without
first checking ctx.get('active_model') == 'helpdesk.ticket'. So
when active_id pointed at the wizard's own id, default_get fed
that to _default_get_single, which raised "Ticket <wizard_id> no
longer exists" — and the user saw a confusing error about a
ticket that obviously DID exist (just not with that id).
Two fixes:
(a) action_generate_summary + action_generate_all_summaries now
return True. The form field write is visible to the client via
the call response; the wizard re-renders with the new
ai_summary populated. No spurious navigation, no context
pollution.
(b) default_get only consults active_id / active_ids when
active_model is helpdesk.ticket. Explicit
default_ticket_id[s] context keys still take precedence and
aren't gated by active_model (they're the caller's strong
signal).
Verified live: opening the wizard with active_id=99999 and NO
active_model no longer raises 'Ticket 99999 no longer exists' —
just creates the wizard cleanly. The normal flow (default_ticket_id
+ active_model='helpdesk.ticket') still works as before.
Bumps fusion_helpdesk_central to 19.0.2.3.3.
Previous fix (col=1 on the group) didn't work — Odoo still rendered
the group's string as a left-column label inside the form sheet's
flow, so the textarea got pushed into a narrow right column. The
summary field looked entirely missing because its content split
between the button row (on the right) and the textarea (collapsed
nowhere visible).
Right idiom (lifted from Odoo's own mail.compose.message wizard):
WIDE textareas live directly at the form level, not inside <group>.
Section titles use <separator string="…"> which renders as a
horizontal divider with the label above. The textarea then takes
the full sheet width naturally.
Same pattern applied to bulk mode for consistency.
Also moved Personal Note into the top compact group with Owner /
Owner Email since it's a one-line input that belongs with the
header info, not pretending to be a wide section.
Bumps fusion_helpdesk_central to 19.0.2.3.2.
Add action_test_connection — a read-only connectivity/schema check that
reports source row counts and imports nothing, the safe first step before
a dry-run. Wire a "Test Connection" button on the wizard. Document the
end-to-end run in the README: least-privilege read-only DB role SQL, the
fusion_billing.nexacloud_dsn system parameter (libpq DSN = NexaCloud's
URL minus +asyncpg), and the Test → dry-run → real-run flow. Refresh the
stale SCAFFOLD status. 53/53 green on odoo-trial.
User reported the rich text editor showing raw HTML tags as literal text
instead of rendering them as formatted prose. Root cause: Odoo's Editor
delegates content insertion to setElementContent() (web/core/utils/html.js),
which only takes the innerHTML branch when the content was flagged as safe
markup via owl's markup() helper. Plain strings fall through to the
textContent branch, which is what the user was seeing:
<p>Ask the client if the stairlift has power. Check:</p> <ul> <li>...
instead of the rendered paragraph + list.
The canonical html_field.js in @html_editor wraps its value with markup()
before passing it to the Wysiwyg config; I missed that detail.
FIX
- import markup from @odoo/owl
- in wysiwygConfig getter, wrap the saved content_html string with
markup() before assigning to config.content
- pass markup("") for empty content (avoids editor confusion with falsy)
- load-bearing comment to keep future refactors from re-introducing the bug
VERIFIED
- upgrade clean
- 7 stale asset bundles flushed, container restarted, login serves 200
- new bundle 014fee9 renders 10029808 bytes
- node --check PARSE_OK
- compiled bundle contains: content:rawHtml?markup(rawHtml):markup("")
which is exactly the markup-wrapped path the Editor wants
Bumped to 19.0.2.2.4.
Co-authored-by: Cursor <cursoragent@cursor.com>
The Findings and Summary fields rendered at half-width because their
enclosing <group> defaulted to col="2" — Odoo reserves a label column
even when the field has nolabel="1", so the textarea was squeezed
into the right half of the dialog while the left half sat empty.
Switch both groups to col="1" so the field uses the entire group
width. Also tag both fields with widget="text" explicitly (it was
inferred from the Text field type, but being explicit makes the
intent obvious to anyone reading the view) and migrate the button
row to a flex div so the helper text aligns with the button vertical
center.
Bumps fusion_helpdesk_central to 19.0.2.3.1.
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.
REGRESSION FROM b22bb11b (Wysiwyg integration).
While inserting the new Wysiwyg methods (wysiwygConfig getter, onWysiwygLoad,
onToggleSource) between setup() and the existing onMounted / onWillUnmount
hook calls, I accidentally closed setup() early with the new
`this.wysiwygEditors = {};` assignment. That left the original
`onMounted(async () => {...});` and `onWillUnmount(...);` calls dangling
INSIDE the class body but OUTSIDE any method - which is invalid JS.
JavaScript's class-body parser sees the bare `onMounted(async () => ...)`
and tries to interpret it as a method declaration where `onMounted` is the
name and the parens are the parameter list. `async () => {...}` is not a
valid parameter, so the bundle fails with:
Uncaught SyntaxError: Unexpected token '('
web.assets_web.min.js:28807
That single parse failure tanks the entire backend asset bundle, leaving
users with a completely blank screen on /odoo (and any other backend
route). Frontend bundle was unaffected.
FIX
Move the onMounted / onWillUnmount calls back inside setup() where they
belong. Add a load-bearing comment explaining why they must stay there so
this regression cannot silently come back during a future refactor.
VERIFIED
- line 51: setup() opens
- lines 87, 93: onMounted, onWillUnmount calls INSIDE setup
- line 142: _initDrawflow as a normal class method (outside setup)
- upgrade clean
- bundle 10029245 bytes, exactly one onMounted( occurrence in
FlowchartDesigner class body
- node --check on the freshly-rendered web.assets_web.min.js -> PARSE_OK
Bumped to 19.0.2.2.3.
Co-authored-by: Cursor <cursoragent@cursor.com>
Resolves findings from the post-build review:
- C1: a partial import was indistinguishable from success. action_run_import
now logs failed rows at ERROR (survives nexa's log_level=warn) and the
wizard shows red/amber banners with failed/skipped counts.
- H3: an unrecognized billing_cycle silently fell back to monthly (wrong
plan AND price). Now raised per-row -> failed[], never silently mis-billed.
- M5: a NULL plan price silently became a $0 line. Prices now preserve
NULL-vs-0.0; a missing price for the subscription's cycle is failed[].
- H2: post-connect query/schema errors now become a clean UserError, not a
raw SQL traceback (matches the connection-error path).
- M4: per-row failures now record the exception type and log a traceback.
- MED#3: charge plan_id set explicitly False so re-runs re-assert the
shadow-safe NULL even if it was changed between runs.
- HIGH-edge: re-run only rewrites x_fc_* on existing subs; partner_id/plan_id/
line are set at creation only (never rewrite immutable fields).
- account_link: partner email match is now case-insensitive (=ilike) to avoid
duplicate partners against a differently-cased pre-existing partner.
Shadow-safety invariant unchanged and re-confirmed. 52/52 green on odoo-trial.
Adding the 'Fusion Helpdesk Central' block to General Settings so the
three ICP keys the engagement flow reads are configurable from a real
form instead of forcing admins to open Technical → System Parameters.
Three settings, all wired via config_parameter= so the existing read
paths (engagement_wizard, _fc_send_engagement_reminders) keep working
unchanged:
- fusion_helpdesk_central.openai_api_key (password widget — doesn't
render plaintext on the form)
- fusion_helpdesk_central.openai_model (default 'gpt-4o-mini')
- fusion_helpdesk_central.engagement_reminder_days (default 3, 0
disables the reminder cron entirely)
Bumps fusion_helpdesk_central to 19.0.2.2.0.
Find under Settings → Fusion Helpdesk Central. The block has two
sub-sections: "Owner Approval — AI Summary" (key + model) and
"Owner Approval — Reminder Cadence" (days).
fusion.billing.import.wizard backfills NexaCloud into Odoo: read-only
psycopg2 reader (_read_nexacloud_rows, DSN from ir.config_parameter)
split from pure-Odoo writes (_import_rows/_do_import) so the logic is
unit-tested headless. Maps users→partners+links (reusing
_resolve_or_create_partner, stashing stripe_customer_id), plans→a
cpu_seconds charge catalog (included_quota=cpu_seconds_quota,
unit_batch=3600, $0.0075/core-hour, plan_id NULL), and deployments→one
DRAFT shadow sale.order per deployment with the flat price set
explicitly. Shadow-safe by construction: draft + no payment token +
charge plan_id NULL (rating cron is a no-op). Idempotent re-runs;
per-row savepoints isolate bad rows; dry-run rolls back. 11 tests,
50/50 green on odoo-trial.
price_per_unit was a Monetary field, so a realistic sub-cent rate like
$0.0075/core-hour was rounded to $0.01 on write, corrupting the rate.
Make it Float(16,6). Also stop _compute_billable from rounding the
overage amount to 2 decimals mid-calc — that lost the half-cent on
sub-cent rates and would drift against the source app, which keeps
usage amounts at 4 decimals and only rounds at the invoice total.
Now rounds to 6 dp (float-noise only); cent-rounding defers to the
invoice line. Exposed while building the NexaCloud importer.
Adds a one-click 'loop the owner into the chatter' shortcut on the
ticket form — separate from the engagement approval flow, just keeps
the owner in the loop on ongoing communication.
What's new on helpdesk.ticket:
- x_fc_owner_display (computed Char): 'Kris Pathinather <kris@…>',
read live from fusion.helpdesk.client.key so a change to the owner
contact reflects immediately on every existing ticket.
- x_fc_owner_email_resolved (computed Char): email-only slice, drives
view visibility (the field + button only render when an owner is
configured).
- x_fc_owner_is_follower (computed Boolean): True when a partner with
the owner email is in message_partner_ids. Swaps the button for a
green 'Following' badge when the owner is already on the thread.
- action_add_owner_as_follower(): find-or-create the owner partner by
email and message_subscribe. Idempotent — second call is a no-op,
no duplicate partner. Raises UserError with a clear message if no
owner is configured.
View extension on the helpdesk ticket form: injects right after the
existing partner_id ('Customer') field in the customer side group,
so it reads as 'Customer | Owner Contact [Add as Follower]' — same
row, no layout shift when the state flips to 'Following'.
Tests cover the compute display in three states (configured,
no-client-label, no-owner-on-key), the action's three paths
(create-and-subscribe, reuse-existing-partner, idempotent-when-
already-following), and the UserError when nothing is configured.
Smoke-tested live on nexa: ticket with x_fc_client_label='ENTECH'
displays 'Kris Pathinather <kris@enplating.ca>'; first click adds
res.partner #723 to followers and flips owner_is_follower to True;
second click is a no-op.
Bumps fusion_helpdesk_central to 19.0.2.1.0.
Smaller UX simplification on the client side: the owner is already a
contact in entech's address book, so picking one is faster + safer than
re-typing their email and name (and avoids typos creeping into the
approval-email To: header).
What changed:
- Entech settings: drop fhd_owner_email + fhd_owner_name char fields;
add fhd_owner_partner_id Many2one to res.partner exposed in the
same "Owner Approval" block as a single partner selector. Quick-create
+ create-and-edit kept enabled so admins can spin up a new partner
inline if the owner isn't already in the system.
- controllers/main.py::_read_config: derives owner_email + owner_name
from the selected partner via the new _resolve_owner_contact helper.
Missing / dangling partner id → blank email + name → central simply
won't see the keys and the Engage button stays disabled (correct
"not configured" behaviour).
- Nexa side: ZERO changes. Still receives owner_email + owner_name
strings on the ticket payload, still upserts client_key.owner_email/
name. The partner abstraction stops at the entech boundary.
- migrations/19.0.2.1.0/post-migration.py auto-resolves the legacy
fusion_helpdesk.owner_email ICP value to an existing res.partner
(lowest-id match on lowercased email), writes the new
fusion_helpdesk.owner_partner_id key, and deletes the obsolete
owner_email + owner_name ICP rows so a future reader doesn't trip
over stale config.
Verified live on entech: kris@enplating.ca → res.partner #2308 ("Kris
Pathinather"), legacy keys purged, controller._resolve_owner_contact
returns the expected (email, name). The piggyback payload is unchanged
so existing client_key sync continues to work without a central
redeploy.
Bumps fusion_helpdesk to 19.0.2.1.0. fusion_helpdesk_central stays at
19.0.2.0.0 (no central-side changes required).
Replace the plain <textarea> in the flowchart designer's node-editor
right-panel with Odoo 19's native rich text editor so admins write
formatted prose / lists / bold / links / inline images without typing
HTML tags. The raw <textarea> stays available behind a toggle for the
power-user case (pasting markup from elsewhere, debugging).
CHANGES
manifest:
- depends += 'html_editor' (provides @html_editor/wysiwyg)
- bumped to 19.0.2.2.1
components/flowchart_designer/flowchart_designer.js:
- import { Wysiwyg } from '@html_editor/wysiwyg'
- import { MAIN_PLUGINS } from '@html_editor/plugin_sets'
- register Wysiwyg in static components
- state.sourceMode boolean (default false = rich text mode)
- wysiwygConfig getter builds the EditorConfig for the SELECTED node;
onChange reads editor.getContent() and writes back into the same
selectedMeta.content_html the rest of the designer already uses,
so the save path is unchanged
- onWysiwygLoad(editor) captures the editor instance per dfId so the
onChange callback can resolve the right one when nodes switch
- onToggleSource flushes the current editor's content before flipping
modes so unsaved keystrokes don't get lost
components/flowchart_designer/flowchart_designer.xml:
- replaced <textarea>...</textarea> with a conditional block:
sourceMode == false -> <Wysiwyg t-key="'wysiwyg-' + selectedNodeId"
config="wysiwygConfig"
onLoad="onWysiwygLoad.bind(this)"/>
sourceMode == true -> <textarea class="font-monospace" rows="10"/>
- t-key forces the editor to re-mount with the freshly-selected node's
content; otherwise switching nodes would keep showing the first
selected node's HTML
- new toolbar row above the editor has a "HTML Source" / "Rich Text"
toggle button (eye / code icons) so the user can flip at will
- hint text updated to reflect what each mode supports
components/flowchart_designer/flowchart_designer.scss:
- widened the right editor panel from 320px to 360px to give the
Wysiwyg toolbar room to breathe
- new .fr-wysiwyg-shell rule frames the embedded editor with the same
border + background as the other form-controls in the panel, with
a min-height of 180px and max-height 320px so it scrolls when the
content grows. Pins .o-we-toolbar inside the shell so it stays in
view as the user scrolls long content.
The save path, the runtime renderer, and the data model are unchanged -
content_html is still sanitised HTML stored on fusion.repair.flowchart.node.
Verified on local westin-v19:
- upgrade clean (no errors, no warnings)
- login serves 200 after restart
- 4 stale asset bundles flushed; Drawflow JS still served 46KB at
/fusion_repairs/static/src/lib/drawflow/drawflow.min.js
- Wysiwyg export confirmed at
/usr/lib/python3/dist-packages/odoo/addons/html_editor/static/src/wysiwyg.js:25
- MAIN_PLUGINS export confirmed at plugin_sets.js:103
Bumped to 19.0.2.2.1.
Co-authored-by: Cursor <cursoragent@cursor.com>
One-time, re-runnable, read-only importer that backfills NexaCloud
customers/plans/deployments into Odoo as a shadow copy for dual-run
reconciliation. Locks the brainstorming decisions: per-deployment
granularity, flat+overage billing, cpu_seconds metric, CPU-only v1,
Odoo-side psycopg2 reader, and shadow-safety by construction (draft
subs + no payment token + charges with NULL plan_id).
Findings from the post-feature code review on commit 396170b4. Addresses
the two CRITICAL + one HIGH + two MEDIUM issues; rest are deferred.
CRITICAL #1 — magic-link token race:
Two near-simultaneous POSTs on the same /engagement/<token>/approve
could both SELECT state='pending' under READ COMMITTED, both post
chatter, and let the last writer flip the outcome. Now the POST path
does an atomic UPDATE helpdesk_ticket SET token=NULL WHERE token=%s
AND state='pending' RETURNING id — the loser gets no row back and
renders the friendly invalid-link page. Verified live: 2 concurrent
POSTs → 1 wins, 1 loses, exactly 1 chatter row.
CRITICAL #2 — reminder cron without per-row savepoint:
Per CLAUDE.md rule #14, a DB failure mid-loop aborts the whole
transaction and silently kills the rest of the batch. Wrap each row's
send_mail+write in `with self.env.cr.savepoint()`. Also corrected the
success-count log (was len(stale), now actual sent count).
HIGH #3 — turnaround pivot summed instead of averaged:
fields.Float defaults to SUM aggregator; meaningless for per-ticket
decision delays. Added aggregator='avg' so the pivot reads "avg
turnaround per ticket" not "summed wait time".
HIGH #4 — added test_concurrent_claim_only_one_wins regression test
that fires two real HTTP POSTs against the same token and asserts
exactly one wins + exactly one approval chatter row exists.
MEDIUM #6 — cron nextcall pinned to 09:00 tomorrow so reminders land
in business hours regardless of when the module was last upgraded.
MEDIUM #10 — escalate failed owner-partner-create from WARNING to
ERROR (via _logger.exception) since silent attribution to the bot
account is a real audit-trail confusion.
Deferred (follow-up commits): #5, #7 (executor cleanup), #8, #9,
#11–#14 — none are bugs, all spec-drift or hardening.
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).
Two big workflow additions:
1. Visual drag-and-drop flowchart designer (Drawflow) + card-by-card runner
(with show-whole-tree toggle) so admins build per-(category, symptom)
decision trees with embedded photos/videos and CS walks callers through
them on the phone. Resolved-on-call closes the repair; escalated copies
the full transcript into internal_notes so the dispatched tech sees what
was already tried before they arrive at the client.
2. Vendor + draft-PO + factory-tracking on the part-order capture. Tech on
the phone with the factory picks the vendor from contacts, types the OEM
part #, cost, ETA date (calendar widget), factory ticket #, RA #, ticks
under_warranty, and the system auto-creates a draft purchase.order with
the right product (looked up or created from OEM) + activity for the
office on the ETA day + client email with ETA prominently shown and
cost intentionally omitted.
NEW MODELS
fusion.repair.symptom.class - lookup table (category + name + code).
Replaces the flat x_fc_issue_category Char on repair.order. Seeded with
7 stairlift symptoms + lighter coverage for hospital bed / porch lift /
lift chair. Equipment Class added to fusion.repair.product.category
(this carried over from the Bundle 10 plan).
fusion.repair.flowchart + .node + .edge - design-time graph.
- flowchart has name, category, symptom, version, published flag,
canvas_layout (Drawflow JSON), node_ids, edge_ids, computed start_node
- node has node_type (question / suggestion / info / outcome),
content_html, media_ids (M2M ir.attachment for photos + videos),
is_start, outcome_kind (resolved / escalate / order_part),
canvas_x/y for Drawflow round-trip
- edge has source, target, label, sequence - supports N-ary branching
(not just Yes/No)
- designer_load() and designer_save(payload) RPC API the OWL component
consumes; save is atomic-replace + bumps version + soft-validates
fusion.repair.flowchart.run + .step - runtime sessions.
- One run per repair, audited; runtime_start_or_resume() returns the
existing in-progress run or creates a fresh one for the matching chart
- runtime_choose(edge_id, cs_note) records a step + advances current_node
- runtime_complete(outcome) snapshots final node + calls _apply_outcome:
resolved -> auto-close via action_repair_start + action_repair_end,
set x_fc_resolved_on_call, post transcript to chatter
escalated -> prepend transcript to repair.internal_notes so the tech
sees it first when they open the form
order_part -> chatter note; tech opens visit-report wizard next
abandoned -> just save transcript
- Each step snapshots node_name + chosen_label at write time so the
transcript survives later chart edits without breaking.
REPAIR.ORDER EXTENSIONS
- x_fc_symptom_class_id (M2O) - new structured symptom field
- x_fc_resolved_on_call (Boolean, tracked) - true after a resolved outcome
- x_fc_flowchart_run_ids + x_fc_flowchart_run_count
- action_start_troubleshoot() - opens the runner client action, raises a
helpful UserError if no symptom set or no published chart exists
- action_view_flowchart_runs() smart button
- x_fc_issue_category renamed string to "(legacy)" - kept for back-compat
+ AI prompt context; new intakes set the M2O
DRAWFLOW DESIGNER (OWL)
static/src/lib/drawflow/drawflow.min.{js,css} - vendored Drawflow 0.0.59
(MIT). Loaded only in web.assets_backend, ~48KB total.
components/flowchart_designer/flowchart_designer.{js,xml,scss}:
- Client action "fusion_repair_flowchart_designer" with full drag-drop
canvas + zoom + pan
- 4 custom node templates color-banded by type (question blue,
suggestion green, info gray, outcome red/green/amber per outcome_kind)
- Right-panel editor for selected node: title, type, outcome kind,
content (HTML), media uploader (drag-drop or click), set-as-start
toggle, per-outgoing-edge label editor
- Save serializes Drawflow JSON to canvas_layout + atomic-replaces the
structured node/edge rows via the designer_save RPC
CARD RUNNER (OWL)
components/flowchart_runner/flowchart_runner.{js,xml,scss}:
- Client action "fusion_repair_flowchart_runner"
- DEFAULT MODE: card-by-card. One big card per node, embedded photos +
inline <video controls>, answer buttons sized for phone use, CS note
textarea (saved as cs_note on the step), running transcript at the
bottom
- TOGGLE: "Show Whole Tree" loads the same Drawflow lib in read-only
fixed mode, imports the canvas_layout JSON, highlights current node
yellow / visited green via .fr-current / .fr-visited classes
- Outcome buttons drive the right runtime_complete() call; success
notifications + auto-return to the parent repair form
- "Abandon & Escalate" header button at all times - transcript is saved
even on bail-out so the dispatched tech still benefits
PART ORDER + VENDOR PO
repair.part.order new fields:
vendor_partner_id (M2O res.partner, is_company domain), purchase_order_id
(auto-created draft PO), product_id (auto-resolved or created),
unit_cost (Monetary) + currency_id, internal_po_ref, factory_ticket_ref,
factory_ra_number, under_warranty.
action_create_draft_po() - resolves product.product by OEM (default_code)
or creates a new one in a "Spare Parts" product.category, creates a
purchase.order in draft state with one line (product + qty + price_unit
+ date_planned from expected_date or +7d), stamps Westin's internal PO
ref as partner_ref so the factory can find it on return. Office reviews
and confirms via the normal Odoo flow.
_schedule_eta_activity() - schedules a Repair: Assign Technician activity
on the parent repair.order due on expected_date, assigned to
repair.user_id, so the office is reminded to call the client and book
the return visit on the day parts arrive.
VISIT-REPORT WIZARD PARTLINE EXTENSIONS
Same new fields exposed inline on the partline list so the tech captures
everything on the phone with the factory in one form:
vendor_partner_id (vendors-only filter), unit_cost + currency,
expected_date (calendar widget) replacing expected_lead_days as the
preferred input, under_warranty, internal_po_ref, factory_ticket_ref,
factory_ra_number, create_draft_po (default True - auto-builds PO on
submit when vendor + cost are both set).
CLIENT EMAIL TIGHTENED
email_template_parts_ordered:
- Subject now includes ETA "Parts ordered for your stairlift - expected 2026-06-06"
- Hero ETA panel: large blue-bordered card with "Expected Arrival" label
and the date in 24px bold
- Cost INTENTIONALLY OMITTED - "Our office will call you to confirm a
return visit time. If you have any questions about pricing or
scheduling, please reach out to our office directly."
- "There is nothing for you to do right now." callout
UI
- repair.order form header: new "Start Troubleshooting" button (info
style, sitemap icon, visible when state in (draft, confirmed,
under_repair) AND symptom is set)
- repair.order form intake row: x_fc_symptom_class_id picker filtered to
the category, x_fc_resolved_on_call display when true
- repair.part.order form: header button "Create Draft Purchase Order"
+ new Vendor / Cost / Warranty group + System group with the PO link
- Intake wizard equipment line: symptom_class_id picker
- New menus:
Configuration > Symptom Classes
Configuration > Troubleshooting Flowcharts
Fusion Repairs > Troubleshooting Sessions (run history)
SECURITY
18 new ACL rows for the 6 new models, scoped Manager-full / User-read /
FieldTech-read. Flowchart runs and steps get write access for User so CS
can record steps; Manager owns flowchart + node + edge CRUD.
POST-MIGRATION (19.0.2.2.0)
Existing installs: walks all distinct (category, x_fc_issue_category) text
pairs on repair.order, creates a placeholder fusion.repair.symptom.class
per pair (or reuses an existing match by code/name), back-fills the new
x_fc_symptom_class_id M2O. Idempotent + safe to re-run.
DEPENDENCY
Added 'purchase' to depends (action_create_draft_po needs purchase.order).
VERIFIED END-TO-END on local westin-v19 (Margaret persona, 0 bugs):
STEP 0 seed: chart v1 8 nodes / 12 edges / published, 7 stairlift
symptoms, stairlift class=lift_elevating
STEP 1 CS creates RO-202605-60 with symptom Not Moving
STEP 2 Start Troubleshooting -> client action tag returned
STEP 3 walk run: Power on? Yes -> Seatbelt? Yes -> Swivel? Yes ->
outcome 'Still not moving - dispatch technician'
(outcome_kind=escalate)
STEP 4 runtime_complete('escalated') -> internal_notes prepended with
CS troubleshooting summary
STEP 5 visit-report parts_needed with vendor Handicare + cost $425 +
warranty + factory refs -> PART-00008 created + draft
PO 26690 auto-built with line "Handicare 1100 control
board" qty 1 @ $425, partner_ref WH-2026-1042
STEP 6 mark_ordered -> client email queued (NO cost mentioned, ETA
shown prominently) + office activity scheduled for
2026-06-06
STEP 7 fresh resume returns same run; resolved outcome auto-closes the
repair (state=done, x_fc_resolved_on_call=True)
Bumped to 19.0.2.2.0.
Co-authored-by: Cursor <cursoragent@cursor.com>
End-to-end spec for the owner-approval feature on fusion_helpdesk +
fusion_helpdesk_central. Captures data model, engagement flow (single +
bulk), magic-link approval portal, OpenAI summary, reminder cron,
reporting dashboard, edge cases, and test plan. Ready for the
writing-plans skill to turn into an implementation plan.
The OWL dialog used <t t-out="m.body"/> on message bodies, but t-out
escapes plain strings — it only renders raw when the value is a Markup
instance. Bodies arrive over JSON-RPC as plain strings (Markup is a
client-side type, doesn't cross the wire), so the customer was seeing
literal "<p>This has been fixed.</p>" in the thread instead of the
rendered HTML.
Wrap incoming bodies in `markup()` at the boundary (openTicket +
sendReply call sites) so the template renders them as the sanitised
HTML the central chatter already produced. Trust is fine — the body is
sanitised server-side by mail.thread before it ever leaves nexa.
Bumps fusion_helpdesk to 19.0.1.7.1.
Three coordinated changes on top of the section grouping:
1. **Mark as Critical** — a red chip on the New tab sets priority='3'
when submitted. The central post-create hook auto-applies a "Critical"
helpdesk.tag (shipped via fusion_helpdesk_central data XML, noupdate=1
so support can recolor without losing it on upgrade), giving support
a kanban-groupable signal that doesn't rely on remembering what
priority='3' means. Scoped to in-app-channel tickets only, so a
support agent manually setting Urgent on their own ticket isn't
silently tagged.
2. **KPI cards above the sections** — Total / Open / Closed / Critical
in a 4-up grid (auto-collapses to 2x2 under 540px). Each card uses
its own saturated gradient so it reads on both light and dark mode —
the dialog backdrop is irrelevant because the gradient brings its
own background. Counts are computed in JS from state.tickets so they
always match what's rendered below.
3. **Colored stage pills** — red Critical, green Solved, dark-yellow New,
orange Cancelled, blue for In Progress / Testing / On Hold. Critical
priority gets a *separate* red pill alongside the stage pill so you
keep stage info even on escalated tickets. Stage matching is
substring-based (lowercased) so a renamed "Resolved" or "Done" stage
on central still maps to the green pill.
Tests cover the new is_critical=True → priority='3' wiring and the
default omission so SLA / stage defaults keep working for normal
tickets. Bumps fusion_helpdesk to 19.0.1.7.0 and
fusion_helpdesk_central to 19.0.1.2.0. End-to-end smoke test verified
live: priority=3 + x_fc_client_label triggers the Critical tag.
The flat write_date-sorted list was hard to scan with 50+ tickets — solved
ones were intermixed with active ones, and there was no signal for
priority. Bucket each ticket server-side into 'critical' (open + priority
High/Urgent), 'solved' (stage marked fold=True on central) or 'open'
(everything else), and render three labelled sections in the dialog with
sticky headers, count badges, and per-group accent colours. Backend keeps
its write_date desc order so latest is always at top within each bucket.
Bucketing uses helpdesk.stage.fold (not the stage name) so renaming
"Solved" to "Done" on the central won't quietly mis-categorise rows.
Adds bucket_ticket() in utils.py with unit tests covering the
folded-wins-over-priority precedence and the missing-priority fallback.
Also surfaces a small Urgent (triangle) / High (arrow) icon on each row
so a critical ticket reads at a glance even after a user scrolls past
the section header.
Bumps fusion_helpdesk to 19.0.1.6.0.
The customer-followup ship left two papercuts that hid 51 historical
tickets from the entech owner:
1. group_reporter_admin had zero members on install — the new XML record
created the group but never granted it. Extend base.group_system's
implied_ids so every system administrator transparently inherits the
admin view of the embedded inbox on install / upgrade. (4, id) tuple
is additive — never replaces base's existing implications.
2. Tickets created before this feature shipped had NULL
x_fc_client_label and NULL partner_email, so the scope filter
excluded them all. The reporter identity was still recoverable from
the description HTML's diag block. Backfill recipe is captured in
CLAUDE.md so future deployments can apply the same one-shot UPDATE
without re-deriving the regex.
Bumps fusion_helpdesk to 19.0.1.5.0. Verified live on entech: all six
base.group_system members now return True for
has_group('fusion_helpdesk.group_reporter_admin').
Adds two Integer fields to res.partner:
- x_fc_default_lead_time_min_days
- x_fc_default_lead_time_max_days
Set once on the customer's Plating Defaults tab (Fulfilment group);
auto-copies onto every new Express Order via the existing
_onchange_partner_id hook. Operator can still override per-order
since the onchange only fills when the wizard field is still blank.
Field declaration lives in fusion_plating_configurator (alongside
the rest of the partner cascade reads). View edit lives in
fusion_plating_invoicing where the Plating Defaults tab already
hosts the other partner-level defaults (invoice strategy, deposit
%, delivery method, deadline-days). Invoicing depends on
configurator, so the fields are registered before the view loads.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Three related fixes on the Express Orders totals card:
1. Totals card now breaks out Subtotal / Tax / Tooling Charge /
Grand Total. Previously the "Subtotal" and "Grand Total" rows
both read from total_amount (same value rendered twice) and no
tax was shown at all. Customers on a fiscal position-mapped
tax rate (Ontario HST, etc.) had their taxes silently dropped
from the preview.
2. tooling_charge now feeds the Grand Total. The total_amount
compute previously summed line subtotals only. Added a real
SO line for the tooling charge in action_create_order so the
eventual sale.order.amount_total matches the preview AND the
invoice carries a "Tooling Charge" line item.
3. tax_ids is now visible as an optional column on the lines
list. Operator can see + override the auto-applied tax per
line. Default still comes from FP-SERVICE product mapped
through partner.property_account_position_id (fiscal position).
New compute fields on fp.direct.order.wizard:
- total_subtotal (sum of line.qty * line.unit_price, pre-tax)
- total_tax (sum of line + tooling taxes via compute_all)
- total_amount (subtotal + tax + tooling — was just subtotal)
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Durable: nexa/entech upgrade commands, central service-account Contact
Creation prerequisite, backup-outside-addons-path gotcha, smoke-tests-must-
call-the-controller lesson. Plus current deploy status + the one remaining
step (browser confirmation of My Tickets / New on entech).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Squash-merge of feat/helpdesk-customer-followup. The billing and
fusion_login_audit work from that branch is already on main (landed
separately); this lands only the helpdesk feature.
- Identity keystone: submit() forwards partner_email/partner_name/
x_fc_client_label so the central Helpdesk find-or-creates the customer
partner and subscribes them as a follower (enables reply emails + magic link).
- Embedded in-app 'My Tickets' inbox: server-side scoped read/reply RPC
endpoints, per-user seen tracking (fusion.helpdesk.ticket.seen), systray
unread badge. Defense-in-depth scope domain + _norm_email normalisation
(wildcard emails cannot widen scope).
- fusion_helpdesk_central: x_fc_client_label field + list/search views +
branded acknowledgement email template.
- Deployed and smoke-tested live: nexa central 19.0.1.1.0, entech client
19.0.1.4.1 (requires Contact Creation on the central service account).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
When the login string resolves to an existing user and the password is
wrong, BOTH overrides used to write a failure row:
- _check_credentials wrapper: result=failure, reason=bad_password
- _login wrapper (catching the propagating AccessDenied): result=
failure, reason=unknown_user
Discovered in production smoke on westin-v19 after the deploy: a
single failed login for info@gsafinancialconsulting.com produced two
audit rows (one bad_password, one unknown_user). The unknown_user
label was wrong — the user IS in the system.
Fix: _login now checks whether the login string resolves to any user
BEFORE writing the unknown_user row. If yes, _check_credentials
already logged the attempt and _login skips. If no, the user lookup
in super() failed and _login is the only chance to log.
Regression test test_login_known_user_bad_password_single_row asserts
exactly one row per attempt and that the row carries bad_password
(not unknown_user) when the user exists.
30 tests green locally; production smoke on westin-v19 confirms:
one row per failed login, bad_password, IP 172.18.0.1 captured.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Capture in the plan the Odoo 19 gotchas discovered during execution
that the original plan template missed:
- Test command requires --http-port=0 --gevent-port=0 (running
container holds 8069).
- Declarative models.Constraint / models.Index (T2).
- res.users.groups_id renamed to group_ids (T3, T6).
- ir.rule groups is additive not restrictive (T3).
- mail.template inline-template ctx IS env.context (T11).
- ir.cron has no numbercall field in 19 (T12).
- registry.cursor() in tests is TestCursor; cr.commit() raises;
use savepoints (T13).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Asserts the smart-button and Login Activity tab fields are stripped
from res.users get_view() for non-admin users, and present for
Settings admins. Locks down the contract behind the
groups="base.group_system" XML attributes on the form-inheritance
view (the inherited view record cannot carry groups itself per
CLAUDE.md rule #11; the gate must live on the inner nodes).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
5-min cron processes up to 100 pending rows per pass: private IPs
short-circuit to state=private_ip; same-IP cache (30 days) avoids
duplicate ip-api.com calls; reverse DNS via socket with 1.5s timeout;
HTTP lookup respects ip-api''s X-Rl rate-limit header. Tests cover
private-IP shortcut, cache hit (no HTTP), and internal-state skip --
no network calls needed.
Per-row isolation uses cr.savepoint() instead of cr.commit() because
Odoo 19 TestCursor raises AssertionError on commit/rollback. Recorded
the gotcha as CLAUDE.md rule #14.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds _fc_retention_gc() that deletes rows older than the configured
horizon (default 365 days; 0 = keep forever). Registered as a daily
ir.cron. Tests verify both the delete path and the "keep forever"
short-circuit.
Also documents the Odoo 19 gotcha that ir.cron dropped the numbercall
field (the legacy "-1 = run forever" pattern now raises ValueError at
install time; just omit the field).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Mail template + helpers (_fc_alert_*, _fc_recent_failure_count,
_fc_send_failure_alert) wired into _check_credentials so that crossing
the consecutive-failure threshold within the window queues exactly one
mail.mail per attempted login per 60-minute cooldown. Master switch
x_fc_login_audit_alert_enabled honoured. Recipients are members of
base.group_system with a non-empty email and share=False; the
__system__ superuser is excluded by Odoo''s default user filter.
Tests (3 new, 22 total green):
test_failure_burst_queues_one_email
test_cooldown_suppresses_second_alert
test_alert_disabled_master_switch
setUp ensures base.user_admin has an email (fusion-dev''s admin user
ships without one; the only user with an email is __system__, which
is filtered out of standard res.users searches).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Four x_fc_* fields on res.config.settings backed by ir.config_parameter:
retention_days (default 365, 0 = forever), alert_threshold (5),
alert_window_min (15), alert_enabled (True). New "Login Audit" block
on the General Settings page (gated by base.group_system on the block,
NOT on the inherited view record per CLAUDE.md rule #11).
CLAUDE.md gotchas added during this task:
#5 Boolean config_parameter fields don't round-trip "False" as a
string — IrConfigParameter.set_param deletes the row on falsy.
Test with assertFalse, never assertEqual(..., "False").
#6 ir.ui.view uses group_ids (Odoo 19 rename mirrored from res.users).
Setting groups_id on an ir.ui.view record raises ValueError at
install. (The XML attribute groups="..." on inner nodes is
unrelated and still works.)
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
List, form, and search views for fusion.login.audit, plus a "Login
Events" full-history action and a "Failed Logins (24h)" pre-filtered
action. Both surface under Settings -> Technical -> Login Audit
(menu items gated by base.group_system). Views are no-create / no-edit
/ no-delete to enforce append-only at the UI layer too.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds four x_fc_* fields on res.users: login_audit_ids (One2many),
login_audit_count (compute), last_successful_login (compute, stored),
last_login_ip (compute, stored). action_fc_view_login_audit returns
a window action scoped to the current user. View inheritance adds a
smart button to the button box and a "Login Activity" page to the
notebook, both gated by base.group_system on the inner XML nodes
(NOT on the view record — Odoo 19 forbids that; see CLAUDE.md rule #11).
Tests (2 new, 18 total green):
test_computed_last_successful_login — uses registry cursor to commit
the audit row so the stored compute picks it up across the
TransactionCase boundary.
test_action_view_login_audit_returns_window_action — smart-button
action shape + domain scoping.
CLAUDE.md rule #11 added: inherited ir.ui.view records cannot have
groups/group_ids on the record; the gate must be on the inner XML nodes.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Overrides res.users._login. When the login string does not resolve to
any user, super() raises AccessDenied; we record a row with user_id=NULL
and failure_reason="unknown_user", then re-raise. Closes the gap where
typo'd or scanned logins would otherwise vanish from the audit trail.
The existing _fc_record_login_event helper writes through an independent
registry.cursor(), so the audit row survives the rollback that follows
the re-raised AccessDenied.
Note: in Odoo 19 _login is a plain instance method (not the classmethod
it was in earlier versions) and takes (credential, user_agent_env). The
original plan was written for the classmethod signature; corrected here
and recorded in CLAUDE.md rule #10 so future-Claude does not waste time
re-discovering it.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Wraps res.users._check_credentials. On AccessDenied, records a row with
result=failure and failure_reason='bad_password' (or '2fa_failed' when
credential['type'] == 'totp'), then re-raises. Regression test asserts
the attempted password value never lands in any audit field.
The audit row is written through registry.cursor() (independent cursor) so
it survives the rollback that follows AccessDenied — in production
odoo/service/model.py::retrying resets the transaction and http.py closes
the cursor without committing, in tests assertRaises opens its own
savepoint. Either way an inline write would vanish. Tests
enter registry_test_mode and use manual try/except to keep the audit row
visible across the savepoint hierarchy.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Overrides res.users._update_last_login to create a fusion.login.audit
row with result=success after the parent runs. The write goes through
sudo() + mail_create_nolog=True. Any exception in the audit path is
caught and logged but never propagates — a broken audit table must
never block a real user from logging in.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Single helper builds vals for fusion.login.audit rows from the live
HTTP request, or falls back to ip=''internal'' + geo_lookup_state=''internal''
when there is no request. Parses UA into browser/os/device_type via the
bundled user_agents library. Never reads credential[''password'']. Tests
cover: no-request fallback, UA parsing on a Chrome/Windows UA, and the
regression that no password value leaks into the vals dict.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Record rule grants admins an unrestricted domain on the audit log;
ACL forbids write/create/unlink for every group (audit is append-only;
sudo() inside auth hooks is the only write path). Defence-in-depth
layering: ACL is the actual gate, the rule documents and locks down
admin access path.
Tests (5, all green) cover:
test_admin_can_read_through_acl_and_rule — positive path through both.
test_acl_blocks_read_for_regular_user — base.group_user denied by ACL.
test_acl_blocks_read_for_portal_user — base.group_portal share user
denied (sensitive data leakage
surface closed at ACL layer).
test_acl_blocks_write_for_admin — append-only at the write boundary.
test_acl_blocks_unlink_for_admin — append-only at the unlink boundary.
Drop the redundant `from . import tests` from the root __init__.py —
Odoo's test loader imports `odoo.addons.<mod>.tests` directly; the
extra import was dead weight (and inconsistent with the repo pattern).
CLAUDE.md gotchas added during this task:
#6 res.users.groups_id -> group_ids rename (test setUp pitfall).
#6 ir.rule `groups` is additive, not restrictive — group-scoped
rules only apply to users in that group, they do not restrict
non-members. Default to letting the ACL gate; use rules for
row-level filters ACLs cannot express.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- All 16 columns per spec (user, attempted_login, result, failure_reason,
event_time, ip/geo fields, user_agent triple, device_type, database).
- Check constraint binds failure_reason presence to result value.
- Three composite indexes (user+time, login+time, geo_state+time) supporting
the per-user, failure-burst, and geo cron queries.
- Minimal admin-read ACL added so subsequent tests can verify writes.
- 3 TransactionCase tests passing: model create, failure_reason nullable on
success, geo_lookup_state='internal' accepted.
Odoo 19 deprecation note: this implementation uses the declarative
models.Constraint and models.Index attributes (Odoo 19 silently drops the
legacy `_sql_constraints = [...]` list and `init()`/raw-SQL pattern with
only a warning). Captured in CLAUDE.md rule #9.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Empty installable module with manifest, package inits, and icon.
Subsequent tasks add the audit model, hooks, views, and tests.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
