Contributions

Contributions are the immutable evidence records that feed into a record's materialised view. Each contribution represents a single data ingestion event — a document extraction, an integration response, a manual entry, or an API write. They are a first-class resource used for provenance, compliance, and audit.

Read endpoints

Method	Path	Description
GET	/contributions	List contributions
GET	/contributions/{contributionId}	Get a contribution
GET	/records/{recordId}/contributions	List contributions bound to a record
GET	/contexts/{contextId}/contributions	List contributions within a context

Write endpoints

Method	Path	Description
POST	/contributions	Ingest a contribution
POST	/contributions/bulk	Batch ingest contributions

Ingest a contribution

POST /contributions

Request

{
  "contextId": "01957c3a-5678-7000-8000-000000000001",
  "modelId": "01957c3a-1234-7000-8000-000000000001",
  "sourceId": "revenue-integration-2026-01",
  "sourceType": "integration",
  "data": {
    "annualIncome": 54800
  },
  "fieldConfidences": {
    "annualIncome": 0.99
  }
}

Field	Required	Description
contextId	Yes	The decisioning context this evidence belongs to
modelId	Yes	The model this evidence is for
sourceId	Yes	Identifier for the data source (e.g., a document ID, integration name)
sourceType	Yes	Category of the source: document_extraction, integration, manual_entry, api
data	Yes	The evidence payload — field names and values
fieldConfidences	No	Per-field confidence scores (0.0 to 1.0). Omitted fields default to 1.0.
supersedes	No	ID of a previous contribution to supersede

Response 201 Created

{
  "contributionId": "01957c3a-9abc-7000-8000-000000000001",
  "contextId": "01957c3a-5678-7000-8000-000000000001",
  "modelId": "01957c3a-1234-7000-8000-000000000001",
  "sourceId": "revenue-integration-2026-01",
  "sourceType": "integration",
  "data": {
    "annualIncome": 54800
  },
  "fieldConfidences": {
    "annualIncome": 0.99
  },
  "createdAt": "2026-02-20T10:05:00Z"
}

The contribution is persisted and the containing context is marked stale for reprocessing. The platform's resolution pipeline will bind this contribution to a record and materialise the result.

Superseding contributions

If a source sends corrected data, create a new contribution that supersedes the previous one:

{
  "contextId": "01957c3a-...",
  "modelId": "01957c3a-...",
  "sourceId": "revenue-integration-2026-01",
  "sourceType": "integration",
  "data": {
    "annualIncome": 48000
  },
  "fieldConfidences": {
    "annualIncome": 0.99
  },
  "supersedes": "01957c3a-9abc-7000-8000-000000000001"
}

The platform validates that both contributions share the same sourceId, marks the old contribution as superseded (excluded from resolution but retained for audit), and processes the new one as active.

The original contribution is never modified. Superseding creates a new contribution and changes how the platform interprets the existing evidence.

Get a contribution

GET /contributions/{contributionId}

Response 200 OK

{
  "contributionId": "01957c3a-9abc-7000-8000-000000000001",
  "contextId": "01957c3a-...",
  "modelId": "01957c3a-...",
  "sourceId": "revenue-integration-2026-01",
  "sourceType": "integration",
  "data": {
    "annualIncome": 54800
  },
  "fieldConfidences": {
    "annualIncome": 0.99
  },
  "targetRecordId": "01957c3a-def0-7000-8000-000000000001",
  "fieldMappings": {
    "annualIncome": "annualIncome"
  },
  "createdAt": "2026-02-20T10:05:00Z"
}

When the contribution has been bound to a record by the resolution pipeline, targetRecordId and fieldMappings are populated.

List contributions

GET /contributions
GET /records/{recordId}/contributions
GET /contexts/{contextId}/contributions

All list endpoints support standard filtering:

GET /contributions?filter[sourceType][eq]=document_extraction
GET /contributions?filter[contextId][eq]=uuid
GET /contributions?filter[createdAt][gte]=2026-01-01T00:00:00Z

Bulk ingest

POST /contributions/bulk

Accepts an array of up to 100 contributions. See Bulk Operations for the response format and partial success semantics.

Binding

Contributions are not bound to records at ingestion time. Binding — associating a contribution with a target record and field mappings — is handled by the platform's resolution pipeline. This ensures bindings are created as part of an atomic resolution transaction.

You can observe binding status by reading a contribution after resolution completes: the targetRecordId and fieldMappings fields will be populated.