mirror of
https://github.com/microsoft/graphrag.git
synced 2025-11-14 00:55:00 +00:00
c02ab0984a
* Remove create_final_nodes * Rename final entity output to "entities" * Remove duplicate code from graph extraction * Rename create_final_relationships output to "relationships" * Rename create_final_communities output to "communities" * Combine compute_communities and create_final_communities * Rename create_final_covariates output to "covariates" * Rename create_final_community_reports output to "community_reports" * Rename create_final_text_units output to "text_units" * Rename create_final_documents output to "documents" * Remove transient snapshots config * Move create_final_entities to finalize_entities operation * Move create_final_relationships flow to finalize_relationships operation * Reuse some community report functions * Collapse most of graph and text unit-based report generation * Unify schemas files * Move community reports extractor * Move NLP report prompt to prompts folder * Fix a few pandas warnings * Rename embeddings config to embed_text * Rename claim_extraction config to extract_claims * Remove nltk from standard graph extraction * Fix verb tests * Fix extract graph config naming * Fix moved file reference * Create v1-to-v2 migration notebook * Semver * Fix smoke test artifact count * Raise tpm/rpm on smoke tests * Update drift settings for smoke tests * Reuse project directory var in api notebook * Format * Format
82 lines
2.4 KiB
Python
82 lines
2.4 KiB
Python
# Copyright (c) 2024 Microsoft Corporation.
|
|
# Licensed under the MIT License
|
|
|
|
|
|
from graphrag.callbacks.noop_workflow_callbacks import NoopWorkflowCallbacks
|
|
from graphrag.config.create_graphrag_config import create_graphrag_config
|
|
from graphrag.config.enums import LLMType
|
|
from graphrag.index.operations.summarize_communities.community_reports_extractor import (
|
|
CommunityReportResponse,
|
|
FindingModel,
|
|
)
|
|
from graphrag.index.workflows.create_community_reports import (
|
|
run_workflow,
|
|
)
|
|
from graphrag.utils.storage import load_table_from_storage
|
|
|
|
from .util import (
|
|
DEFAULT_MODEL_CONFIG,
|
|
compare_outputs,
|
|
create_test_context,
|
|
load_test_table,
|
|
)
|
|
|
|
MOCK_RESPONSES = [
|
|
CommunityReportResponse(
|
|
title="<report_title>",
|
|
summary="<executive_summary>",
|
|
rating=2,
|
|
rating_explanation="<rating_explanation>",
|
|
findings=[
|
|
FindingModel(
|
|
summary="<insight_1_summary>", explanation="<insight_1_explanation"
|
|
),
|
|
FindingModel(
|
|
summary="<insight_2_summary>", explanation="<insight_2_explanation"
|
|
),
|
|
],
|
|
)
|
|
]
|
|
|
|
|
|
async def test_create_community_reports():
|
|
expected = load_test_table("community_reports")
|
|
|
|
context = await create_test_context(
|
|
storage=[
|
|
"covariates",
|
|
"relationships",
|
|
"entities",
|
|
"communities",
|
|
]
|
|
)
|
|
|
|
config = create_graphrag_config({"models": DEFAULT_MODEL_CONFIG})
|
|
llm_settings = config.get_language_model_config(
|
|
config.community_reports.model_id
|
|
).model_dump()
|
|
llm_settings["type"] = LLMType.StaticResponse
|
|
llm_settings["responses"] = MOCK_RESPONSES
|
|
llm_settings["parse_json"] = True
|
|
config.community_reports.strategy = {
|
|
"type": "graph_intelligence",
|
|
"llm": llm_settings,
|
|
}
|
|
|
|
await run_workflow(
|
|
config,
|
|
context,
|
|
NoopWorkflowCallbacks(),
|
|
)
|
|
|
|
actual = await load_table_from_storage("community_reports", context.storage)
|
|
|
|
assert len(actual.columns) == len(expected.columns)
|
|
|
|
# only assert a couple of columns that are not mock - most of this table is LLM-generated
|
|
compare_outputs(actual, expected, columns=["community", "level"])
|
|
|
|
# assert a handful of mock data items to confirm they get put in the right spot
|
|
assert actual["rank"][:1][0] == 2
|
|
assert actual["rank_explanation"][:1][0] == "<rating_explanation>"
|