Decision Logs

Every fraud decision is recorded permanently. The records exist for reconciliation: afterwards you read which decision a payment received, which rules matched, and which channel the decision came from.

GET /api/v1/fraud/decision-logs            (list)
GET /api/v1/fraud/decision-logs/{traceId}  (single)

Authentication is the same as the Decision API: a bearer token carrying the fraud:view_decisions scope. Only your own account's records are returned — the tenant scope is resolved from the token.

List

Filters

Parameter	Type	Meaning
`paymentId`	string	The payment reference you sent in the decision request
`decision`	string	`allow` · `review` · `block` · `force_3d`
`integrationMode`	string	`standalone` for external callers
`productContext`	string	The product surface (for example `direct_api`, `payment_form`)
`limit`	number	Page size. Default 50, valid range 1–200
`cursor`	string	The `pageInfo.nextCursor` value of the previous response

limit is a page size, not a ceiling. Pass the page size you want and fetch the following pages with cursor; you reach every record however many there are.

Paging

The order is newest first, and the sort key is the record's creation instant combined with its id — so two decisions created in the same millisecond are never skipped at a page boundary.

pageInfo is a sibling of data, not nested inside it:

{
  "status": "success",
  "code": "SUCCESS",
  "data": [ { "traceId": "…" } ],
  "pageInfo": {
    "hasMore": true,
    "nextCursor": "eyJ2IjoxLCJ0IjoiMjAyNi0xMC0wNVQwOToxMjo0MC4xMjM0NTZaIiwiaWQiOiIwMUo4WlUifQ",
    "totalCount": null
  },
  "requestId": "req_01J8ZV..."
}

When hasMore is false the list is exhausted.
totalCount is always null on this endpoint: a total count is expensive and is not needed for paging.
The cursor is opaque. Do not parse it; send it back verbatim. A corrupted cursor does not silently fall back to the first page — it returns 400 INVALID_CURSOR.

The full platform paging contract: Pagination.

Single record

GET /api/v1/fraud/decision-logs/{traceId}

You take traceId from the data.traceId field of the decision response. If you want to keep a reference to a decision, that is the only value you need to store.

Record fields

Field	Meaning
`traceId`	The decision's unique id; the value to quote in a support request
`decision`	`allow` · `review` · `block` · `force_3d`
`score`	Risk score
`reasonCodes`	The reason codes that produced the decision
`matchedRules` / `matchedRuleGroups`	Matched rules and rule groups
`flowPath`	The flow nodes the decision passed through
`flowId` / `flowName` / `flowVersion`	Id, name and version of the flow that decided
`explainabilityJson`	Explainability summary and group results
`tags` / `actions`	Tags attached to the decision and the actions that fired
`channel` / `surface` / `stage` / `operationKind`	The request's channel taxonomy
`authModeRequested` / `authModeApplied`	Requested and applied authentication mode
`integrationMode` / `productContext`	Integration mode and product surface
`subMerchantRef`	Sub-merchant reference, if you sent one
`transactionRef`	Transaction reference, if you sent one
`assertionOutcome`	Verification outcome of the identity artefact (on the `merchant_form` channel)
`paymentId` / `customerId`	The references you sent in the request
`latencyMs`	How long the decision took to produce
`createdAt`	When the decision was created

The record also carries the request that produced the decision. That is part of the audit trail: when asked why a decision was made, the context it was made with can be seen too.

Reading the decision for a payment

For payments taken through the Lunixi payment orchestrator, fraud already runs inside authorisation; you do not need to call evaluate separately. You read that payment's decision with the paymentId filter:

curl "https://api-gateway.lunixi.com/api/v1/fraud/decision-logs?paymentId=pi_123&limit=1" \
  -H "Authorization: Bearer $LUNIXI_TOKEN"

Decision API — request fields and the decision response
Event Ingest API — feeding the rules that count
Errors, repeated requests and retries

Event Ingest API Errors and Retries