Lineage and graph APIs

Use SQLINK_BASE_URL=https://<host>/sqlink for lineage and GRAPH_BASE_URL=https://<host>/graph for cross-object graphs. Lineage explains origins, transformations, and dependencies. A graph can also contain standards or asset associations; not every edge represents computational lineage.

SQLink queries

Method and complete path	Inputs	Result
POST /sqlink/item/searchItems	Body itemName, folderId, String modelId, String typeId, currentPage, pageSize	content, totalItems
POST /sqlink/lineageGraph/query	Body String id, queryType, direction, depth, hasTemp	nodes, relationships
POST /sqlink/lineageGraph/getTransform	Query Boolean isGraph, default true; graph conditions body	startIds, endId, transforms, lineageId, etc.
POST /sqlink/lineageGraph/getImpact	Body id, direction, depth, resourceTypes	Impact objects with identity, type, namespace, direction

Map source system, source, schema, table/field, and SQLink node ID explicitly. Numeric metadata IDs do not substitute for node IDs.

Assess a field change before release

1. Locate the corresponding SQLink node and confirm recent harvesting and parsing.
2. Query downstream with FORWARD, explicit depth, and type constraints.
3. Read transformations and verify expressions and connecting nodes.
4. Generate an impact list for business owner review.
5. Reharvest, parse, and verify key paths after the change.

Example impact body with a placeholder ID:

{
  "id": "example-sqlink-node-id",
  "queryType": "COLUMN",
  "direction": "FORWARD",
  "depth": 3,
  "resourceTypes": ["TABLE", "COLUMN", "REPORT"],
  "hasTemp": false
}

Common direction values are FORWARD, BACKWARD, and ALL; types can include SCHEMA, TABLE, COLUMN, REPORT, REPORT_COLUMN, and FILE. Actual enums may be narrower. Confirm the deployed contract, especially differences from the legacy SQLink reference.

Missing downstream results may reflect incomplete harvesting, parsing, depth, filters, or permissions rather than no impact.

Optional graph and unified asset access

Confirm these deployment-dependent capabilities. Path v1/v2 labels are API names, not product release claims. If unavailable, use the appropriate business service without assuming automatic fallback.

Method and complete path	Inputs	Result
POST /graph/graph/v2/nodes/search	keyword, typeId, page, size	content, total; confirm page origin
POST /graph/graph/v2/relations/query	typeId, String objectId, sourceSystem, layers, scope	nodes, relations, lastUpdateTime, truncated
POST /graph/asset-access/v1/discovery/search	query, typeIds, limit, cursor, scope	items, nextCursor, hasMore
POST /graph/asset-access/v1/context/resolve	Array assetKeys, profile	results; inspect each item

Pass the complete returned assetKey unchanged:

{
  "assetKeys": [{
    "sourceSystem": "<sourceSystem from search result>",
    "typeId": 123,
    "objectId": "<objectId from search result>"
  }]
}

Replace all placeholders, including type ID. Valid profile, scope, and scope types are not fully enumerated; obtain confirmed values. truncated=true means an incomplete relationship set. Follow nextCursor rather than constructing tokens. Retain available sourceVersion and sourceUpdatedAt for freshness.

DIG boundaries

Reading graphs is separate from publishing a DIG space MCP. Use context publishing to expose a governed, accepted business scope. Direct graph queries do not grant space documents, approvals, or knowledge access. The source guide checked definitions without parsing, modifying graphs, rebuilding projections, or publishing; validate business output with known deployment samples.

Parser plugin call chain

New job formats use SQLink Parser SPI plugins rather than ordinary REST controllers. REST handles upload, parsing, and queries; the plugin produces LineageContainer, steps, and relationships.

Method and complete path	Purpose
POST /sqlink/lineage/getLineageParseEngines	Installed typeName, descriptions, plugin information
POST /sqlink/lineage/uploadLineageFile	Form lineageEngine, objectType, file, folderId, defaultSchema, parse
POST /sqlink/lineage/parseLineage	lineageIds, status, onlyFailed, incrementalParsing
POST /sqlink/task/getTaskParseDetail	Progress, errors, missing tables
POST /sqlink/lineage/getLineageTableByPage	Parsed tables and state
POST /sqlink/lineage/getLineage	File and parsing details
POST /sqlink/lineageGraph/getTransform	Transformation logic
POST /sqlink/lineageGraph/getImpact	Impact

Plugin	Manifest	Input
WhaleTunnel	type=WHALETUNNEL, loadClass=com.datablau.sqlink.plugin.seatunnel.WhaleTunnelLineageParser, build=1.0.0, .json	whaletunnel.job_config, exportData source/transform/sink; datasource_name, source_table_name, result_table_name
Apache SeaTunnel	type=SEATUNNEL, loadClass=com.datablau.sqlink.plugin.seatunnel.SeatunnelLineageParser, build=1.0.0, .json	seatunnel.job_config, plugin_name, input/output, table_path, SQL, CTEs, aliases, transformations

Depend on SPI and standard output objects rather than copying JAR implementation. Report unsupported tasks clearly instead of creating misleading partial lineage. Supply default schema explicitly or use the supported DEFAULT_SCHEMA.

Query a complete table path

POST /sqlink/llm/getLineageOfTables requires namespace, schema, and table:

{
  "namespace": "<catalog-or-source>",
  "schema": "sales",
  "table": "customer",
  "direction": "ALL"
}

Use supported FORWARD, BACKWARD, or ALL. Retain full paths, node types, and source systems. A table name alone is ambiguous across sources and schemas.