graphrag/tests/verbs/test_extract_graph.py

# Copyright (c) 2024 Microsoft Corporation.
# Licensed under the MIT License

from graphrag.config.enums import ModelType
from graphrag.config.models.graph_rag_config import GraphRagConfig
from graphrag.index.workflows.extract_graph import (
    run_workflow,
)
from graphrag.utils.storage import load_table_from_storage

from .util import (
    DEFAULT_CHAT_MODEL_CONFIG,
    DEFAULT_EMBEDDING_MODEL_CONFIG,
    create_test_context,
)

MOCK_LLM_ENTITY_RESPONSES = [
    """
    ("entity"<|>COMPANY_A<|>COMPANY<|>Company_A is a test company)
    ##
    ("entity"<|>COMPANY_B<|>COMPANY<|>Company_B owns Company_A and also shares an address with Company_A)
    ##
    ("entity"<|>PERSON_C<|>PERSON<|>Person_C is director of Company_A)
    ##
    ("relationship"<|>COMPANY_A<|>COMPANY_B<|>Company_A and Company_B are related because Company_A is 100% owned by Company_B and the two companies also share the same address)<|>2)
    ##
    ("relationship"<|>COMPANY_A<|>PERSON_C<|>Company_A and Person_C are related because Person_C is director of Company_A<|>1))
    """.strip()
]

MOCK_LLM_SUMMARIZATION_RESPONSES = [
    """
    This is a MOCK response for the LLM. It is summarized!
    """.strip()
]


async def test_extract_graph():
    context = await create_test_context(
        storage=["text_units"],
    )

    extraction_model = DEFAULT_CHAT_MODEL_CONFIG.copy()
    extraction_model["type"] = ModelType.MockChat
    extraction_model["responses"] = MOCK_LLM_ENTITY_RESPONSES  # type: ignore
    config = GraphRagConfig(
        models={
            "default_chat_model": extraction_model,
            "default_embedding_model": DEFAULT_EMBEDDING_MODEL_CONFIG,
        }  # type: ignore
    )

    summarize_llm_settings = config.get_language_model_config(
        config.summarize_descriptions.model_id
    ).model_dump()
    summarize_llm_settings["type"] = ModelType.MockChat
    summarize_llm_settings["responses"] = MOCK_LLM_SUMMARIZATION_RESPONSES
    config.summarize_descriptions.max_input_tokens = 1000
    config.summarize_descriptions.max_length = 100

    await run_workflow(config, context)

    nodes_actual = await load_table_from_storage("entities", context.output_storage)
    edges_actual = await load_table_from_storage(
        "relationships", context.output_storage
    )

    assert len(nodes_actual.columns) == 5
    assert len(edges_actual.columns) == 5

    # TODO: with the combined verb we can't force summarization
    # this is because the mock responses always result in a single description, which is returned verbatim rather than summarized
    # we need to update the mocking to provide somewhat unique graphs so a true merge happens
    # the assertion should grab a node and ensure the description matches the mock description, not the original as we are doing below
    assert nodes_actual["description"].to_numpy()[0] == "Company_A is a test company"