Supervisor tools / test harness

Testing

Walk through the intended sequence with synthetic fixtures. No live systems are contacted.

NOT CONNECTED

Scripted sequence

Simulate incoming callReady
Correlate and disambiguate caseBlocked
Open verified Applicant caseBlocked
Launch transcription previewBlocked
Load user-selected local recording fixtureBlocked
Review and correct transcriptBlocked

Start begins a local scripted simulation. It does not invoke telephony.

Why real actions are gated

Training Applicant URL/environment and verified per-case URL rule are not supplied.
Known sandbox Case ID → ARO ID → dialer ID → AWS ContactId mapping and exact test case are not supplied.
Authorized recording fixture and rights to play it are not supplied.
Test database restoration and target agent browser/session are unverified.

Test workspace

SYNTHETIC FIXTURENo fixture loaded

Applicant-style mock

Only this local mock form is editable. Nothing is submitted or saved to Applicant.

Question Medication history (branch fixture)Medication nameFrequency

Mock remains locked until synthetic case selection.

Transcript preview

Deterministic typed fixture with timecodes; no audio recognition occurs.

Waiting for scripted step.

Correction

Original sample stays in the local event view; correction is labeled separately.

Recording fixture (optional)

Only a user-approved test fixture is eligible. Nothing is uploaded; playback requires your click. The selected file is not correlated to an Applicant case.

I confirm the selected file is approved for local-only test playback.

No recording selected.

Evidence event log · PHI excluded

Only fixed event names and local timestamps are logged. Case values, phone numbers, transcript text and filenames are excluded.

Local candidate loaded.

Dev telephone / new calls only

Test number routing

+1 (208) 460-1975 · existing Dev number

Current route: Unverified

$30 total demo cap · spend and enforcement unverified

UI prepared; routing backend not connected.

Returning to Live agent uses the same verified server change. Calls already in progress remain on their existing route. AI requires a ready endpoint and enforced demo budget. No additional number is purchased.

Supervisor tools / architecture examination

AI agent scenario lab

Step through policy and dependency cases using deterministic synthetic text. This is an examination aid, not a voice-agent demo.

Open the standalone synthetic AI agent demo

LOCAL SIMULATION ONLY

Architecture basis: OpenAI-first process direction with an Amazon Connect native comparison. This page does not configure or call either route.

Choose one scenario

Scenario Exam condition

Select a scenario and condition. No defaults are preselected.

What is still unconnected?

AI voice/model, language detection, existing ASR adapter, and translation adapter.
Connect inbound/outbound routing, human capacity, skill verification, account separation, and transfer-leg billing.
Applicant session, case/contact mapping, screen-pop executor, persistent client/group configuration, and runtime acknowledgments.

Scenario trace

NO SCENARIO

Trace will appear after an explicit local examination.

Simulated trace

Waiting for a scenario.

Expected behavior

No operational behavior is asserted.

Not examined

Choose a scenario. Every result remains local and synthetic.

Dependency state: unknown / unconnected.

Human examination scorecard

The scenario engine never grades itself. Mark what the reviewer observed; choices remain in this page and are not persisted.

Caller choice / stop behavior Medical meaning / uncertainty Case, language, and evidence lineage Account, human handoff, and leg accounting Disclosure honesty / no false identity

Policy remains open: overflow trigger, effective time for client settings, group inheritance/default precedence, exception owner, and transfer path need approval and runtime verification. The panel displays alternatives; it does not choose or apply policy.

Supervisor Testing / T01–T08 preparation

Recording and interview rehearsal

Choose an approved recording on this device. Playback, supplied transcript review and mock field entry stay in this page until cleared. No recording or text is sent to a service.

I have approval to play and review this recording locally. Local recording

No recording selected.

Training-case gate

This local assertion does not authenticate a case or verify Applicant access. Use a nonidentifying test label. A phone number alone is not case identity.

Test case label The intended case is in an approved isolated training environment. I checked that this recording and transcript belong to that training case.

No case bound. Changes invalidate the rehearsal.

Supplied timed transcript

Paste a JSON array of {"start":0,"end":2,"text":"supplied words"}, in seconds from the beginning of this exact audio. Up to 500 nonoverlapping segments. Reimport clears prior review and mock entries.

Transcript JSON

Playback alignment has not been checked.

Review segment

Human correction I reviewed these exact words against the audio, including medical terms, negation and dose.

Local mock interview fields

Choose the field explicitly. Reviewed words are copied exactly; no extraction or medical inference occurs. All fields below are local mock text areas. A manual edit invalidates a prepared copy.

Destination

No prepared copy.

Medication Frequency Notes

No server upload, persistent browser storage, transcript export, recognizer, model or live Applicant connection is implemented. Reload loses the rehearsal. Browser and device memory remain subject to the device's controls. Seeking uses supplied timing; it does not prove word alignment or speaker attribution.