Case Review Console

Sentra Console is a Next.js application for analyst and admin workflows. It is used for demos, external walkthroughs, analyst queue operation, manual scoring, data import, model inspection, and threshold configuration.

Cases page

Path: /cases

The Cases page is the analyst queue. It supports:

• Search by case ID or transaction ID.
• Filter by AI signal state: all, pending, ready.
• Filter by queue status: OPEN, IN_REVIEW, CLOSED.
• Filter by decision: PASS, REVIEW, HOLD, BLOCK.
• Pagination with a default page size of 20.

Rows show:

• Case ID.
• Transaction ID.
• Queue status.
• Amount and currency.
• Effective outcome.
• Model decision and manual override state.
• AI signal badge.
• Last update time.

Case detail page

Path: /cases/{case_id}

The detail page includes:

• Transaction fields: amount, currency, route, PSP, status, fee, sender, receiver.
• Sender and receiver entity context.
• Latest risk output: score, decision, reasons, evidence, model version, LLM version, latency, request ID.
• Recent score history for the same transaction.
• Feedback label history.
• Activity timeline.
• Analyst notes.
• Final decision override.
• Case status update.
• Admin-only Run AI Analysis button.

Analyst workflow

1. Open /cases.
2. Filter to AI Signals Ready for enriched review, or AI Signals Pending for items needing refresh/trigger.
3. Open a case.
4. Review risk score, reasons, evidence, transaction details, and entity context.
5. Set status to IN_REVIEW if investigation is active.
6. Add analyst notes.
7. Choose a final decision when resolved.
8. Submit a feedback label so the outcome can improve future training.
9. Close the case.

Manual Risk Check

Path: /manual-score

Manual Risk Check is a JSON editor for the live scoring contract.

Actions:

• Load example: loads a valid sample request.
• Generate random example: creates a realistic payment-risk request.
• Score Now: calls /v1/risk/score.
• Save as Case: calls /v1/cases/from-score for the current transaction.

Use Manual Risk Check to:

• Smoke test a deployment.
• Reproduce a payment platform request.
• Validate new feature overrides.
• Demonstrate low-latency decisioning and async AI behavior.

Admin Console

Path: /admin

Required role: Admin.

Admin includes:

• Data Ops: live data contract and downloadable templates.
• Data Imports: ingest run list, row counts, quality checks, and errors.
• Models: model registry, metrics, source dataset, and activation controls.
• Config: decision thresholds and prompt version.

Demo vs production mode

Production mode must not depend on mock auth, mock LLM, local file storage, or in-memory queues.