CREATE MATERIALIZED VIEW fusion_unreconciled_bank_line_mv pre-computes
the data the kanban widget needs (top suggestion, confidence band,
attachment count, partner reconcile hint) so that listing 50-100 lines
is one indexed query instead of N+1.
Refresh strategy:
- Triggered on fusion.reconcile.suggestion create/write (best-effort,
never poisons the originating transaction)
- Cron (every 5 min) — added in Task 25
The MV is created in the model's init() (Odoo calls this on
install/upgrade). The SQL DDL is idempotent
(CREATE MATERIALIZED VIEW IF NOT EXISTS / CREATE INDEX IF NOT EXISTS)
and includes a UNIQUE(id) index so REFRESH MATERIALIZED VIEW
CONCURRENTLY is supported. _refresh() falls back to a blocking refresh
on the first call after creation.
Made-with: Cursor
Per-step audit caught real enforcement bugs across all 9 WO kinds.
Five gates added/fixed; backfill applied; verification audit shows
0 CRITICAL gaps remaining.
**1. Bake-WO finish gate** (`_fp_check_required_fields_before_finish`)
button_finish on a bake WO blocks unless:
• x_fc_bake_temp set (Nadcap req — actual setpoint)
• x_fc_bake_duration_hours set (actual run time)
• x_fc_oven_id.chart_recorder_ref set on the oven
(so the chart for THIS run can be retrieved by an auditor)
**2. Rack-WO start gate** added to button_start.
**3. Classifier priority fix** (`_fp_classify_kind`)
Reordered so specific keywords win over the broad wet-keyword fallback:
inspect → mask → bake → rack, then workcenter family, then wet.
"Post-plate Inspection" now → inspect (was wrongly → wet).
"Oven bake (Post de-rack)" now → bake (was wrongly → rack).
**4. Auto-populate** target_thickness + dwell_time at WO generation.
Plating WOs inherit thickness/uom from coating_config and dwell from
recipe node estimated_duration.
**5. Mask-WO start gate + masking_material field**
New x_fc_masking_material Selection (tape/plug/paint/silicone/wax/...).
Required to start mask/de-mask WO. Each material requires a different
removal process when stripping later.
**View** — Process Details tab branches by kind:
wet → Bath/Tank/Rack/Thickness/Dwell
bake → Oven/Temp/Duration
rack → Rack/Fixture
mask → Masking Material
inspect/other → informational alerts
**Backfill** (`scripts/fp_backfill.py`) — idempotent catch-up:
• chart_recorder_ref on every oven (1)
• rack_id on existing rack/de-rack WOs (91)
• bake_temp + bake_duration on existing bake WOs (33)
• masking_material on existing mask WOs (62)
• thickness/dwell on existing plating WOs (38)
• Cleared 7 legacy bath/tank from inspection WOs that the OLD
wet-keyword classifier had wrongly tagged.
**Per-step audit** (`scripts/fp_per_step_audit.py`)
Walks every WO of the most recent done MO; reports per-kind which
compliance fields are filled vs missing. Re-runnable for regressions.
**Final verification** on freshly-run MO:
• 0 CRITICAL gaps across all 9 WO steps
• 2 IMPORTANT (dwell_time + rack_id on E-Nickel Plating — both
inherited from recipe node data, not enforcement bugs)
• Classifier correct for all 9 step types
12 negative tests still passing.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Per-step audit caught real enforcement bugs across all 9 WO kinds in
the recipe (Masking, Racking, Plating, De-Masking, Oven baking, etc.).
Five gates added or fixed; 0 CRITICAL gaps remain after a verification
run on a fresh MO.
**1. Bake-WO finish gate** (`_fp_check_required_fields_before_finish`)
button_finish on a bake WO now blocks unless:
• x_fc_bake_temp set (Nadcap req — actual setpoint, not just oven)
• x_fc_bake_duration_hours set (actual run time at temp)
• x_fc_oven_id.chart_recorder_ref set (so the chart for THIS run
can be retrieved by an auditor — required for AS9100/Nadcap)
Run-time data lives at FINISH, not START — operators don't know
temp/duration until the bake is done.
**2. Rack-WO start gate** added to the existing button_start gate.
Per-rack life tracking + which physical fixture handled the parts.
**3. Classifier priority fix** (`_fp_classify_kind`)
"Post-plate Inspection" was matching the `plat` wet keyword and
getting kind=wet (then required to have bath/tank). Reordered:
1. Explicit equipment links (bath_id/oven_id)
2. Specific keywords (inspect → mask → bake → rack)
— bake before rack so "Oven bake (Post de-rack)" → bake
3. Workcenter wet families
4. Wet name keywords as last fallback
**4. Auto-populate target_thickness + dwell_time** at recipe→WO
generation. Plating WOs inherit:
• thickness_target from coating_config.thickness_max
• thickness_uom from coating_config.thickness_uom
• dwell_time_minutes from recipe node's estimated_duration
So aerospace QC has the spec target on every WO without paper.
**5. Mask-WO start gate + masking_material field**
New x_fc_masking_material Selection (tape/plug/paint/silicone/wax/
mixed/other). Required to start a mask WO. Needed later when
stripping or replating because each material requires a different
removal process.
**View** (`mrp_workorder_views.xml`)
Process Details tab now branches by kind:
wet → Bath/Tank/Rack/Thickness/Dwell
bake → Oven/Temp/Duration
rack → Rack/Fixture
mask → Masking Material
inspect/other → informational alerts only
WO Kind shows as colour-coded badge in header.
**Backfill** (`scripts/fp_backfill.py`)
Idempotent script that catches up existing data:
• chart_recorder_ref on every oven
• rack_id on existing rack/de-rack WOs (91 backfilled)
• bake_temp + bake_duration_hours on existing bake WOs (33)
• masking_material on existing mask WOs (62)
• thickness/dwell on existing plating WOs (38)
• Cleared 7 legacy bath/tank from inspection WOs that had been
misclassified by the OLD wet-keyword classifier.
**Per-step audit** (`scripts/fp_per_step_audit.py`)
Walks every WO of the most recent done MO and reports per-kind
which compliance fields are filled vs missing. Re-runnable to
catch regressions.
**Final state on freshly-run MO 00049:**
• 0 CRITICAL gaps
• 2 IMPORTANT gaps (dwell_time + rack_id on E-Nickel Plating —
both inherited from recipe node data, not enforcement bugs)
Negative tests still passing (12 total).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
When fusion_accounting_bank_rec is installed, match_bank_line_to_payments
and auto_reconcile_bank_lines now use fusion.reconcile.engine via the
BankRecAdapter, gaining precedent recording, AI suggestion superseding,
and shared validation. Legacy paths preserved for Enterprise/Community-
only installs (engine model absent -> fall back to set_line_bank_statement_line
and _try_auto_reconcile_statement_lines).
Also wraps engine.reconcile_batch's per-line loop in a savepoint so a
single bad line's DB error (e.g. check-constraint violation) no longer
poisons the whole batch transaction; the existing per-line try/except
now isolates failures as originally intended.
Made-with: Cursor
Adds fusion_suggest_matches, fusion_accept_suggestion,
fusion_reconcile_bank_line, fusion_unreconcile, and
fusion_get_pending_suggestions. All route through the BankRecAdapter
(or direct engine for ones the adapter doesn't expose), giving the AI
chat the same reconciliation surface a human operator gets in the OWL UI.
Made-with: Cursor
Enhances list_unreconciled_via_fusion to include fusion fields
(top_suggestion_id, confidence_band, attachment_count). Adds 3 new
adapter methods that proxy the engine: suggest_matches, accept_suggestion,
unreconcile. AI tools (Task 22+) and OWL controller (Task 26) will call
these adapter methods instead of touching the engine directly.
Made-with: Cursor
Provider-agnostic system + user prompt builder for the confidence
scoring pipeline's Pass 3 (AI re-rank). Output contract is JSON with
"ranked" array; works with OpenAI, Claude, and local OpenAI-compatible
servers (LM Studio, Ollama).
Made-with: Cursor
Two engine bugs caught by Task 19's integration tests:
1. Partial reconcile (bank_amount < invoice_residual) was creating an
unbalanced bank move. Counterpart balance now clamped to
min(remaining_bank_amount, abs(invoice_residual)) so the move stays
balanced; Odoo's reconcile() handles the resulting partial. The
counterpart's amount_currency is scaled proportionally so multi-
currency lines stay consistent.
2. Unreconcile only removed account.partial.reconcile rows but didn't
restore the suspense line on the bank move, leaving is_reconciled=True
after unreconcile. Now delegates to V19's standard
account.bank.statement.line.action_undo_reconciliation for any
affected bank line, which both deletes partials and restores the
suspense state in one shot.
Made-with: Cursor
Tests engine behavior using factories (Task 18) instead of SQL fixtures.
Covers simple match, partial chain, multi-invoice batch, suggest-then-
accept flow, unreconcile reversal, and edge cases.
Two tests are intentionally failing — they expose real engine bugs
that should be fixed in a follow-up:
- TestReconcilePartialChain.test_partial_reconcile_leaves_residual:
reconcile_one() builds counterpart vals using the full invoice
residual, which leaves the bank move unbalanced when bank amount
is smaller than the invoice (UserError: entry not balanced).
- TestUnreconcile.test_unreconcile_removes_partial: unreconcile()
unlinks partial.reconcile rows but does not restore the suspense
line on the bank move, so account.bank.statement.line.is_reconciled
remains True after reversal.
Made-with: Cursor
Provides make_bank_journal, make_bank_statement, make_bank_line,
make_invoice, make_vendor_bill, make_suggestion, make_pattern,
make_precedent, make_reconcileable_pair helpers used across the
bank-rec test suite. Replaces the original plan's SQL-fixture capture
with programmatic factories — same testing intent, simpler maintenance,
no real Westin data baked into the repo.
Note: the original plan called for 5 SQL fixtures captured from the local
DB (westin_simple_match.sql, westin_partial_chain.sql, etc.). Those are
replaced by factory-driven test creation in Task 19 — eliminates fragile
hand-curated SQL while testing the same code paths.
Made-with: Cursor
Replaces the upstream Odoo icons with the purple-pink-orange V mark
so all three modules show consistent Fusion branding in the Apps list
and settings UI.
Same icon file across all three so they read as a family. Upstream
had its own icon.png on the `iot` module which this overwrites.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds fusion.reconcile.engine — the AbstractModel orchestrator for all
bank-line reconciliations. Six public methods (reconcile_one,
reconcile_batch, suggest_matches, accept_suggestion, write_off,
unreconcile) form the only sanctioned write path to
account.partial.reconcile from the rest of the module (controllers, AI
tools, wizards).
Implementation follows V19's bank_rec_widget pattern: rewrite the bank
move's suspense line into one counterpart per matched invoice (or a
write-off line) on the appropriate receivable / payable / write-off
account, then call account.move.line.reconcile() on each pair. Records
a precedent row per reconcile for downstream pattern learning.
16 new unit tests cover all six methods across happy paths, the
precedent side effect, suggestion lifecycle, batch auto-strategy, and
write-off line clearance. 67 total tests, 0 failed.
Made-with: Cursor
User caught two related issues from screenshots of the WO form:
1. The "Plating Details" tab was meaningless for non-wet WOs —
bath/tank/dwell/thickness all show as empty for masking, oven
bake, racking, and inspection steps. A shop with multiple ovens
had no way to record which oven a bake WO ran in.
2. When there's only ONE option (single oven, single bath), forcing
the manager to pick it on every WO is busywork — pin it
automatically.
**1. WO classification + per-kind equipment**
New `x_fc_wo_kind` (compute, non-stored) Selection field that buckets
each WO into one of: wet / bake / mask / rack / inspect / other.
Classification by priority:
• bath linked → wet
• oven linked → bake
• workcenter's process families wet → wet
• WO name keyword match (bake/oven/cure → bake;
mask/de-mask → mask; rack/de-rack → rack;
inspect/qa/qc/fai → inspect; default → other)
New equipment fields per kind:
• `x_fc_oven_id` (m2o fp.bake.oven) for bake WOs
• `x_fc_bake_temp`, `x_fc_bake_duration_hours` — bake parameters
• Existing bath/tank/rack/thickness reused for wet
• Existing rack reused for rack WOs
**2. Required-fields gate extended**
button_start now also requires `x_fc_oven_id` for bake WOs (alongside
the existing operator + bath/tank rules). Without an oven the
chart-recorder trail can't be tied back to the WO for compliance.
**3. View reorganized**
Process Details tab now shows only the equipment groups that apply
to this WO's kind (using `invisible="x_fc_wo_kind != 'bake'"` etc.).
Mask + Inspection + Other show informational alerts instead of
empty form fields. WO header shows a colour-coded kind badge.
**4. Smart auto-fill defaults**
New `_fp_autofill_default_equipment()` method on mrp.workorder. When
the facility has exactly ONE active option, it pre-pins:
• Bath → if facility has 1 active bath
• Tank → if the chosen bath has 1 active tank
• Oven → if facility has 1 active oven
Hooked from:
• `@api.onchange('workcenter_id', 'x_fc_facility_id', 'x_fc_bath_id')`
→ fills as user edits in the form
• Recipe → WO generation `_generate_workorders_from_recipe()`
→ fills at creation time so single-line shops never see an
empty bath/oven field
None of this overwrites an already-set value. Multi-line shops still
get a blank field to choose from.
**Simulator updates** (scripts/fp_e2e_workforce.py)
• Creates an oven if none exists
• Pins per-kind equipment in Hannah's planning step
• New PASS check: bake-WO auto-pinned to default oven
• New negative test 2b: bake WO with oven stripped → blocked
**Final E2E**: 54 PASS / 2 WARN / 0 FAIL out of 56 checks.
12 negative tests passing — all gates fire when triggered:
Tests 1-2 + 2b: WO start (operator + bath/tank + oven)
Tests 3-7: MO facility, cert spec, delivery POD, invoice
payment terms, thickness cal std
Tests 8-11: NCR close, CAPA close, discharge close, invoice ref
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Phase A of the IoT initiative — gets the server-side infrastructure
in place before the Raspberry Pi hardware arrives, so the iot admin
UI + /fp/iot/ingest endpoint are ready to accept the first real
temperature reading as soon as the Pi is wired up.
New top-level folder: fusion_iot/
1. **iot_base/** — Odoo S.A. iot_base module, copied from
RePackaged-Odoo verbatim. LGPL-3 upstream, no changes needed.
2. **iot/** — Odoo S.A. iot module, repackaged:
- `models/update.py` neutralised (removed the publisher_warranty
IoT-Box-counting report that phones home to odoo.com for
enterprise licence enforcement)
- `iot_handlers/lib/load_worldline_library.sh` deleted (proprietary
Worldline payment lib fetch from download.odoo.com, not needed)
- `wizard/add_iot_box.py._connect_iot_box_with_pairing_code` —
upstream called odoo.com's iot-proxy to resolve pairing codes;
replaced with a no-op. Pi-side iot_drivers proxy registers
directly with this Odoo server instead.
- Manifest rebranded with an explicit changelog preamble.
3. **fusion_plating_iot/** — new plating-specific wrapper:
- `fp.tank.sensor` — maps an iot.device (or a direct-HTTP-ingest
sensor) to a fusion.plating.tank + fusion.plating.bath.parameter.
Supports DS18B20, PT100/1000, pH, conductivity, level. Per-sensor
alert_min/max overrides.
- `fp.tank.reading` — append-only time-series. On create, evaluates
against sensor's alert range. On in-spec → out-of-spec TRANSITION,
auto-raises a fusion.plating.quality.hold (once per excursion,
no spam during sustained out-of-spec).
- `POST /fp/iot/ingest` — shared-secret HTTP endpoint for sensors
bypassing the Pi proxy. Token via X-FP-IOT-Token header OR body.
Accepts single-reading or batch payloads.
- Menu under Plating → Operations → Sensors & Readings.
- Tank form inherits get a Sensors tab inline.
Deployed to entech. Verified end-to-end:
- Install: iot_base + iot + fusion_plating_iot all 'installed'
- Smoke test: in-spec → out-of-spec → hold raised (HOLD-0010);
continued excursion → NO duplicate hold; back-in-spec → NEW
excursion → NEW hold (HOLD-0011) ✓
- HTTP endpoint: correct token → 200 accepted; wrong token → 401;
unknown device_serial → 404; batch payload → 200 accepted=N ✓
Phase B (when Raspberry Pi hardware arrives): DS18B20 iot_handler
driver for the Pi-side iot_drivers proxy + systemd service on
vanilla Raspberry Pi OS + first live reading from physical probe.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Task 11 of Phase 1 Bank Reconciliation. Adds the brain that ranks
candidate journal-item matches for a bank statement line.
Pass 1 — SQL filter (done by caller's _fetch_candidates).
Pass 2 — Statistical scoring: weighted blend of amount-delta,
partner pattern fit, and precedent similarity.
Pass 3 — Optional AI re-rank when an LLM provider is configured;
gracefully no-ops when provider missing, prompt module not
yet present (Task 20), or the JSON response is malformed.
Pass 4 — Persistence (handled by engine.suggest_matches).
Returns top-K ScoredCandidate dataclasses with per-feature scores
exposed for transparency and future learning.
7 new tests added; full module suite green (51 tests, 0 failures).
Made-with: Cursor
Adds the foundation for AI confidence scoring:
- fusion.reconcile.pattern: per-(company, partner) aggregate profile
(volume, cadence, preferred matching strategy, memo signature,
write-off habits) — recomputed nightly from precedents.
- fusion.reconcile.precedent: per-historical-decision memory holding
full feature vector + outcome, used by precedent_lookup for KNN
scoring of new bank lines.
Includes ACL rows for fusion accounting user (read) and admin (CRUD)
groups. Manifest bumped to 19.0.1.0.1.
Note: switched the pattern uniqueness rule from the deprecated
_sql_constraints attribute to models.Constraint (Odoo 19 native API)
so the unique(company_id, partner_id) is actually enforced at the
PG level — _sql_constraints is silently ignored in 19.
Made-with: Cursor
Phase 1 prerequisite for local LLM support. Adapters now declare
capability flags (supports_tool_calling, max_context_tokens, etc.) so
the engine can reason about what backend is available.
OpenAI adapter accepts fusion_accounting.openai_base_url config -- point
it at LM Studio (http://host.docker.internal:1234/v1) or Ollama
(http://host.docker.internal:11434/v1) and the existing OpenAI adapter
works unchanged.
Implementation note: existing Odoo AbstractModel adapters
(fusion.accounting.adapter.openai/claude) are preserved untouched to
avoid breaking the chat panel; the new plain-Python OpenAIAdapter and
ClaudeAdapter classes (LLMProvider subclasses) are added alongside them.
Made-with: Cursor
group_fusion_show_when_enterprise_absent has membership = all internal
users iff no Enterprise accounting module is installed. Membership is
recomputed on module install/uninstall via overrides on ir.module.module.
Used by Phase 1 fusion_bank_rec menus to auto-hide when Enterprise is
active and auto-appear after Enterprise uninstall.
Made-with: Cursor
Declare account.bank.statement.line.cron_last_check on
fusion_accounting_core so the column survives Enterprise
account_accountant uninstall. Mirrors the existing pattern used
for account.move and account.reconcile.model shared fields.
- Add models/account_bank_statement_line.py declaring cron_last_check
as fields.Datetime(copy=False)
- Wire model into models/__init__.py
- Add post_install regression test verifying field presence and type
- Bump manifest 19.0.1.0.0 -> 19.0.1.0.1
Made-with: Cursor
Scaffold the fusion_accounting_bank_rec sub-module with directory
tree, manifest, empty package __init__ files, empty ACL CSV, icon,
and Enterprise reference snapshots. No models, controllers, or
business logic yet — installs cleanly on V19 westin-v19 dev DB.
Made-with: Cursor
**7a. NCR close gate** (fusion.plating.ncr.action_close)
Block close unless these are filled in:
• Description (what happened)
• Containment Actions (immediate response)
• Root Cause (why it happened)
• Disposition (use-as-is / rework / scrap / RTV decision)
A closed NCR without these is useless for AS9100 audits — it's
the entire point of an NCR to document what went wrong, why, and
how we responded. Empty-HTML strings like "<p><br></p>" are
detected as empty too.
**7b. CAPA close gate** (fusion.plating.capa.action_close)
Block close unless:
• Root Cause Analysis filled in
• Action Plan filled in
• Verification (date + verifier) recorded
• Effectiveness Notes filled when CAPA was marked Not Effective
AS9100 §10.2 / Nadcap require evidence of root-cause analysis,
the corrective/preventive action plan, AND that effectiveness
was verified before the loop is closed.
**8. Invoice ref defensive default** (account.move.create)
Auto-fills `ref` from the source SO's client_order_ref or
x_fc_po_number when the invoice is created with invoice_origin set
but no ref. Already populated on the SO confirm path; this catches
manually-created invoices that would otherwise miss it. Customer
AP teams reject invoices that don't quote their PO# back.
**9. Discharge sample close gate** (fusion.plating.discharge.sample.action_close)
Block close unless:
• Lab Report # set
• Results Received Date set
• At least one parameter reading on file
• Lab certificate/report attached
Without lab evidence the record fails any environmental compliance
audit — the whole point is to document the test was performed and
what the lab said.
**Simulator** (scripts/fp_e2e_workforce.py)
Adds 4 new negative tests (Test 8-11), all wrapped in savepoints:
✓ Test 8 : NCR close without RC/containment/disposition → blocked
✓ Test 9 : CAPA close without analysis/plan/verification → blocked
✓ Test 10: Discharge sample close without lab evidence → blocked
✓ Test 11: Invoice ref auto-fills from SO.client_order_ref → asserted
**Final E2E**: 52 PASS / 2 WARN / 0 FAIL out of 54 checks.
Both remaining WARNs are expected (bake-window auto-create,
first-piece gate — coating-driven, this coating doesn't trigger them).
11 negative tests in total now, every gate fires when triggered.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Following the workforce-E2E + required-fields audit, ship the first 6
high-priority gates so critical workflow + compliance fields can no
longer be left empty by accident.
**1. Invoice payment terms (account.move)**
- create() now auto-inherits `invoice_payment_term_id` from
partner.property_payment_term_id when missing
- action_post() raises UserError if still missing — accountant must
pick one before posting (prevents silent "immediate" due-date)
**2. MO facility (mrp.production)**
- action_confirm() auto-derives `x_fc_facility_id` if unset, in order:
SO override → res.company.x_fc_default_facility_id → first active
facility — then HARD GATES: raises UserError if still empty.
Without facility every downstream record (WO, batch, bath log,
cert) is missing the "where" half of the audit trail.
**3. WO facility (mrp.workorder)**
- Switched `x_fc_facility_id` from related (workcenter only) to a
proper compute that falls back to production_id.x_fc_facility_id.
Stub workcenters auto-created from process node names usually have
no facility — the MO always does (from #2 above).
**4. Thickness reading calibration_std (fp.thickness.reading)**
- `calibration_std_ref` is now `required=True` with sensible default
("NiP/Al STD SET SN 100174568"). Nadcap mandates which calibration
standard the gauge was checked against — without it the cert
data has no chain back to a metrology record.
**5. Delivery POD gate (fusion.plating.delivery)**
- action_mark_delivered() raises UserError if no `pod_id`. Driver
must capture POD on the iPad (recipient signature + photos +
notes) BEFORE marking delivered. Without POD there's no signed
receipt to back the invoice or defend a delivery dispute.
**6. Certificate spec_reference gate (fp.certificate)**
- action_issue() raises UserError if no `spec_reference`. The cert
ATTESTS to a spec — leaving it blank produces a piece of paper
that AS9100 / Nadcap auditors will (rightfully) reject.
**Simulator updated**: scripts/fp_e2e_workforce.py
- Sets net-30 on the test customer + ensures a default facility
- New PHASE 4c: 5 negative tests (one per new gate), each wrapped
in a SAVEPOINT so SQL constraint violations don't abort the txn
- Driver now creates POD on iPad BEFORE marking delivered
**Final E2E**: 48 PASS / 2 WARN / 0 FAIL out of 50 checks.
The 2 remaining WARNs (bake-window auto-create, first-piece gate)
are expected behaviour — both are coating-driven and the test
coating intentionally doesn't trigger them.
All 7 negative tests now pass:
✓ Test 1: WO start without operator → blocked
✓ Test 2: WO start on wet WO without bath/tank → blocked
✓ Test 3: MO confirm without facility → blocked
✓ Test 4: Cert issue without spec_reference → blocked
✓ Test 5: Delivery delivered without POD → blocked
✓ Test 6: Invoice post without payment terms → blocked
✓ Test 7: Thickness reading without cal std → blocked (DB NOT NULL)
Audit script (scripts/fp_required_fields_audit.py) committed too —
it's the diagnostic that surfaced these gaps and can be re-run to
catch new ones.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
User audit caught: in the workforce E2E run we had no idea which bath /
which tank ran the job. For aerospace traceability that's a deal-
breaker. Add a validation gate on mrp.workorder.button_start so
operators can't tap START without the data the shop floor MUST capture.
**Three new pieces on mrp.workorder:**
1. `_fp_is_wet_process()` — best-effort "does this WO involve a
chemistry bath?" check. Three signals in priority order:
a. A bath is already linked → definitely wet
b. The workcenter's FP work-centre supports a wet process family
(plating, pre/post-treatment, strip, passivation)
c. WO name contains a wet-process keyword (plat, nickel, chrome,
anodiz, zinc, etch, clean, rinse, strip, passivat, electroless…)
The keyword fallback is needed because most existing recipes have
no process_type_id set on their operation nodes.
2. `_fp_check_required_fields_before_start()` — runs before the
existing certification check. Rules:
• Every WO needs an assigned operator (x_fc_assigned_user_id).
Without it, productivity records can't be attributed and the
proficiency tracker has no employee to credit.
• Wet WOs additionally need x_fc_bath_id + x_fc_tank_id. So we
know exactly which chemistry bath ran the job and which physical
tank it sat in.
Raises a clear UserError listing the missing fields if any.
3. `x_fc_requires_bath` (compute, non-stored) — surfaces the wet check
to the form view so bath + tank fields render with `required=`.
**View changes:**
- `x_fc_assigned_user_id` is now `required="1"` on the form
- `x_fc_bath_id` + `x_fc_tank_id` use `required="x_fc_requires_bath"`
→ red asterisk only when the WO is actually wet
**Simulator updates** (scripts/fp_e2e_workforce.py):
- Hannah now explicitly assigns bath + tank to wet WOs during planning,
AND pre-issues operator certifications for the bath's process type
(real shop manager workflow).
- Two negative tests added that PROVE the gates fire:
• Test 1: strip the operator → button_start raises "missing Assigned Operator"
• Test 2: strip bath/tank on a wet WO → button_start raises "missing Bath/Tank"
**Final E2E:** 42 PASS / 2 WARN / 0 FAIL out of 44 checks.
Both remaining WARNs (bake-window auto-create, first-piece gate) are
expected behaviour — those are coating-driven and the test coating
intentionally doesn't trigger them.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
51 tasks across 17 groups covering the full Phase 1 build:
Group 1 (5 tasks): Foundation — branch, sub-module skeleton, shared
fields on _core, LLMProvider contract for local LLM readiness
Group 2 (8 tasks): Reconcile engine — TDD-layered build of
matching_strategies, exchange_diff, memo_tokenizer, precedent_lookup,
pattern_extractor, confidence_scoring 4-pass pipeline, the AbstractModel
engine with 6-method API, and Hypothesis property-based tests
Group 3 (4 tasks): Models — fusion.reconcile.pattern,
fusion.reconcile.precedent, fusion.reconcile.suggestion, widget transient,
and inherits on Community account.bank.statement.line + account.reconcile.model
Group 4-5 (6 tasks): Integration tests with SQL fixtures from real Westin
reconciles + AI prompts + adapter fill-ins + AI tools refactor
Group 6-7 (3 tasks): Materialized view, cron schedules, and 10-endpoint
JSON-RPC controller with auth guards
Group 8-10 (10 tasks): Frontend — SCSS tokens, service, kanban controllers,
all 18 Enterprise-mirror OWL components, and 5 fusion-only components
(ai_suggestion folder, batch_action_bar, attachment_strip,
partner_history_panel, reconcile_model_picker)
Group 11-13 (5 tasks): Wizards (auto-reconcile + bulk), migration wizard
inheritance with bootstrap of 16,500 historical reconciliations + audit
report PDF + round-trip test, coexistence menu/group + tests
Group 14-16 (3 tasks): 5 OWL tour tests, performance benchmarks against
P95 targets, local LLM compatibility test against LM Studio
Group 17 (4 tasks): Closeout — meta-module manifest update, sub-module
docs, end-to-end smoke test, completion tag
TDD discipline throughout: every code task is red test → impl → green
→ commit. Property-based tests for amount invariants. Migration round-
trip test asserts byte-identical reconciliation state pre/post Enterprise
uninstall. All testing on local OrbStack VM only (environment-safety
rule applies).
Made-with: Cursor
Built a comprehensive simulator (scripts/fp_e2e_workforce.py) that
role-plays 10 employees driving an order quote → invoice using real
operator timers (button_start / button_finish with elapsed time.sleep).
Initial run: 31 PASS / 2 WARN / 0 FAIL exposed two gaps that would
hurt a real shop:
**Gap 1 — Thickness readings never reached the CoC**
The Fischerscope readings inspectors take during post-plate inspection
had no path to the CoC. The cert came out empty, useless for AS9100
or aerospace audits.
Fixes:
- New tablet endpoint `/fp/shopfloor/log_thickness_reading` so the
inspector can record one reading at a time during the inspection WO
(auto-numbers, defaults the operator, supports microscope image).
- mrp_production._fp_mark_done_post_actions now bulk-links any
orphan thickness readings (those with production_id=mo.id but no
certificate_id) to the freshly-created CoC. So inspectors can log
during inspection AND the cert PDF picks them up automatically.
**Gap 2 — Operator queue leaked other people's work + simulator missed it**
fusion.plating.operator.queue.build_for_user pulled EVERY ready /
in-progress WO regardless of assignment. Tom would see John's masking
WO in his "Up Next" list — bad for aerospace traceability where you
want strict per-operator accountability.
Fix: build_for_user now filters MRP WOs by
`(x_fc_assigned_user_id == user_id OR x_fc_assigned_user_id == False)`.
Operators see their own assigned tasks first, plus any unassigned
tasks anyone can grab. Other operators' assigned WOs no longer leak
through.
Also caught: simulator was using wrong field name on the queue model.
Fixed and added a "queue isolation" check that verifies no operator
sees another operator's assigned WOs.
After fixes: **39 PASS / 2 WARN / 0 FAIL** (out of 41 checks).
Remaining WARNs are both expected behaviour:
- bake-window auto-create: this coating doesn't require_bake_relief
(the recipe has an inline Oven step instead)
- first-piece gate: same — coating-driven, only fires when needed
Areas validated end-to-end:
- quote → SO with PO# carried into client_order_ref
- SO confirm → MO + portal job auto-created
- receiving qty prefill + accept
- 9 WOs generated from recipe + assigned to specific operators
- All 9 WOs ran with real elapsed timers + 17 productivity records
across 4 distinct operators
- MO done triggers CoC auto-issue with 5 thickness readings linked,
319 KB rich PDF, customer-slug filename
- Delivery auto-created with prefilled date + driver + CoC link
- Delivery delivered, 2 chain-of-custody entries
- Invoice posted (NOT auto-paid)
- All 5 customer notifications fired (so_confirmed +
parts_received + mo_complete + shipped + invoice_posted) with
correct attachments
- Portal job → complete, SO workflow_stage → invoicing
- Chemistry log persisted, operator proficiency tracked
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Drafts the design for fusion_accounting_bank_rec — a native bank
reconciliation widget that replaces Odoo Enterprise account_accountant
in V19 OWL architecture, with a clean-room reconcile engine reading and
writing Community account.partial.reconcile rows.
Key design decisions captured:
- CORE scope (~5.5-6 weeks): manual + auto reconcile, write-offs,
partial, multi-currency, chatter, model picker
- Strict mirror of all 18 Enterprise OWL units (zero functional loss)
plus 5 fusion-only additions for AI/history visibility
- Hybrid AI badge layout: inline strip with one-click Accept plus
expandable ranked-alternatives panel
- Behavioural learning via fusion.reconcile.pattern (per-partner) and
fusion.reconcile.precedent (per-decision memory) with bootstrap from
the 16,500 historical reconciliations
- Local LLM ready via OpenAI-compatible adapter base_url config and
per-feature provider routing — works against LM Studio, Ollama, vLLM
- Statistical-mode-without-API-key as a first-class path
- Coexistence with Enterprise: Enterprise wins by default, fusion
menu hides until uninstall, then auto-appears
- Migration wizard step bootstraps pattern memory and produces an
audit report PDF proving every reconciliation preserved
- TDD on engine algorithms with Hypothesis property-based tests for
amount invariants; migration round-trip integration test
Builds on Phase 0 (commit c450bb2, range pre-phase-0..phase-0-complete).
Made-with: Cursor
Users were seeing both Odoo's stock PDFs and FP's branded equivalents
in the Print dropdown side-by-side, and accidentally sending the wrong
(unbranded, missing PO# / job ref / plating fields) PDF to customers.
Add fp_hide_default_reports.xml that drops the Print-menu binding on:
| Model | Hidden | FP replacement |
|-----------------|-------------------------------------------------------------|---------------------------------|
| sale.order | sale.action_report_saleorder | action_report_fp_sale_* |
| sale.order | sale_pdf_quote_builder.action_report_saleorder_raw | action_report_fp_sale_* |
| account.move | account.account_invoices | action_report_fp_invoice_* |
| account.move | account.account_invoices_without_payment | action_report_fp_invoice_* |
| stock.picking | stock.action_report_delivery | action_report_fp_packing_slip_* |
| mrp.production | mrp.action_report_production_order | action_report_fp_job_traveller_*|
| account.payment | account.action_report_payment_receipt | action_report_fp_receipt_* |
Mechanism: set binding_model_id=False + binding_type=action — removes
from the Print dropdown but leaves the report record + template intact.
Fully reversible from Settings → Technical → Reports if anyone needs
the stock PDF back.
Intentionally NOT touched:
- sale.action_report_pro_forma_invoice (no FP pro-forma yet)
- account.action_account_original_vendor_bill (vendor bills, internal)
- stock.action_report_picking / picking_packages / return_label_report
(internal warehouse ops, not customer-facing)
- mrp.action_report_finished_product / mrp.label_manufacture_template
(production labels — ZPL, not customer-facing)
- sale_timesheet.* (timesheet integration)
Added sale_pdf_quote_builder to depends so the data file always finds
that record when applied (it ships in entech's repackaged enterprise
bundle and was already installed there).
Verified on entech: re-running the print-menu audit shows zero stock
Odoo customer-facing PDFs left where FP has an equivalent.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The floating message-action toolbar (reaction / reply / star / link
icons) appearing on hover renders white-icons-on-white-background in
dark mode — Odoo's own dark.scss sets the icon hover color to white
but never gives the toolbar itself a dark background. Result: the
icons vanish entirely in dark mode.
Add fp_chatter_dark.scss that branches at compile time on
$o-webclient-color-scheme == dark (Odoo 19 compiles every SCSS file
into both web.assets_backend with `bright` AND web.assets_web_dark
with `dark`) and gives the toolbar:
- Solid dark background (#2b2f33 fallback, var(--o-component-bgcolor))
- Subtle 1px white-alpha border + drop shadow so it floats nicely
- Icon color rgba(255,255,255,.78) at full opacity (not 35%)
- Brighter hover state with a subtle bg highlight
Light bundle output is empty (the @if branch doesn't fire), so the
light theme is untouched.
Verified: dark bundle includes our rule with #2b2f33 marker present.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two fixes from a single SO walkthrough screenshot:
**1. "Current stage" banner**
- Was placed `inside sheet` so it rendered at the BOTTOM of the form
where users miss it. Moved to `before form/header` (same xpath
pattern as the Account Hold banner) — now it's the first thing
visible above the SO header.
- Was still showing "Shipped — awaiting invoice" after the invoice
was posted because `_compute_workflow_stage` only advanced to
`complete` when shipped + ALL paid; an unpaid posted invoice left
the SO stuck on `shipped`. Added an `invoicing` branch: shipped +
has_posted_invoice → invoicing. Banner invisible-list now also
includes `invoicing` and `paid`, so the banner only shows for
in-progress steps.
**2. Chatter messages rendering raw HTML tags as text**
Odoo 19 escapes any string passed to `message_post(body=...)`
unless wrapped in `markupsafe.Markup`. We had ~10 places posting
HTML (`<a href>`, `<b>`, `<br/>`, `<code>`, `<pre>`) that all
showed up as `<a href=...>` literal text in the chatter.
Wrapped each one with `Markup(_(...))` so the tags render. Files
touched:
- fusion_plating_bridge_mrp/models/sale_order.py
(auto-MO failure code block, "Draft MO created" link,
"Job assigned to <b>" message)
- fusion_plating_bridge_mrp/models/mrp_production.py
("Recipe steps" pre/br block on each WO)
- fusion_plating_bridge_mrp/models/fp_proficiency.py
(operator promotion announcement)
- fusion_plating_configurator/models/fp_quote_configurator.py
(SO link, 3D model attached, drawing attached, save to catalog)
- fusion_plating_configurator/models/fp_part_catalog.py
(3D/drawing change tracking + propagation to linked quotes)
- fusion_plating_portal/models/fp_quote_request.py
(RFQ → SO link)
- fusion_plating_quality/models/fp_quality_hold.py
(hold status change)
- fusion_plating_shopfloor/controllers/manager_controller.py
(worker / tank / manager-takeover assignments)
Verified on entech: SO S00038 stage now reads `invoicing` (banner
hidden), and a freshly posted message shows `<a href>` and `<b>`
as actual link + bold instead of escaped text.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The 'Fusion Accounting' top-level menu was missing the web_icon attribute,
so the app switcher grid showed a placeholder instead of the branded icon.
ir.ui.menu.web_icon is separate from ir.module.module.icon (Apps page) —
both need to be set for full icon coverage.
Made-with: Cursor
User reported "multiple unwanted vertical lines in the boxes" on the
portrait BoL. Pixel analysis confirmed it: previous design had 3
separate `<div class="sig-box">` each with its own 1px border, with a
4-8px gap between adjacent boxes — visually those adjacent borders
read as a doubled / "duplicate" line between cells.
Fix: replace 3-box layout with a single `<table class="bordered
sig-table">` containing 3 td cells. With border-collapse: collapse,
adjacent cells share their border — so the row now shows 4 vertical
lines (1 outer left + 2 internal dividers + 1 outer right) instead
of 6 close-together border lines.
- Dropped `.sig-box` class entirely (no per-box border anymore)
- Added `.sig-table` + `.sig-cell` with explicit 1px borders so the
layout works without depending on `.bordered` class inheritance
- Applied to both portrait + landscape variants
- Landscape sig-row was still using the OLD Bootstrap row+col-4
layout (never got replaced earlier) — also migrated to the new
table layout
Verified: page count unchanged (portrait 1, landscape 1), all
labels and content present, structure clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Drops the 73KB icon.png into each of the four sub-modules
(fusion_accounting meta, _core, _ai, _migration) so Odoo's Apps page
renders the branded icon for each. Meta-module manifest 'icon' path
now points to its own icon instead of the AI sub-module's.
Made-with: Cursor
Last fix kept signatures intact but the landscape BoL still overflowed
to a second page (with the signature row pushed entirely to page 2).
The real ask was for the landscape variant to fit on one page since
landscape has plenty of vertical room.
Aggressive landscape compaction:
- Body font 11pt → 10pt, td font 10 → 9.5pt, th font 10 → 9pt
- Cell padding 8/10px → 4/8px
- Table margin-bottom 12px → 6px
- h2 title 26pt → 18pt with tighter top/bottom margins
- BoL # subtitle 14pt → 11pt
- Shipper/consignee row height 120 → 70px
- highlight-box (cert) padding 10px → 6/10, font 10 → 9pt
- sig-box padding 12 → 8/10px
- sig-line height 70 → 45px
Verified with pypdf: landscape BoL now renders as exactly 1 page
with cert + all 3 signature labels + company info all present.
137 KB clean PDF.
Portrait variant left untouched (it already fit on one page and
the bigger title is appropriate for portrait).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Never again touch production without explicit confirmation. This rule
codifies the hard-won lesson from 2026-04-19: ssh odoo-westin goes to
PRODUCTION (192.168.1.40, erp.westinhealthcare.ca), not dev, despite
the container being named odoo-dev-app.
alwaysApply: true.
Made-with: Cursor
Last fix added page-break-inside: avoid but the boxes still split
because wkhtmltopdf 0.12 ignores that rule inside flex containers,
and BOTH the .sig-box (display: flex) AND the Bootstrap .row
wrapper were flex.
Replace both with non-flex equivalents:
- .sig-box: dropped `display: flex` + `flex-direction: column` +
`justify-content: flex-end`. Layout now uses padding + a fixed-
height .sig-line block + the muted label below. Same visual
result, but a plain block element so wkhtmltopdf honors the
page-break rule.
- Replaced `<div class="row">` + 3 `<div class="col-4">` (Bootstrap
flex grid) with a `<table class="sig-table">` containing one row
of three 33% tds. wkhtmltopdf treats table rows as atomic for
page-breaking, so the whole signature row now stays on a single
page.
Verified with pypdf: page 1 has the cert statement, page 2 has
all three signature labels together — no more sliced boxes.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Landscape BoL was splitting the signature row down the middle —
boxes half on page 1, half on page 2. Two complementary fixes:
1. **Per-element rule**: added `page-break-inside: avoid` +
`break-inside: avoid` to `.sig-box` (both portrait + landscape
styles) so an individual signature box can never split across
pages.
2. **Wrapper rule**: introduced `.fp-keep-together` utility +
wrapped the BoL's certification statement + signature row in
it, so the whole "sign here" block moves to the next page as
one unit if it doesn't fit. Also applied
`page-break-inside: avoid` to `table tr` so cargo lines don't
split mid-row either.
Lives in shared `report_base_styles.xml` so any FP template that
opts into `.fp-keep-together` benefits automatically.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Five fixes applied to the Bill of Lading and (where relevant) all
report templates:
1. **Bigger title + BoL #** — portrait now uses h2 24pt (was h4 16pt),
landscape h2 26pt; BoL # ticker is 13/14pt instead of body size.
2. **Shipper info missing** — root cause: `_fp_build_delivery_vals`
was creating deliveries without `company_id`, so the BoL's
`<span t-field="doc.company_id.name"/>` rendered empty. Two fixes:
- Hook now sets `company_id = mo.company_id.id or env.company.id`.
- Template falls back defensively to `env.company` when
`doc.company_id` is empty (covers any legacy delivery that
somehow slips through without it).
- Backfilled 14 existing deliveries via SQL on entech.
3. **Uniform header backgrounds** — replaced mixed `info-header`
(gray) + default-th (brand black) headers with a single
`fp-header-primary` (brand black) across all sub-tables for a
consistent look.
4. **Cargo description alignment + missing column** — added a QTY
column (matches landscape variant), pulled from the linked MO
via job_ref → mrp.production.product_qty. Added `.fp-cell-mid`
utility class with `vertical-align: middle !important;` and
applied it to every cargo + info cell so values sit centred
instead of jammed against the top border.
5. **Signature box too short** — bumped `.sig-box` from 70 → 110 px
(portrait) / 130 px (landscape), `.sig-line` from 28 → 60/70 px,
added flex layout so the label sits at the bottom and signers
have a real space to write in. Lives in the shared
`report_base_styles.xml` so EVERY FP template benefits, not just
the BoL.
Verified: BoL portrait renders cleanly at 140 KB with full shipper
block + uniform headers + middle-aligned cargo cells.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Odoo 19 renamed the m2m-to-groups fields on res.users:
- groups_id (Odoo <=18) was split into group_ids (direct) +
all_group_ids (direct + implied)
The /book-assessment route was raising KeyError: 'groups_id' on every hit,
returning HTTP 500. Switched to all_group_ids so any user with the sales
salesman group access (direct OR via implied manager/admin groups) is
matched when resolving available sales reps.
Verified by curl: /book-assessment now returns HTTP 200.
Made-with: Cursor
Task 18 — empirical verification of the data-preservation claims in
Section 3 of the Enterprise Takeover Roadmap.
Key empirical findings (verified on westin-v19 live DB + clone):
1. Safety guard blocks Enterprise uninstall (Scenario A, verified on
throwaway clone) — UserError fires with the correct migration-wizard
guidance message.
2. Bank reconciliation tables (account.partial.reconcile,
account.full.reconcile) are owned exclusively by Community account
module. 30,874 reconciliation rows (16,500 partial + 14,374 full)
confirmed immune to any Enterprise uninstall.
3. All 5 Enterprise extension fields on account.move (deferred_move_ids,
deferred_original_move_ids, deferred_entry_type, signing_user,
payment_state_before_switch) are dual-owned by account_accountant
AND fusion_accounting_core. Odoo's module-ownership ledger will
preserve columns/relations when Enterprise uninstalls.
4. account.reconcile.model is triple-owned (account + account_accountant
+ fusion_accounting_core). Reconciliation rules survive.
5. account.move has 36 module owners; table cannot be dropped by any
realistic uninstall scenario.
A full destructive uninstall cycle on a clone was attempted but blocked
by pre-existing data-integrity issues in westin-v19 (orphan FK references
in payslip_tags_table + account_account_res_company_rel — outside fusion
scope). The schema-ownership verification approach provides stronger
evidence than a point-in-time count comparison — it proves the invariants
hold for any real-world data shape, not just a single fixture.
Test clone westin-v19-phase0-empirical dropped after testing. No live
data was modified.
Phase 0 data-preservation design is empirically validated. Phase 1 can
proceed.
Made-with: Cursor
The Bill of Lading template assigned a temp variable
`<t t-set="dest" t-value="doc.delivery_address_id or doc.partner_id"/>`
and then tried `<div t-field="dest" .../>`. Odoo 19 QWeb asserts
t-field must be `record.field_name` (have a dot) — the temp variable
form fails compilation and the report renders as a multi-page
"Oops! Something went wrong" PDF stuffed with the traceback.
Fix: branch with `t-if`/`t-else` and call `t-field="doc.delivery_address_id"`
or `t-field="doc.partner_id"` directly. Same pattern in both header
and second-page-header sections (lines 49/235).
Verified: BoL render goes from 39 KB error page to 138 KB clean PDF.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
