Ingest lineage

Use OpenLineage for runtime producers and the artifact endpoints for systems that already emit a complete project or catalog snapshot.

HTTP receiver

Both routes require a bearer token with write scope and workspace editor access. Requests are limited to 8 MiB. Batches accept 1 to 1,000 events. Compressed requests are not accepted.

A workspace-scoped API token can omit the workspace_id query parameter. This lets standard OpenLineage emitters that only support a base URL plus API key target Embrasure directly: set the emitter's URL to https://api.embrasure.ai (the client appends /api/v1/lineage) and its API key to a workspace-scoped token. Without connector_id, dataset identity resolves through the provider's namespace binding rules.

curl --request POST \
  --url "https://api.embrasure.ai/api/v1/lineage?workspace_id=$WORKSPACE_ID&connector_id=$CONNECTOR_ID" \
  --header "Authorization: Bearer $EMBRASURE_API_TOKEN" \
  --header "Content-Type: application/json" \
  --header "Idempotency-Key: orders-refresh-0190" \
  --data-binary @openlineage-event.json

Accepted delivery returns a durable receipt:

{
  "delivery_id": "c5ecf66d-1ca4-4e86-ae60-51c9108490cf",
  "state": "accepted",
  "duplicate": false,
  "item_count": 1,
  "invalid_item_count": 0
}

Processing is asynchronous. A receipt confirms durable acceptance, not completed graph projection.

Service-to-table reads

Send an input-only OpenLineage event to trace an application or service that reads a table. Use a stable, per-service provider_key and include job and inputs without outputs.

curl -X POST "$EMBRASURE_API/api/v1/lineage?workspace_id=$WORKSPACE_ID&connector_id=$CONNECTOR_ID&provider_key=risk-api" \
  -H "Authorization: Bearer $EMBRASURE_PAT" \
  -H "Content-Type: application/json" \
  -d '{"eventType":"COMPLETE","eventTime":"2026-08-04T12:00:00Z","job":{"namespace":"production","name":"risk-api"},"run":{"runId":"8a3c"},"inputs":[{"namespace":"warehouse","name":"analytics.transactions"}]}'

Embrasure creates the service identity from job.namespace and job.name, then records a reads_from edge for each resolved input. The service appears downstream when you inspect a table it reads.

Query-history discovery creates the same service nodes for service accounts. BigQuery uses referenced_tables from INFORMATION_SCHEMA.JOBS_BY_PROJECT; Snowflake and Databricks use collected query observations. Human users do not create service-read edges. Reads are aggregated into one evidence window per service and table each hour.

This route accepts enabled, user-bound dpt_ personal access tokens.

Idempotency

Send Idempotency-Key on retries. Reusing a key with different bytes returns a conflict. Without a key, byte-identical deliveries deduplicate by payload digest.

MCP submission

submit_lineage uses the same delivery contract as HTTP:

{
  "workspace_id": "workspace-id",
  "connector_id": "connector-id",
  "provider_key": "application:risk-api",
  "events": [{"eventType": "COMPLETE"}],
  "idempotency_key": "risk-api-run-0190"
}

events accepts 1 to 1,000 OpenLineage objects. The result contains delivery_id, state, duplicate, item_count, and invalid_item_count. The tool requires write scope and is idempotent but not read-only.

Artifact imports

Every artifact request includes workspace_id, connector_id, completeness, and an idempotency key for complete acquisitions.