API Reference

`datameshops_core.mcp_client.MCPClientManager`

MCPClientManager(
    gms_url: str = None,        # Defaults to DATAHUB_GMS_URL env var
    gms_token: str = None,      # Defaults to DATAHUB_GMS_TOKEN env var
    max_retries: int = 3,
    backoff_factor: float = 1.5,
)

Methods

Method	Returns	Description
`execute_graphql(query, variables)`	`Dict`	Executes GraphQL against GMS with backoff
`call_mcp_tool(tool_name, arguments)`	`Dict`	Routes to appropriate DataHub MCP handler

Supported MCP Tools

search, get_entities, get_lineage, list_schema_fields, get_dataset_queries, add_tags, add_structured_properties, update_description

`datameshops_core.graph_traverser.GraphTraverser`

GraphTraverser(mcp_client: MCPClientManager)

Method	Returns	Description
`trace_upstream_lineage(urn, max_hops=3)`	`Dict`	Multi-hop BFS lineage traversal
`inspect_schema_fields(urn)`	`List[Dict]`	Returns all schema fields for entity
`get_historical_queries(urn)`	`List[Dict]`	Returns historic SQL queries
`search_catalog(query)`	`List[Dict]`	Full-text catalog search

`datameshops_core.mutation_engine.MutationEngine`

MutationEngine(mcp_client: MCPClientManager)

Method	Returns	Description
`apply_remediation_mutations(urn, agent_id, root_cause, sql_fix)`	`Dict`	Healer atomic mutation sequence
`apply_quarantine_mutations(model_urn, agent_id, health_score, leakage_field, drift_detail)`	`Dict`	Guard quarantine mutation sequence

`datameshops_core.batch_mutator.DataHubBatchMutatorSkill`

Open-source skill contribution for datahub-project/datahub-skills.

DataHubBatchMutatorSkill(client: DataHubClient)

Method	Returns	Description
`execute_batch_mutation(mutations: List[MutationPayload])`	`Dict`	Batch mutation with per-payload error isolation

`MutationPayload`

@dataclass
class MutationPayload:
    urn: str
    tags_to_add: Optional[List[str]] = None
    tags_to_remove: Optional[List[str]] = None
    structured_properties: Optional[Dict[str, Any]] = None
    markdown_description: Optional[str] = None

`datameshops_core.llm_adapter.LLMAdapter`

LLMAdapter(
    provider: str = None,   # "openai" | "gemini" | "anthropic" (env: LLM_PROVIDER)
    model: str = None,      # env: LLM_MODEL
)

Method	Returns	Description
`generate_remediation_sql(dataset_urn, schema_fields, historical_queries, error_context)`	`str`	SQL remediation patch
`generate_leakage_audit(model_urn, feature_schemas, upstream_lineage)`	`Dict`	Leakage & drift audit report