Skip to main content

Sandbox decision flow

End-to-end technical proof of the Decision Platform path in sandbox / founder-beta. Field names match the Decision API contract (packages/contracts, OpenAPI at GET /openapi.json on a provisioned host). This is not a live lending, payments, or capital facility. Hosts may still be coming online, use the base URL Vence issues for your org.

1. Create the entity

Exact response envelope may include additional bookkeeping fields; keep external_id stable.
Wallet links feed the Trust Graph. Sandbox demos use simulated addresses, not production chain data. Consent is scoped to the product you will underwrite:

4. Ingest transactions (feature evidence inputs)

Repeat for the entity’s history. Prefer normalized transactions over opaque blobs so features and policy stay explainable.

5. Build a feature snapshot

The Feature Engine (baseline-v1) produces versioned features with lineage (completeness, stability, anomaly indicators, concentration, repayment history, and related keys). Underwrite can also compute a snapshot from stored transactions if none exists yet.

6. Underwrite (policy → decision)

Product selects the policy pack (Stage 2 product dimension). Example: business_credit_linesandbox-policy-v1.
id is the decision id for retrieve / replay / review / outcomes. Decision evidence (feature snapshot id, policy checksum, rule trace) is persisted server-side, use replay, not a free-form evidence array on this body. Score vs decision: an optional external credit score (bureau or third-party scorer) can plug in via a provider interface later. It is an input signal, not the Vence verdict. The same entity can receive different verdicts under different product packs (payroll advance vs business line vs market-maker facility). Vence does not claim partnership with any specific scorer.

7. Replay (reproducibility)

Replay re-evaluates the stored feature snapshot against the same policy version. Exact match confirms auditability. Illustrative shape:

8. Audit trail

Also retrieve the canonical decision: GET /v1/decisions/{decision_id}.

9. Record an outcome (Trust Graph feedback)

Outcomes close the loop: monitoring and future underwrites can consume updated Trust Graph evidence.

What “good” looks like

You can walk a risk committee from entity and wallet links through features, policy version, checksum, reasons, replay, audit, and outcome without a black box, in sandbox today, with honest founder-beta host labels.