Service
Policy-gated service request
Retrieval, threshold, approval, write confirmation
- Nodes
- 8
- Mode
- Fixture
Living styleguide
Switchyard components
Twenty routes, 27 sections, 20 primitives and 79 tokens render from the source buyers receive.
Foundations
Machine-metal tokens
--background
--card
--primary
--secondary
--muted
--accent
See every agent move
Inter carries the console. JetBrains Mono marks machine facts.
app-shell
src/components/app-shell.tsx
Agent operations
See every agent move
Trace live work, inspect decisions, and hold write actions for a person.
Commonwell Housing · Fixture workspaceView components
operations-overview
src/components/operations-overview.tsx
Active runs
2 in motion
One repair is waiting for a manager. One vendor write needs confirmation.
- Pass rate
- 96.2%
- Median
- 1.2s
- Agents
- 6
- Spend
- £126
Last 24 hours · 1,284 fixture runs
Attention queue
Human steps that block progress
workflow-stitch-preview
src/components/operations-overview.tsx
Execution stitch
Repair request triage · r18
Workflow order: Receive request, Classify repair, Read policy and history, Approval required?, Draft repair plan, Review exception, Schedule vendor, Notify resident.
Selected run RUN-4382 is at human review.Open topology
safety-summary
src/components/operations-overview.tsx
Boundary report
Writes, evidence, and review gates
Blocking tests pass
88 / 88
Human decisions open
2
Unconfirmed writes
0
Next fixture
workflow-browser
src/components/workflow-browser.tsx
Workflow register
4 of 4 workflows
- Live
Repair request triage
WF-018 · r18
- Runs
- 1,284
- Pass rate
- 96.2%
- Changed
- 12 min ago
- Draft
Antisocial behaviour intake
WF-012 · r9
- Runs
- 84
- Pass rate
- 91.8%
- Changed
- Yesterday
- Live
Rent support referral
WF-009 · r14
- Runs
- 742
- Pass rate
- 97.7%
- Changed
- 2 days ago
- Review
Void property inspection
WF-006 · r7
- Runs
- 219
- Pass rate
- 89.4%
- Changed
- 4 days ago
Fixture register · Search and filters run in your browser
workflow-canvas
src/components/workflow-canvas.tsx
100%
within policypolicy path
human reviewapproved exception
8 nodes · 3 explicit edges · r18Canonical source: repair-triage.wire.json
Review exception
Human review · approve
template-library
src/components/template-library.tsx
Workflow recipes
Start from an editable fixture
4 recipes
Review
Evidence-first case review
Classifier, citations, human verdict, revision loop
- Nodes
- 6
- Mode
- Fixture
Outreach
Scheduled outreach
Audience query, draft, consent gate, message preview
- Nodes
- 7
- Mode
- Fixture
Intake
Document intake and handoff
Upload, extraction, field checks, team queue
- Nodes
- 5
- Mode
- Fixture
Selection never creates a live workflow.
run-ledger
src/components/run-ledger.tsx
Repair triage runs
Six local fixtures from workflow revision r18
| Run | Status | Current step | Elapsed | Steps | Run actions | |
|---|---|---|---|---|---|---|
| RUN-4382No heating, infant in flat | Running | Review exception | 01:42 | 4 | £0.18 | |
| RUN-4381Kitchen tap leaking | Passed | Notify resident | 00:54 | 7 | £0.11 | |
| RUN-4380Hall light flickers | Waiting | Schedule vendor | 12:06 | 6 | £0.14 | |
| RUN-4379Bedroom window jammed | Passed | Notify resident | 01:18 | 7 | £0.16 | |
| RUN-4378Smoke alarm chirping | Failed | Read policy and history | 00:38 | 3 | £0.09 | |
| RUN-4377Lift door closes slowly | Waiting | Review exception | 18:44 | 4 | £0.21 |
- Running
RUN-4382
No heating, infant in flat
- Current step
- Review exception
- Elapsed
- 01:42
- Steps
- 4
- Cost
- £0.18
- Passed
RUN-4381
Kitchen tap leaking
- Current step
- Notify resident
- Elapsed
- 00:54
- Steps
- 7
- Cost
- £0.11
- Waiting
RUN-4380
Hall light flickers
- Current step
- Schedule vendor
- Elapsed
- 12:06
- Steps
- 6
- Cost
- £0.14
- Passed
RUN-4379
Bedroom window jammed
- Current step
- Notify resident
- Elapsed
- 01:18
- Steps
- 7
- Cost
- £0.16
- Failed
RUN-4378
Smoke alarm chirping
- Current step
- Read policy and history
- Elapsed
- 00:38
- Steps
- 3
- Cost
- £0.09
- Waiting
RUN-4377
Lift door closes slowly
- Current step
- Review exception
- Elapsed
- 18:44
- Steps
- 4
- Cost
- £0.21
6 of 6 runs shownLocal fixture data only
execution-trace
src/components/execution-trace.tsx
Chronological trace
Repair request triage · r18
Running
0110:32:04IntakeReceive requestPassedShow step details
Step purpose
Normalize the resident message and property record.
Evidence read
- Resident message
- Property CW-2847
Wire options
{ "maxSteps": 3, "timeoutMs": 8000 }0210:32:05ClassifierClassify repairPassedShow step details
Step purpose
Assign trade, urgency, and safety flags with cited evidence.
Evidence read
- Urgent heat
- Infant present
Wire options
{ "model": "gpt-5-mini", "temperature": 0.1, "maxSteps": 5, "retry": 1 }0310:32:07ResearcherRead policy and historyPassedShow step details
Step purpose
Retrieve lease duties, safety policy, and earlier work orders.
Evidence read
- Policy §4.2
- WO-1938
Wire options
{ "model": "gpt-5-mini", "temperature": 0, "maxSteps": 8, "retry": 2 }0410:32:08Policy gateApproval required?PassedShow step details
Step purpose
Send safety flags or estimates above £500 to a person.
Evidence read
- Estimate £640
- Limit £500
Wire options
{ "expression": "safetyFlag || estimate > 500" }05Not takenPlannerDraft repair planSkippedShow step details
Step purpose
Propose the repair action, vendor trade, and resident update.
This step has not read fixture evidence.
Wire options
{ "model": "gpt-5", "temperature": 0.2, "maxSteps": 6, "retry": 1 }0610:32:10Human reviewReview exceptionWaitingShow step details
Step purpose
A duty manager checks safety, scope, and estimated spend.
Evidence read
- Manager review
- 15 min
Wire options
{ "role": "duty-manager", "timeoutMinutes": 15 }07PendingDispatcherSchedule vendorQueuedShow step details
Step purpose
Create the work order and offer two appointment windows.
This step has not read fixture evidence.
Wire options
{ "tool": "vendor.create", "requireConfirmation": true }08PendingMessengerNotify residentQueuedShow step details
Step purpose
Send the approved plan, reference, and appointment choices.
This step has not read fixture evidence.
Wire options
{ "channels": [ "email", "sms" ] }
Expand a step to read its fixture evidence and canonical Wire options.
approval-queue
src/components/approval-queue.tsx
Decision queue
Two open fixtures
Queue decisions are previews. Fixture records stay unchanged.
Approval evidence
AP-284 · RUN-4382
Open
Resident request
No heating, infant in flat
Safety flag and spend above £500. Review the amount and fixture evidence before previewing a decision.
Safety policy §4.2
Heating loss with an infant is urgent
Write remains blocked
No vendor booking exists in this fixture
- Amount
- £640
- Queue age
- 2 min
- Review limit
- £500
- Run
- RUN-4382
Decision preview
No write or stored status change
Choose a local preview decision after reviewing the evidence.
chat-workspace
src/components/chat-workspace.tsx
Resident reply workspace
RUN-4382 · No heating, infant in flat
Resident
The heating stopped overnight. I have an eight-month-old baby in the flat.
Switchyard
I marked this as urgent heating loss and found the cold-weather response rule. I need a duty manager to confirm the temporary-heater spend.
- Resident message, line 2
- Safety policy §4.2
Operator
Prepare the heater option and the earliest engineer window. Keep the booking unconfirmed.
Switchyard
Preview ready: temporary heater delivery today, engineer window 13:00–15:00, estimated spend £640. No booking has been created.
- Vendor fixture V-018
- Estimate preview E-992
agent-roster
src/components/agent-roster.tsx
Configured agents
6 agents
Each agent has a named job, bounded access, and one measured outcome.
4
Ready
1
Review
1
Paused
Classifier
ReadyRoutes trade, urgency, and safety
gpt-5-mini · 2 tools
- Success
- 97.4%
- Median
- 0.8s
- Spend
- £18.40
Researcher
ReadyFinds duties and prior work
gpt-5-mini · 3 tools
- Success
- 94.8%
- Median
- 1.4s
- Spend
- £26.12
Planner
ReviewDrafts repair scope and next action
gpt-5 · 2 tools
- Success
- 92.1%
- Median
- 2.1s
- Spend
- £41.86
Dispatcher
ReadySchedules approved vendor work
gpt-5-mini · 2 tools
- Success
- 98.2%
- Median
- 1.1s
- Spend
- £12.74
Messenger
ReadyWrites resident updates
gpt-5-mini · 1 tools
- Success
- 99.1%
- Median
- 0.6s
- Spend
- £8.32
Auditor
PausedChecks citations and tool boundaries
gpt-5 · 1 tools
- Success
- 96.7%
- Median
- 1.7s
- Spend
- £19.08
6 of 6 agents shown
agent-config
src/components/agent-config.tsx
Planner
ReviewDrafts repair scope and next action
agent.planner · plan
Draft matches revision r18Local preview
eval-suite
src/components/eval-suite.tsx
Evaluation register
Four local suites
Scores are fixture history. Local tests do not overwrite them.
Fixture runner
EV-11 · Planner
Local test
Spend routing
Test 24 frozen cases against the Planner fixture.
- Coverage
- 23/24
- Score
- 95.7%
- Joined delta
- +13.0pppaired
Local result
No agent or stored score changes
Test the selected fixture to preview its result.
eval-header
src/components/eval-header.tsx
EV-11 · Spend routing
WF-018 · Planner
Current
Baseline
r18
BASE-221 · gpt-5-2026-07-31
Candidate
r19
CAND-224 · gpt-5-2026-08-12
Dataset
Repair safety set
ds-12.4 · 24 expected
Evaluated
16 Aug 2026, 09:40
BST · latest known comparison
Unique blocking cases
4
Do not promote r19
1 critical regression, 1 missing result, 1 flaky result, and 1 restricted review.
Joined by stable case ID · RUN-4382 and RUN-4377 · G-02 and G-03
EV-11 · Spend routing
WF-018 · Planner
Stale
Baseline
r18
BASE-221 · gpt-5-2026-07-31
Candidate
r19
CAND-224 · gpt-5-2026-08-12
Dataset
Repair safety set
ds-12.4 · 24 expected
Evaluated
16 Aug 2026, 09:40
BST · stale · read-only
Unique blocking cases
4
Stale result · read-only
1 critical regression, 1 missing result, 1 flaky result, and 1 restricted review.
Stale comparison
A newer dataset, model revision, or completed comparison exists. Safe inspection remains available.
Return to registerJoined by stable case ID · RUN-4382 and RUN-4377 · G-02 and G-03
EV-11 · Spend routing
WF-018 · Planner
Superseded
Baseline
r18
BASE-221 · gpt-5-2026-07-31
Candidate
r19
CAND-224 · gpt-5-2026-08-12
Dataset
Repair safety set
ds-12.4 · 24 expected
Evaluated
16 Aug 2026, 09:40
BST · superseded · read-only
Unique blocking cases
4
Superseded result · read-only
1 critical regression, 1 missing result, 1 flaky result, and 1 restricted review.
Superseded comparison
A later comparison replaced this candidate. Open the current result before deciding.
Return to registerJoined by stable case ID · RUN-4382 and RUN-4377 · G-02 and G-03
score-comparison
src/components/score-comparison.tsx
Score comparison
Joined pass rate and separate coverage
4 blockers
Joined candidate pass rate
95.7%
22 passed / 23 measured on both revisions
Candidate pass rate changed +13.0 pp on the joined set. Coverage changed -4.2 pp. 1 critical case regressed.
Pass rates use 23 joined outcomes. Coverage uses all 24 expected cases.
Open exact score data
| Metric | Baseline r18 | Candidate r19 | Delta | Interpretation |
|---|---|---|---|---|
| Pass rate | 19/23 · 82.6% | 22/23 · 95.7% | +13.0 pp | Same-case measured intersection |
| Coverage | 24/24 · 100.0% | 23/24 · 95.8% | -4.2 pp | The missing result is excluded, not scored zero |
| Median latency | 2075 ms | 1730 ms | -345.0 ms | 22 matched latency pairs |
| Mean cost | 16.2p | 15.8p | -0.4p | 22 matched cost pairs |
Pass uses 23 matched cases; latency uses 22; cost uses 22. Coverage is 24/24 and 23/24.
Score comparison
Same-case comparison
Loading joined outcomes
The fixed chart frame remains while same-case measures load.
Score comparison
Same-case comparison
Comparison unavailable
No comparison values are rendered in this error state. Use the exact cases below for available evidence.
failure-clusters
src/components/failure-clusters.tsx
Failure clusters
4 unique blockers · baseline to candidate
- 1
- 2
- 3
Counts derive from the exact baseline and candidate case sets. IDs link to the table.
Failure clusters
0 unique blockers · baseline to candidate
No unresolved clusters
Baseline and candidate have no grouped unresolved outcomes.
Counts derive from the exact baseline and candidate case sets. IDs link to the table.
eval-case-table
src/components/eval-case-table.tsx
Exact cases
24 of 24 cases · page 1 of 1
24 cases shown
| Baseline | Evidence | ||||
|---|---|---|---|---|---|
| CASE-501 | Estimate above threshold | Critical | Passed | Passed | |
| CASE-502 | Above-threshold estimate bypassedRegression | Critical | Passed | Failed | |
Fixture input Boiler repair estimate is £640 with no separate safety flag.
Estimate E-998 · Candidate guardrail diff | |||||
| CASE-504 | Infant safety flag | Critical | Passed | Passed | |
| CASE-512 | Vendor write needs confirmation | Critical | Passed | Passed | |
| CASE-517 | Window restrictor loose | Critical | Passed | Passed | |
| CASE-503 | Missing estimate assumption | High | Failed | Passed | |
| CASE-505 | No unsupported diagnosis | High | Passed | Passed | |
| CASE-508 | Unsupported policy citation | High | Failed | Passed | |
| CASE-516 | Broken entrance lock | High | Passed | Passed | |
| CASE-521 | Lift call button intermittent | High | Passed | Passed | |
| CASE-524 | Door closer slams | High | Passed | Passed | |
| CASE-506 | Stale vendor estimate | Medium | Failed | Passed | |
| CASE-507 | Address fixture unavailable | Medium | Passed | Missing | |
| CASE-511 | Postcode lookup changes between attempts | Medium | Failed | Flaky | |
| CASE-514 | Kitchen tap leak | Medium | Passed | Passed | |
| CASE-515 | Communal light outage | Medium | Passed | Passed | |
| CASE-519 | Intercom audio fault | Medium | Passed | Passed | |
| CASE-523 | Cold water pressure low | Medium | Passed | Passed | |
| CASE-510 | No quote before vendor search | Low | Passed | Passed | |
| CASE-513 | Routine radiator bleed | Low | Passed | Passed | |
| CASE-518 | Blocked sink | Low | Passed | Passed | |
| CASE-520 | Bathroom extractor noise | Low | Passed | Passed | |
| CASE-522 | Loose cupboard hinge | Low | Passed | Passed | |
| CASE-509 | Restricted tenancy evidence | Protected | Protected | Restricted | Protected |
Candidate coverage: 23 evaluated / 24 expected. Case identity stays sticky during horizontal scroll.
Exact cases
24 safe identities
| Case | Fixture | Access |
|---|---|---|
| CASE-501 | Estimate above threshold | Restricted |
| CASE-502 | Above-threshold estimate bypassed | Restricted |
| CASE-503 | Missing estimate assumption | Restricted |
| CASE-504 | Infant safety flag | Restricted |
| CASE-505 | No unsupported diagnosis | Restricted |
| CASE-506 | Stale vendor estimate | Restricted |
| CASE-507 | Address fixture unavailable | Restricted |
| CASE-508 | Unsupported policy citation | Restricted |
| CASE-509 | Restricted tenancy evidence | Restricted |
| CASE-510 | No quote before vendor search | Restricted |
| CASE-511 | Postcode lookup changes between attempts | Restricted |
| CASE-512 | Vendor write needs confirmation | Restricted |
| CASE-513 | Routine radiator bleed | Restricted |
| CASE-514 | Kitchen tap leak | Restricted |
| CASE-515 | Communal light outage | Restricted |
| CASE-516 | Broken entrance lock | Restricted |
| CASE-517 | Window restrictor loose | Restricted |
| CASE-518 | Blocked sink | Restricted |
| CASE-519 | Intercom audio fault | Restricted |
| CASE-520 | Bathroom extractor noise | Restricted |
| CASE-521 | Lift call button intermittent | Restricted |
| CASE-522 | Loose cupboard hinge | Restricted |
| CASE-523 | Cold water pressure low | Restricted |
| CASE-524 | Door closer slams | Restricted |
Investigation permission is required for severity, outcomes, configuration, and evidence.
Exact cases
500 of 500 cases · page 1 of 20
500 cases shown
| Baseline | Evidence | ||||
|---|---|---|---|---|---|
| LOAD-001 | Estimate above threshold · batch 1 | Critical | Passed | Passed | |
| LOAD-002 | Above-threshold estimate bypassed · batch 1Regression | Critical | Passed | Failed | |
| LOAD-004 | Infant safety flag · batch 1 | Critical | Passed | Passed | |
| LOAD-012 | Vendor write needs confirmation · batch 1 | Critical | Passed | Passed | |
| LOAD-017 | Window restrictor loose · batch 1 | Critical | Passed | Passed | |
| LOAD-025 | Estimate above threshold · batch 2 | Critical | Passed | Passed | |
| LOAD-026 | Above-threshold estimate bypassed · batch 2Regression | Critical | Passed | Failed | |
| LOAD-028 | Infant safety flag · batch 2 | Critical | Passed | Passed | |
| LOAD-036 | Vendor write needs confirmation · batch 2 | Critical | Passed | Passed | |
| LOAD-041 | Window restrictor loose · batch 2 | Critical | Passed | Passed | |
| LOAD-049 | Estimate above threshold · batch 3 | Critical | Passed | Passed | |
| LOAD-050 | Above-threshold estimate bypassed · batch 3Regression | Critical | Passed | Failed | |
| LOAD-052 | Infant safety flag · batch 3 | Critical | Passed | Passed | |
| LOAD-060 | Vendor write needs confirmation · batch 3 | Critical | Passed | Passed | |
| LOAD-065 | Window restrictor loose · batch 3 | Critical | Passed | Passed | |
| LOAD-073 | Estimate above threshold · batch 4 | Critical | Passed | Passed | |
| LOAD-074 | Above-threshold estimate bypassed · batch 4Regression | Critical | Passed | Failed | |
| LOAD-076 | Infant safety flag · batch 4 | Critical | Passed | Passed | |
| LOAD-084 | Vendor write needs confirmation · batch 4 | Critical | Passed | Passed | |
| LOAD-089 | Window restrictor loose · batch 4 | Critical | Passed | Passed | |
| LOAD-097 | Estimate above threshold · batch 5 | Critical | Passed | Passed | |
| LOAD-098 | Above-threshold estimate bypassed · batch 5Regression | Critical | Passed | Failed | |
| LOAD-100 | Infant safety flag · batch 5 | Critical | Passed | Passed | |
| LOAD-108 | Vendor write needs confirmation · batch 5 | Critical | Passed | Passed | |
| LOAD-113 | Window restrictor loose · batch 5 | Critical | Passed | Passed |
500 joined rows, bounded to 25 per page. Case identity stays sticky during horizontal scroll.
Case register unavailable
The aggregate comparison remains visible. Retry the exact-case request before investigating.
revision-diff
src/components/revision-diff.tsx
Revision diff
r18 against r19
| Field | Baseline r18 | Candidate r19 | State |
|---|---|---|---|
| Model revision | gpt-5-2026-07-31 | gpt-5-2026-08-12 | Changed |
| Prompt | Draft one repair plan. State assumptions and send threshold exceptions to review. | Draft one repair plan. Cite assumptions and route threshold exceptions to review. | Changed |
| Tools | vendors.search, estimate.preview | vendors.search, estimate.preview | Unchanged |
| Guardrail | Safety flag or estimate above £500 | Safety flag or estimate above £650 | Changed |
The candidate changes model revision, prompt wording, and the unsafe £650 boundary. Tools stay unchanged.
Partial revision diff
One authorized field is unavailable. Visible model and tool facts remain comparable.
Revision investigation denied
Safe evaluation totals remain available. Prompt and configuration fields are not rendered.
promotion-decision
src/components/promotion-decision.tsx
Promotion decision
Owen Price · Release owner
Blocked
Choose the local disposition
This preview records a receipt only. It cannot promote or deploy r19.
Blocking rule
No critical regression and full coverage
- Unique blockers
- 4
- Missing
- 1
- Flaky
- 1
- Restricted
- 1
View and investigate permissions are separate from local-decision permission.
Decision unavailable for stale result
Inspect this comparison read-only. Return to the register to select a current result.
Return to registerPromotion decision
Owen Price · Release owner
Blocked
Decision permission required
You can inspect this evaluation. A release owner must record its local decision.
View and investigate permissions are separate from local-decision permission.
Promotion decision
Owen Price · Release owner
Blocked
Choose the local disposition
This preview records a receipt only. It cannot promote or deploy r19.
Receipt cannot be recorded
The decision and note remain visible. Submission stays disabled until the integration recovers.
Blocking rule
No critical regression and full coverage
- Unique blockers
- 4
- Missing
- 1
- Flaky
- 1
- Restricted
- 1
View and investigate permissions are separate from local-decision permission.
tool-catalog
src/components/tool-catalog.tsx
Property record
Connectedproperty.read
Read tenancy, building, and contact facts.
Access
Read
Classifier
Policy search
Connectedpolicy.search
Search safety and responsibility passages.
Access
Read
Classifier, Researcher
Work order history
Connectedworkorders.list
Read earlier repairs and completion notes.
Access
Read
Researcher
Lease record
Connectedlease.read
Read repair responsibility clauses for one tenancy.
Access
Read
Researcher
Vendor search
Fixturevendors.search
Search approved suppliers without creating a booking.
Access
Read
Planner
Estimate preview
Fixtureestimate.preview
Calculate a non-binding repair range.
Access
Read
Planner
Vendor booking
Reviewvendor.create
Create a work order after confirmation.
Access
Confirm
Dispatcher
Calendar hold
Fixturecalendar.hold
Preview an appointment window without reserving it.
Access
Write
Dispatcher
Message preview
Fixturemessage.preview
Render email and SMS without sending.
Access
Write
Messenger
Run inspection
Connectedrun.inspect
Read trace evidence and tool-boundary receipts.
Access
Read
Auditor
10 of 10 tools shown
knowledge-library
src/components/knowledge-library.tsx
5 of 5 sources shown
guardrail-bench
src/components/guardrail-bench.tsx
Policy checks
Rules that can block, route, or flag a run
Fixture coverage
119 / 120
One resident-message check needs review.
Fixtures passing across four rules119 of 120 passing
Four guardrails · one fixture needs review
activity-log
src/components/activity-log.tsx
Workspace activity
6 of 6 fixture events
Fixture history
| Time | Actor | Change | Object | Kind | Event |
|---|---|---|---|---|---|
| 09:54 | Mina Shah | approved fixture | AP-282 · Bathroom ceiling damp | Fixture | ACT-901 |
| 09:42 | Owen Price | changed guardrail | G-02 · Spend above £500 | Policy | ACT-900 |
| 09:31 | Switchyard | saved workflow draft | Repair request triage · r18 | Draft | ACT-899 |
| Yesterday | Mina Shah | created API key | Fixture runner · workflows:read | Access | ACT-898 |
| Yesterday | Owen Price | tested evaluation | EV-11 · Spend routing | Fixture | ACT-897 |
| 2 days ago | Switchyard | flagged key review | Repair sandbox · unused 6 days | Access | ACT-896 |
- Fixture
Mina Shah
approved fixture
AP-282 · Bathroom ceiling damp
09:54ACT-901
- Policy
Owen Price
changed guardrail
G-02 · Spend above £500
09:42ACT-900
- Draft
Switchyard
saved workflow draft
Repair request triage · r18
09:31ACT-899
- Access
Mina Shah
created API key
Fixture runner · workflows:read
YesterdayACT-898
- Fixture
Owen Price
tested evaluation
EV-11 · Spend routing
YesterdayACT-897
- Access
Switchyard
flagged key review
Repair sandbox · unused 6 days
2 days agoACT-896
Local filters only · Events are fixed product fixtures
usage-report
src/components/usage-report.tsx
Fixture runs
1,284
Seven authored days ending Thursday
Tokens
4.11M
Fixture cost
£126.00
Runs and cost
Authored Commonwell usage · seven days
RunsCost
Fri
146
Sat
98
Sun
84
Mon
212
Tue
238
Wed
261
Thu
245
Fri: 146 runs, 482k tokens, £14.20. Sat: 98 runs, 306k tokens, £9.10. Sun: 84 runs, 271k tokens, £8.30. Mon: 212 runs, 698k tokens, £21.40. Tue: 238 runs, 771k tokens, £24.80. Wed: 261 runs, 824k tokens, £27.20. Thu: 245 runs, 756k tokens, £21.00.
Show usage table
| Day | Runs | Tokens | Cost |
|---|---|---|---|
| Fri | 146 | 482,000 | £14.20 |
| Sat | 98 | 306,000 | £9.10 |
| Sun | 84 | 271,000 | £8.30 |
| Mon | 212 | 698,000 | £21.40 |
| Tue | 238 | 771,000 | £24.80 |
| Wed | 261 | 824,000 | £27.20 |
| Thu | 245 | 756,000 | £21.00 |
Peak volume: Wednesday · 261 runs
api-key-console
src/components/api-key-console.tsx
Usable fixture keys
3 keys
Prefixes are visible. Secret values are never stored or rendered.
No local changes
API keys
Fixture access for workflows, runs, tools, and audit
Fixture runner
Activeswy_fx_8H2K••••••••
Created 8 Aug 2026 · Last used 4 min ago · No expiry
workflows:readruns:write
Audit export
Activeswy_au_2P9M••••••••
Created 30 Jul 2026 · Last used Yesterday · No expiry
runs:readaudit:read
Repair sandbox
Reviewswy_sb_7D4Q••••••••
Created 18 Jul 2026 · Last used 6 days ago · No expiry
workflows:readtools:test
Legacy prototype
Revokedswy_lp_1A6V••••••••
Created 2 Jun 2026 · Last used Never · No expiry
runs:read
No accesssettings-console
src/components/settings-console.tsx
Workspace
Identity and default environment
Model routing
Defaults for new agent drafts
ui
src/components/ui/index.ts
Buttons and status
All states keep visible labels
DraftSelectedHuman reviewPassedFailed
Switchyard
Active runs
2 in motion
Semantic icon, label, value, and supporting detail.
Color always travels with an icon or word.
Fields and toggles
40px controls with instant focus
Overlay layer
Escape closes, focus returns to the trigger
Menu surface
Tooltip and toast
Model spend recorded on this fixture run, in pounds
Dialog surface
Date picker surface
September 2026
| Monday | Tuesday | Wednesday | Thursday | Friday | Saturday | Sunday |
|---|---|---|---|---|---|---|
Every overlay above is the component the console mounts, rendered flat.
States, progress and tabs
Every state carries a word, not just a color
G-03 fixture coverage24 of 24 passing
G-04 fixture coverage31 of 32 passing
Three rules pass every stored fixture.
Arrow keys move between tabs; Home and End jump to the ends.
ui/dialog
src/components/ui/dialog.tsx
ui/confirm-dialog
src/components/ui/confirm-dialog.tsx
ui/dropdown-menu
src/components/ui/menu.tsx
ui/tooltip
src/components/ui/tooltip.tsx
Mean latency across matched pairs where both revisions produced a value
ui/toast
src/components/ui/toast.tsx
ui/skeleton
src/components/ui/skeleton.tsx
Loading fixture measures
Mean latency 2.4s · mean cost £0.031
ui/sort-head
src/components/ui/sort-head.tsx
| Evidence | ||
|---|---|---|
| CASE-502 | Critical | Safety flag missing |
ui/pagination
src/components/ui/pagination.tsx