Skip to main content

AddIndividualPipeline

What it is​

A pipeline that adds a named individual (instance) to an RDF triple store, or returns an existing matching individual if one is already present (based on a search workflow and score threshold).

Public API​

  • AddIndividualPipelineConfiguration(PipelineConfiguration) (dataclass)

    • Purpose: Provide dependencies for the pipeline.
    • Fields:
      • triple_store: ITripleStoreService - triple store service used to insert and fetch graphs.
      • search_individual_configuration: SearchIndividualWorkflowConfiguration - configuration passed to SearchIndividualWorkflow.

  • AddIndividualPipelineParameters(PipelineParameters) (Pydantic model)

    • Purpose: Input parameters for adding/searching an individual.
    • Fields:
      • individual_label: str - label for the individual (stored as rdfs:label).
      • class_uri: str - class URI the individual will be typed as.
      • threshold: Optional[int] = 80 - score threshold (0–100) to accept an existing individual from search results.

  • AddIndividualPipeline(Pipeline)

    • __init__(configuration: AddIndividualPipelineConfiguration)
      • Purpose: Initialize pipeline and internal SearchIndividualWorkflow.
    • run(parameters: PipelineParameters) -> rdflib.Graph
      • Purpose:
        • Searches for an existing individual matching individual_label within class_uri.
        • If a match is found with score >= threshold, returns the subject graph from the triple store.
        • Otherwise, creates a new individual with a UUID-based ABI URI, inserts it into the triple store, and returns the created graph.
      • Raises:
        • ValueError if parameters is not AddIndividualPipelineParameters.
    • as_tools() -> list[langchain_core.tools.BaseTool]
      • Purpose: Exposes the pipeline as LangChain StructuredTools:
        • add_individual_to_triple_store (generic: requires class_uri and individual_label)
        • Convenience tools that pre-fill class_uri:
          • add_commercial_organization
          • add_person
          • add_website
          • add_skill
          • add_legal_name
          • add_ticker_symbol
          • add_linkedin_page
    • as_api(...) -> None
      • Purpose: Present but not implemented (always returns None and does not register routes).

Configuration/Dependencies​

  • Requires an ITripleStoreService implementation providing:
    • insert(graph: rdflib.Graph) -> ...
    • get_subject_graph(subject_uri: str) -> rdflib.Graph
  • Uses SearchIndividualWorkflow from naas_abi.workflows.SearchIndividualWorkflow:
    • Called via search_individual(...) with class_uri and search_label.
    • Expects results containing score and individual_uri.
  • RDF libraries/terms:
    • Builds an rdflib.Graph, binds namespaces (bfo, cco, abi, dcterms), and adds triples:
      • (individual_uri, rdf:type, owl:NamedIndividual)
      • (individual_uri, rdf:type, <class_uri>)
      • (individual_uri, rdfs:label, "label")
  • Namespaces used:
    • ABI = "http://ontology.naas.ai/abi/"
    • CCO = "https://www.commoncoreontologies.org/"
    • BFO = "http://purl.obolibrary.org/obo/"

Usage​

from naas_abi.pipelines.AddIndividualPipeline import (
    AddIndividualPipeline,
    AddIndividualPipelineConfiguration,
    AddIndividualPipelineParameters,
)
from naas_abi.workflows.SearchIndividualWorkflow import SearchIndividualWorkflowConfiguration

# Provide concrete implementations/configuration from your environment:
triple_store = ...  # must implement ITripleStoreService
search_cfg = SearchIndividualWorkflowConfiguration(...)

pipeline = AddIndividualPipeline(
    AddIndividualPipelineConfiguration(
        triple_store=triple_store,
        search_individual_configuration=search_cfg,
    )
)

g = pipeline.run(
    AddIndividualPipelineParameters(
        individual_label="Naas.ai",
        class_uri="https://www.commoncoreontologies.org/ont00000443",
        threshold=80,
    )
)

print(len(g), "triples")

Using the LangChain tools:

tools = pipeline.as_tools()
add_person = next(t for t in tools if t.name == "add_person")
result_graph = add_person.run({"individual_label": "Ada Lovelace"})

Caveats​

  • Search filtering only runs when threshold is not None. If threshold=None, no search result will be accepted and a new individual will always be created.
  • When multiple matches exceed the threshold, the first result is used.
  • as_api is a no-op (no routes are exposed).