Local Search

In [1]:

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

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

In [2]:

import os

import pandas as pd
import tiktoken

from graphrag.query.context_builder.entity_extraction import EntityVectorStoreKey
from graphrag.query.indexer_adapters import (
    read_indexer_covariates,
    read_indexer_entities,
    read_indexer_relationships,
    read_indexer_reports,
    read_indexer_text_units,
)
from graphrag.query.question_gen.local_gen import LocalQuestionGen
from graphrag.query.structured_search.local_search.mixed_context import (
    LocalSearchMixedContext,
)
from graphrag.query.structured_search.local_search.search import LocalSearch
from graphrag.vector_stores.lancedb import LanceDBVectorStore

import os import pandas as pd import tiktoken from graphrag.query.context_builder.entity_extraction import EntityVectorStoreKey from graphrag.query.indexer_adapters import ( read_indexer_covariates, read_indexer_entities, read_indexer_relationships, read_indexer_reports, read_indexer_text_units, ) from graphrag.query.question_gen.local_gen import LocalQuestionGen from graphrag.query.structured_search.local_search.mixed_context import ( LocalSearchMixedContext, ) from graphrag.query.structured_search.local_search.search import LocalSearch from graphrag.vector_stores.lancedb import LanceDBVectorStore

---------------------------------------------------------------------------
ModuleNotFoundError                       Traceback (most recent call last)
Cell In[2], line 3
      1 import os
----> 3 import pandas as pd
      4 import tiktoken
      6 from graphrag.query.context_builder.entity_extraction import EntityVectorStoreKey

ModuleNotFoundError: No module named 'pandas'

Local Search Example¶

Local search method generates answers by combining relevant data from the AI-extracted knowledge-graph with text chunks of the raw documents. This method is suitable for questions that require an understanding of specific entities mentioned in the documents (e.g. What are the healing properties of chamomile?).

Load text units and graph data tables as context for local search¶

• In this test we first load indexing outputs from parquet files to dataframes, then convert these dataframes into collections of data objects aligning with the knowledge model.

Load tables to dataframes¶

In [3]:

INPUT_DIR = "./inputs/operation dulce"
LANCEDB_URI = f"{INPUT_DIR}/lancedb"

COMMUNITY_REPORT_TABLE = "community_reports"
ENTITY_TABLE = "entities"
COMMUNITY_TABLE = "communities"
RELATIONSHIP_TABLE = "relationships"
COVARIATE_TABLE = "covariates"
TEXT_UNIT_TABLE = "text_units"
COMMUNITY_LEVEL = 2

INPUT_DIR = "./inputs/operation dulce" LANCEDB_URI = f"{INPUT_DIR}/lancedb" COMMUNITY_REPORT_TABLE = "community_reports" ENTITY_TABLE = "entities" COMMUNITY_TABLE = "communities" RELATIONSHIP_TABLE = "relationships" COVARIATE_TABLE = "covariates" TEXT_UNIT_TABLE = "text_units" COMMUNITY_LEVEL = 2

Read entities¶

In [4]:

# read nodes table to get community and degree data
entity_df = pd.read_parquet(f"{INPUT_DIR}/{ENTITY_TABLE}.parquet")
community_df = pd.read_parquet(f"{INPUT_DIR}/{COMMUNITY_TABLE}.parquet")

entities = read_indexer_entities(entity_df, community_df, COMMUNITY_LEVEL)

# load description embeddings to an in-memory lancedb vectorstore
# to connect to a remote db, specify url and port values.
description_embedding_store = LanceDBVectorStore(
    collection_name="default-entity-description",
)
description_embedding_store.connect(db_uri=LANCEDB_URI)

print(f"Entity count: {len(entity_df)}")
entity_df.head()

# read nodes table to get community and degree data entity_df = pd.read_parquet(f"{INPUT_DIR}/{ENTITY_TABLE}.parquet") community_df = pd.read_parquet(f"{INPUT_DIR}/{COMMUNITY_TABLE}.parquet") entities = read_indexer_entities(entity_df, community_df, COMMUNITY_LEVEL) # load description embeddings to an in-memory lancedb vectorstore # to connect to a remote db, specify url and port values. description_embedding_store = LanceDBVectorStore( collection_name="default-entity-description", ) description_embedding_store.connect(db_uri=LANCEDB_URI) print(f"Entity count: {len(entity_df)}") entity_df.head()

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
Cell In[4], line 2
      1 # read nodes table to get community and degree data
----> 2 entity_df = pd.read_parquet(f"{INPUT_DIR}/{ENTITY_TABLE}.parquet")
      3 community_df = pd.read_parquet(f"{INPUT_DIR}/{COMMUNITY_TABLE}.parquet")
      5 entities = read_indexer_entities(entity_df, community_df, COMMUNITY_LEVEL)

NameError: name 'pd' is not defined

Read relationships¶

In [5]:

relationship_df = pd.read_parquet(f"{INPUT_DIR}/{RELATIONSHIP_TABLE}.parquet")
relationships = read_indexer_relationships(relationship_df)

print(f"Relationship count: {len(relationship_df)}")
relationship_df.head()

relationship_df = pd.read_parquet(f"{INPUT_DIR}/{RELATIONSHIP_TABLE}.parquet") relationships = read_indexer_relationships(relationship_df) print(f"Relationship count: {len(relationship_df)}") relationship_df.head()

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
Cell In[5], line 1
----> 1 relationship_df = pd.read_parquet(f"{INPUT_DIR}/{RELATIONSHIP_TABLE}.parquet")
      2 relationships = read_indexer_relationships(relationship_df)
      4 print(f"Relationship count: {len(relationship_df)}")

NameError: name 'pd' is not defined

In [6]:

# NOTE: covariates are turned off by default, because they generally need prompt tuning to be valuable
# Please see the GRAPHRAG_CLAIM_* settings
covariate_df = pd.read_parquet(f"{INPUT_DIR}/{COVARIATE_TABLE}.parquet")

claims = read_indexer_covariates(covariate_df)

print(f"Claim records: {len(claims)}")
covariates = {"claims": claims}

# NOTE: covariates are turned off by default, because they generally need prompt tuning to be valuable # Please see the GRAPHRAG_CLAIM_* settings covariate_df = pd.read_parquet(f"{INPUT_DIR}/{COVARIATE_TABLE}.parquet") claims = read_indexer_covariates(covariate_df) print(f"Claim records: {len(claims)}") covariates = {"claims": claims}

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
Cell In[6], line 3
      1 # NOTE: covariates are turned off by default, because they generally need prompt tuning to be valuable
      2 # Please see the GRAPHRAG_CLAIM_* settings
----> 3 covariate_df = pd.read_parquet(f"{INPUT_DIR}/{COVARIATE_TABLE}.parquet")
      5 claims = read_indexer_covariates(covariate_df)
      7 print(f"Claim records: {len(claims)}")

NameError: name 'pd' is not defined

Read community reports¶

In [7]:

report_df = pd.read_parquet(f"{INPUT_DIR}/{COMMUNITY_REPORT_TABLE}.parquet")
reports = read_indexer_reports(report_df, community_df, COMMUNITY_LEVEL)

print(f"Report records: {len(report_df)}")
report_df.head()

report_df = pd.read_parquet(f"{INPUT_DIR}/{COMMUNITY_REPORT_TABLE}.parquet") reports = read_indexer_reports(report_df, community_df, COMMUNITY_LEVEL) print(f"Report records: {len(report_df)}") report_df.head()

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
Cell In[7], line 1
----> 1 report_df = pd.read_parquet(f"{INPUT_DIR}/{COMMUNITY_REPORT_TABLE}.parquet")
      2 reports = read_indexer_reports(report_df, community_df, COMMUNITY_LEVEL)
      4 print(f"Report records: {len(report_df)}")

NameError: name 'pd' is not defined

Read text units¶

In [8]:

text_unit_df = pd.read_parquet(f"{INPUT_DIR}/{TEXT_UNIT_TABLE}.parquet")
text_units = read_indexer_text_units(text_unit_df)

print(f"Text unit records: {len(text_unit_df)}")
text_unit_df.head()

text_unit_df = pd.read_parquet(f"{INPUT_DIR}/{TEXT_UNIT_TABLE}.parquet") text_units = read_indexer_text_units(text_unit_df) print(f"Text unit records: {len(text_unit_df)}") text_unit_df.head()

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
Cell In[8], line 1
----> 1 text_unit_df = pd.read_parquet(f"{INPUT_DIR}/{TEXT_UNIT_TABLE}.parquet")
      2 text_units = read_indexer_text_units(text_unit_df)
      4 print(f"Text unit records: {len(text_unit_df)}")

NameError: name 'pd' is not defined

In [9]:

from graphrag.config.enums import ModelType
from graphrag.config.models.language_model_config import LanguageModelConfig
from graphrag.language_model.manager import ModelManager

api_key = os.environ["GRAPHRAG_API_KEY"]
llm_model = os.environ["GRAPHRAG_LLM_MODEL"]
embedding_model = os.environ["GRAPHRAG_EMBEDDING_MODEL"]

chat_config = LanguageModelConfig(
    api_key=api_key,
    type=ModelType.OpenAIChat,
    model=llm_model,
    max_retries=20,
)
chat_model = ModelManager().get_or_create_chat_model(
    name="local_search",
    model_type=ModelType.OpenAIChat,
    config=chat_config,
)

token_encoder = tiktoken.encoding_for_model(llm_model)

embedding_config = LanguageModelConfig(
    api_key=api_key,
    type=ModelType.OpenAIEmbedding,
    model=embedding_model,
    max_retries=20,
)

text_embedder = ModelManager().get_or_create_embedding_model(
    name="local_search_embedding",
    model_type=ModelType.OpenAIEmbedding,
    config=embedding_config,
)

from graphrag.config.enums import ModelType from graphrag.config.models.language_model_config import LanguageModelConfig from graphrag.language_model.manager import ModelManager api_key = os.environ["GRAPHRAG_API_KEY"] llm_model = os.environ["GRAPHRAG_LLM_MODEL"] embedding_model = os.environ["GRAPHRAG_EMBEDDING_MODEL"] chat_config = LanguageModelConfig( api_key=api_key, type=ModelType.OpenAIChat, model=llm_model, max_retries=20, ) chat_model = ModelManager().get_or_create_chat_model( name="local_search", model_type=ModelType.OpenAIChat, config=chat_config, ) token_encoder = tiktoken.encoding_for_model(llm_model) embedding_config = LanguageModelConfig( api_key=api_key, type=ModelType.OpenAIEmbedding, model=embedding_model, max_retries=20, ) text_embedder = ModelManager().get_or_create_embedding_model( name="local_search_embedding", model_type=ModelType.OpenAIEmbedding, config=embedding_config, )

---------------------------------------------------------------------------
ModuleNotFoundError                       Traceback (most recent call last)
Cell In[9], line 1
----> 1 from graphrag.config.enums import ModelType
      2 from graphrag.config.models.language_model_config import LanguageModelConfig
      3 from graphrag.language_model.manager import ModelManager

ModuleNotFoundError: No module named 'graphrag'

Create local search context builder¶

In [10]:

context_builder = LocalSearchMixedContext(
    community_reports=reports,
    text_units=text_units,
    entities=entities,
    relationships=relationships,
    # if you did not run covariates during indexing, set this to None
    covariates=covariates,
    entity_text_embeddings=description_embedding_store,
    embedding_vectorstore_key=EntityVectorStoreKey.ID,  # if the vectorstore uses entity title as ids, set this to EntityVectorStoreKey.TITLE
    text_embedder=text_embedder,
    token_encoder=token_encoder,
)

context_builder = LocalSearchMixedContext( community_reports=reports, text_units=text_units, entities=entities, relationships=relationships, # if you did not run covariates during indexing, set this to None covariates=covariates, entity_text_embeddings=description_embedding_store, embedding_vectorstore_key=EntityVectorStoreKey.ID, # if the vectorstore uses entity title as ids, set this to EntityVectorStoreKey.TITLE text_embedder=text_embedder, token_encoder=token_encoder, )

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
Cell In[10], line 1
----> 1 context_builder = LocalSearchMixedContext(
      2     community_reports=reports,
      3     text_units=text_units,
      4     entities=entities,
      5     relationships=relationships,
      6     # if you did not run covariates during indexing, set this to None
      7     covariates=covariates,
      8     entity_text_embeddings=description_embedding_store,
      9     embedding_vectorstore_key=EntityVectorStoreKey.ID,  # if the vectorstore uses entity title as ids, set this to EntityVectorStoreKey.TITLE
     10     text_embedder=text_embedder,
     11     token_encoder=token_encoder,
     12 )

NameError: name 'LocalSearchMixedContext' is not defined

Create local search engine¶

In [11]:

# text_unit_prop: proportion of context window dedicated to related text units
# community_prop: proportion of context window dedicated to community reports.
# The remaining proportion is dedicated to entities and relationships. Sum of text_unit_prop and community_prop should be <= 1
# conversation_history_max_turns: maximum number of turns to include in the conversation history.
# conversation_history_user_turns_only: if True, only include user queries in the conversation history.
# top_k_mapped_entities: number of related entities to retrieve from the entity description embedding store.
# top_k_relationships: control the number of out-of-network relationships to pull into the context window.
# include_entity_rank: if True, include the entity rank in the entity table in the context window. Default entity rank = node degree.
# include_relationship_weight: if True, include the relationship weight in the context window.
# include_community_rank: if True, include the community rank in the context window.
# return_candidate_context: if True, return a set of dataframes containing all candidate entity/relationship/covariate records that
# could be relevant. Note that not all of these records will be included in the context window. The "in_context" column in these
# dataframes indicates whether the record is included in the context window.
# max_tokens: maximum number of tokens to use for the context window.


local_context_params = {
    "text_unit_prop": 0.5,
    "community_prop": 0.1,
    "conversation_history_max_turns": 5,
    "conversation_history_user_turns_only": True,
    "top_k_mapped_entities": 10,
    "top_k_relationships": 10,
    "include_entity_rank": True,
    "include_relationship_weight": True,
    "include_community_rank": False,
    "return_candidate_context": False,
    "embedding_vectorstore_key": EntityVectorStoreKey.ID,  # set this to EntityVectorStoreKey.TITLE if the vectorstore uses entity title as ids
    "max_tokens": 12_000,  # change this based on the token limit you have on your model (if you are using a model with 8k limit, a good setting could be 5000)
}

model_params = {
    "max_tokens": 2_000,  # change this based on the token limit you have on your model (if you are using a model with 8k limit, a good setting could be 1000=1500)
    "temperature": 0.0,
}

# text_unit_prop: proportion of context window dedicated to related text units # community_prop: proportion of context window dedicated to community reports. # The remaining proportion is dedicated to entities and relationships. Sum of text_unit_prop and community_prop should be <= 1 # conversation_history_max_turns: maximum number of turns to include in the conversation history. # conversation_history_user_turns_only: if True, only include user queries in the conversation history. # top_k_mapped_entities: number of related entities to retrieve from the entity description embedding store. # top_k_relationships: control the number of out-of-network relationships to pull into the context window. # include_entity_rank: if True, include the entity rank in the entity table in the context window. Default entity rank = node degree. # include_relationship_weight: if True, include the relationship weight in the context window. # include_community_rank: if True, include the community rank in the context window. # return_candidate_context: if True, return a set of dataframes containing all candidate entity/relationship/covariate records that # could be relevant. Note that not all of these records will be included in the context window. The "in_context" column in these # dataframes indicates whether the record is included in the context window. # max_tokens: maximum number of tokens to use for the context window. local_context_params = { "text_unit_prop": 0.5, "community_prop": 0.1, "conversation_history_max_turns": 5, "conversation_history_user_turns_only": True, "top_k_mapped_entities": 10, "top_k_relationships": 10, "include_entity_rank": True, "include_relationship_weight": True, "include_community_rank": False, "return_candidate_context": False, "embedding_vectorstore_key": EntityVectorStoreKey.ID, # set this to EntityVectorStoreKey.TITLE if the vectorstore uses entity title as ids "max_tokens": 12_000, # change this based on the token limit you have on your model (if you are using a model with 8k limit, a good setting could be 5000) } model_params = { "max_tokens": 2_000, # change this based on the token limit you have on your model (if you are using a model with 8k limit, a good setting could be 1000=1500) "temperature": 0.0, }

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
Cell In[11], line 28
      1 # text_unit_prop: proportion of context window dedicated to related text units
      2 # community_prop: proportion of context window dedicated to community reports.
      3 # The remaining proportion is dedicated to entities and relationships. Sum of text_unit_prop and community_prop should be <= 1
   (...)
     13 # dataframes indicates whether the record is included in the context window.
     14 # max_tokens: maximum number of tokens to use for the context window.
     17 local_context_params = {
     18     "text_unit_prop": 0.5,
     19     "community_prop": 0.1,
     20     "conversation_history_max_turns": 5,
     21     "conversation_history_user_turns_only": True,
     22     "top_k_mapped_entities": 10,
     23     "top_k_relationships": 10,
     24     "include_entity_rank": True,
     25     "include_relationship_weight": True,
     26     "include_community_rank": False,
     27     "return_candidate_context": False,
---> 28     "embedding_vectorstore_key": EntityVectorStoreKey.ID,  # set this to EntityVectorStoreKey.TITLE if the vectorstore uses entity title as ids
     29     "max_tokens": 12_000,  # change this based on the token limit you have on your model (if you are using a model with 8k limit, a good setting could be 5000)
     30 }
     32 model_params = {
     33     "max_tokens": 2_000,  # change this based on the token limit you have on your model (if you are using a model with 8k limit, a good setting could be 1000=1500)
     34     "temperature": 0.0,
     35 }

NameError: name 'EntityVectorStoreKey' is not defined

In [12]:

search_engine = LocalSearch(
    model=chat_model,
    context_builder=context_builder,
    token_encoder=token_encoder,
    model_params=model_params,
    context_builder_params=local_context_params,
    response_type="multiple paragraphs",  # free form text describing the response type and format, can be anything, e.g. prioritized list, single paragraph, multiple paragraphs, multiple-page report
)

search_engine = LocalSearch( model=chat_model, context_builder=context_builder, token_encoder=token_encoder, model_params=model_params, context_builder_params=local_context_params, response_type="multiple paragraphs", # free form text describing the response type and format, can be anything, e.g. prioritized list, single paragraph, multiple paragraphs, multiple-page report )

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
Cell In[12], line 1
----> 1 search_engine = LocalSearch(
      2     model=chat_model,
      3     context_builder=context_builder,
      4     token_encoder=token_encoder,
      5     model_params=model_params,
      6     context_builder_params=local_context_params,
      7     response_type="multiple paragraphs",  # free form text describing the response type and format, can be anything, e.g. prioritized list, single paragraph, multiple paragraphs, multiple-page report
      8 )

NameError: name 'LocalSearch' is not defined

Run local search on sample queries¶

In [13]:

result = await search_engine.search("Tell me about Agent Mercer")
print(result.response)

result = await search_engine.search("Tell me about Agent Mercer") print(result.response)

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
Cell In[13], line 1
----> 1 result = await search_engine.search("Tell me about Agent Mercer")
      2 print(result.response)

NameError: name 'search_engine' is not defined

In [14]:

question = "Tell me about Dr. Jordan Hayes"
result = await search_engine.search(question)
print(result.response)

question = "Tell me about Dr. Jordan Hayes" result = await search_engine.search(question) print(result.response)

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
Cell In[14], line 2
      1 question = "Tell me about Dr. Jordan Hayes"
----> 2 result = await search_engine.search(question)
      3 print(result.response)

NameError: name 'search_engine' is not defined

Inspecting the context data used to generate the response¶

In [15]:

result.context_data["entities"].head()

result.context_data["entities"].head()

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
Cell In[15], line 1
----> 1 result.context_data["entities"].head()

NameError: name 'result' is not defined

In [16]:

result.context_data["relationships"].head()

result.context_data["relationships"].head()

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
Cell In[16], line 1
----> 1 result.context_data["relationships"].head()

NameError: name 'result' is not defined

In [17]:

if "reports" in result.context_data:
    result.context_data["reports"].head()

if "reports" in result.context_data: result.context_data["reports"].head()

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
Cell In[17], line 1
----> 1 if "reports" in result.context_data:
      2     result.context_data["reports"].head()

NameError: name 'result' is not defined

In [18]:

result.context_data["sources"].head()

result.context_data["sources"].head()

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
Cell In[18], line 1
----> 1 result.context_data["sources"].head()

NameError: name 'result' is not defined

In [19]:

if "claims" in result.context_data:
    print(result.context_data["claims"].head())

if "claims" in result.context_data: print(result.context_data["claims"].head())

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
Cell In[19], line 1
----> 1 if "claims" in result.context_data:
      2     print(result.context_data["claims"].head())

NameError: name 'result' is not defined

Question Generation¶

This function takes a list of user queries and generates the next candidate questions.

In [20]:

question_generator = LocalQuestionGen(
    model=chat_model,
    context_builder=context_builder,
    token_encoder=token_encoder,
    model_params=model_params,
    context_builder_params=local_context_params,
)

question_generator = LocalQuestionGen( model=chat_model, context_builder=context_builder, token_encoder=token_encoder, model_params=model_params, context_builder_params=local_context_params, )

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
Cell In[20], line 1
----> 1 question_generator = LocalQuestionGen(
      2     model=chat_model,
      3     context_builder=context_builder,
      4     token_encoder=token_encoder,
      5     model_params=model_params,
      6     context_builder_params=local_context_params,
      7 )

NameError: name 'LocalQuestionGen' is not defined

In [21]:

question_history = [
    "Tell me about Agent Mercer",
    "What happens in Dulce military base?",
]
candidate_questions = await question_generator.agenerate(
    question_history=question_history, context_data=None, question_count=5
)
print(candidate_questions.response)

question_history = [ "Tell me about Agent Mercer", "What happens in Dulce military base?", ] candidate_questions = await question_generator.agenerate( question_history=question_history, context_data=None, question_count=5 ) print(candidate_questions.response)

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
Cell In[21], line 5
      1 question_history = [
      2     "Tell me about Agent Mercer",
      3     "What happens in Dulce military base?",
      4 ]
----> 5 candidate_questions = await question_generator.agenerate(
      6     question_history=question_history, context_data=None, question_count=5
      7 )
      8 print(candidate_questions.response)

NameError: name 'question_generator' is not defined