mirror of
https://github.com/microsoft/graphrag.git
synced 2025-07-27 10:50:08 +00:00
61b3d6d56a
* Remove genid * Move snapshot_rows * Move snapshot * Delete spread_json * Delete unzip * Delete zip * Move unpack_graph * Move compute_edge_combined_degree * Delete create_graph * Delete concat * Delete text replace * Delete text_translate * Move text_split * Inline aggregate override * Move cluster_graph * Move merge_graphs * Semver * Move text_chunk * Move layout_graph and fix some __init__s * Move extract_covariates * Rename text_split -> split_text * Move extract_entities * Move summarize_descriptions * Rename text_chunk -> chunk_text * Move community report creation * Remove verb-level packing operators * Streamline some naming * Streamline param name/order * Move mock LLM data to tests * Fixed missed rename * Update some strategy refs * Rename run_gi * Inject mock responses into integ test config
130 lines
3.4 KiB
Python
130 lines
3.4 KiB
Python
# Copyright (c) 2024 Microsoft Corporation.
|
|
# Licensed under the MIT License
|
|
|
|
import networkx as nx
|
|
import pytest
|
|
|
|
from graphrag.config.enums import LLMType
|
|
from graphrag.index.storage.memory_pipeline_storage import MemoryPipelineStorage
|
|
from graphrag.index.workflows.v1.create_summarized_entities import (
|
|
build_steps,
|
|
workflow_name,
|
|
)
|
|
|
|
from .util import (
|
|
get_config_for_workflow,
|
|
get_workflow_output,
|
|
load_expected,
|
|
load_input_tables,
|
|
)
|
|
|
|
MOCK_LLM_RESPONSES = [
|
|
"""
|
|
This is a MOCK response for the LLM. It is summarized!
|
|
""".strip()
|
|
]
|
|
|
|
MOCK_LLM_CONFIG = {
|
|
"type": LLMType.StaticResponse,
|
|
"responses": MOCK_LLM_RESPONSES,
|
|
}
|
|
|
|
|
|
async def test_create_summarized_entities():
|
|
input_tables = load_input_tables([
|
|
"workflow:create_base_extracted_entities",
|
|
])
|
|
expected = load_expected(workflow_name)
|
|
|
|
storage = MemoryPipelineStorage()
|
|
|
|
config = get_config_for_workflow(workflow_name)
|
|
|
|
config["summarize_descriptions"]["strategy"]["llm"] = MOCK_LLM_CONFIG
|
|
|
|
steps = build_steps(config)
|
|
|
|
actual = await get_workflow_output(
|
|
input_tables,
|
|
{
|
|
"steps": steps,
|
|
},
|
|
storage=storage,
|
|
)
|
|
|
|
# the serialization of the graph may differ so we can't assert the dataframes directly
|
|
assert actual.shape == expected.shape, "Graph dataframe shapes differ"
|
|
|
|
# let's parse a sample of the raw graphml
|
|
actual_graphml_0 = actual["entity_graph"][:1][0]
|
|
actual_graph_0 = nx.parse_graphml(actual_graphml_0)
|
|
|
|
expected_graphml_0 = expected["entity_graph"][:1][0]
|
|
expected_graph_0 = nx.parse_graphml(expected_graphml_0)
|
|
|
|
assert (
|
|
actual_graph_0.number_of_nodes() == expected_graph_0.number_of_nodes()
|
|
), "Graphml node count differs"
|
|
assert (
|
|
actual_graph_0.number_of_edges() == expected_graph_0.number_of_edges()
|
|
), "Graphml edge count differs"
|
|
|
|
# ensure the mock summary was injected to the nodes
|
|
nodes = list(actual_graph_0.nodes(data=True))
|
|
assert (
|
|
nodes[0][1]["description"]
|
|
== "This is a MOCK response for the LLM. It is summarized!"
|
|
)
|
|
|
|
assert len(storage.keys()) == 0, "Storage should be empty"
|
|
|
|
|
|
async def test_create_summarized_entities_with_snapshots():
|
|
input_tables = load_input_tables([
|
|
"workflow:create_base_extracted_entities",
|
|
])
|
|
expected = load_expected(workflow_name)
|
|
|
|
storage = MemoryPipelineStorage()
|
|
|
|
config = get_config_for_workflow(workflow_name)
|
|
|
|
config["summarize_descriptions"]["strategy"]["llm"] = MOCK_LLM_CONFIG
|
|
config["graphml_snapshot"] = True
|
|
|
|
steps = build_steps(config)
|
|
|
|
actual = await get_workflow_output(
|
|
input_tables,
|
|
{
|
|
"steps": steps,
|
|
},
|
|
storage=storage,
|
|
)
|
|
|
|
assert actual.shape == expected.shape, "Graph dataframe shapes differ"
|
|
|
|
assert storage.keys() == [
|
|
"summarized_graph.graphml",
|
|
], "Graph snapshot keys differ"
|
|
|
|
|
|
async def test_create_summarized_entities_missing_llm_throws():
|
|
input_tables = load_input_tables([
|
|
"workflow:create_base_extracted_entities",
|
|
])
|
|
|
|
config = get_config_for_workflow(workflow_name)
|
|
|
|
del config["summarize_descriptions"]["strategy"]["llm"]
|
|
|
|
steps = build_steps(config)
|
|
|
|
with pytest.raises(ValueError): # noqa PT011
|
|
await get_workflow_output(
|
|
input_tables,
|
|
{
|
|
"steps": steps,
|
|
},
|
|
)
|