Outcome sources

An outcome source connects a system of record and keeps an evaluation file current from it. Where a source maps records onto the chat format to train on, an outcome source names the two fields that make a scored case: the field holding the case and the field holding the result your business already wrote down for it. Point it at the CRM, the ticketing queue or the ERP where the answer lives, and every refresh writes a fresh set of cases with their real outcomes.

That is what turns a score into a business measure. The model is not being judged against a public benchmark; it is being judged against what actually happened, on the same records your team is already accountable for.

Reference it from a benchmark and every evaluation scores against the current rows, so a model can be re-scored as new records arrive and the result compared with every release before it.

An outcome source starts in connecting while its credentials are checked and its first records are read, then settles at ready or error. Each refresh runs running and then succeeded or failed, and the file it writes goes through the same validation every upload does. The source's file_id always names the latest ready version.

type is the catalogue slug of the system you are connecting, the same catalogue the training sources draw on, for example salesforce, zendesk or postgresql. The catalogue of 95 sources is on the Data connectors page of the Nucleus AI website.

Create an outcome source

curl https://api.nucleus-ai.io/v1/outcome-sources \
  -H "Authorization: Bearer $NUCLEUS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "type": "zendesk",
    "credentials": { "subdomain": "acme", "api_token": "..." },
    "case_field": "description",
    "outcome_field": "assignee_group",
    "filter": "status == \"closed\"",
    "refresh": "daily"
  }'

{
  "id": "osrc_2Km",
  "type": "zendesk",
  "status": "connecting",
  "case_field": "description",
  "outcome_field": "assignee_group",
  "refresh": "daily",
  "file_id": null,
  "cases": null,
  "created_at": "2026-09-08T09:04:00Z"
}

List outcome sources

curl "https://api.nucleus-ai.io/v1/outcome-sources?status=ready" \
  -H "Authorization: Bearer $NUCLEUS_API_KEY"

{
  "outcome_sources": [
    {
      "id": "osrc_2Km",
      "type": "zendesk",
      "status": "ready",
      "cases": 1200,
      "refresh": "daily",
      "refreshed_at": "2026-09-08T04:00:00Z"
    }
  ]
}

Retrieve an outcome source

curl https://api.nucleus-ai.io/v1/outcome-sources/osrc_2Km \
  -H "Authorization: Bearer $NUCLEUS_API_KEY"

{
  "id": "osrc_2Km",
  "type": "zendesk",
  "status": "ready",
  "case_field": "description",
  "outcome_field": "assignee_group",
  "filter": "status == \"closed\"",
  "refresh": "daily",
  "file_id": "file_4kEv1",
  "cases": 1200,
  "refreshed_at": "2026-09-08T04:00:00Z",
  "created_at": "2026-09-08T09:04:00Z"
}

Update an outcome source

curl -X PATCH https://api.nucleus-ai.io/v1/outcome-sources/osrc_2Km \
  -H "Authorization: Bearer $NUCLEUS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "refresh": "hourly" }'

{
  "id": "osrc_2Km",
  "type": "zendesk",
  "status": "ready",
  "refresh": "hourly",
  "refreshed_at": "2026-09-08T04:00:00Z"
}

Refresh an outcome source

curl -X POST https://api.nucleus-ai.io/v1/outcome-sources/osrc_2Km/refresh \
  -H "Authorization: Bearer $NUCLEUS_API_KEY"

{
  "id": "osrf_8Vt",
  "outcome_source": "osrc_2Km",
  "status": "running",
  "started_at": "2026-09-08T10:15:00Z"
}

Disconnect an outcome source

curl -X DELETE https://api.nucleus-ai.io/v1/outcome-sources/osrc_2Km \
  -H "Authorization: Bearer $NUCLEUS_API_KEY"

{
  "id": "osrc_2Km",
  "deleted": true
}