[Bug]: not able to retrieve keyword based search #17074
Description
Bug Description
from langchain_experimental.graph_transformers import LLMGraphTransformer
llm_transformer=LLMGraphTransformer(llm=llm,
node_properties=True,
relationship_properties=True)
graph_documents=llm_transformer.convert_to_graph_documents(docs[:10])
graph.add_graph_documents(
graph_documents,
baseEntityLabel=True,
include_source=True
)
Initialize embeddings with debug information
print("Initializing SentenceTransformerEmbeddings...")
embeddings = SentenceTransformerEmbeddings(model_name="intfloat/e5-small-v2")
print(f"Embeddings model initialized: {embeddings.model_name}")
vector_index = Neo4jVector.from_existing_graph(
embedding=embeddings,
url=NEO4J_URI,
username=NEO4J_USERNAME,
password=NEO4J_PASSWORD,
database="neo4j",
node_label="Document", # Adjust node_label as needed
text_node_properties=["id","text"],
embedding_node_property="embedding",
index_name="vector_index",
# keyword_index_name="entity_index",
search_type="hybrid",
pre_delete_collection=True
)
print("Neo4j vector index created successfully.")
def structured_retriever(question: str) -> str:
print(f"Received question: {question}")
result = ""
# Step 1: Retrieve entities from the entity chain
# print("Invoking entity chain to extract entities...")
entities = entity_chain.invoke({"question": question})
print(f"Extracted entities: {entities.names}")
# Step 2: Iterate through entities and query the graph
for entity1 in entities.names:
# print(f"\nProcessing entity: {entity}")
# Generate the full-text query for the current entity
full_text_query = generate_full_text_query(entity1)
print(f"Generated full text query for entity '{entity1}': {full_text_query}")
# Query the graph
print(f"Querying the graph for entity '{entity1}'...")
response = graph.query(
"""CALL db.index.fulltext.queryNodes('entity_index', $query, {limit:2})
YIELD node,score
CALL {
WITH node
MATCH (node)-[r:!MENTIONS]->(neighbor)
RETURN node.id + ' - ' + type(r) + ' -> ' + neighbor.id AS output
UNION ALL
WITH node
MATCH (node)<-[r:!MENTIONS]-(neighbor)
RETURN neighbor.id + ' - ' + type(r) + ' -> ' + node.id AS output
}
RETURN output LIMIT 50
""",
{"query": full_text_query},
)
print(f"Graph query response for entity '{entity1}': {response}")
# Append the outputs from the query to the result
outputs = [el['output'] for el in response]
# print(f"Outputs for entity '{entity}': {outputs}")
result += "\n".join(outputs)
# Final result
print(f"\nFinal result:\n{result}")
return result
please check in above code I am successfully taking response from similarity search but when trying to get output from keysearch that time its not able to give answer
but when I am running below query then it is able to giving answer please sujjest
graph.query(
"CREATE FULLTEXT INDEX entity_index IF NOT EXISTS FOR (n:__Entity__) ON EACH [n.id, n.text]"
)
thank you
Version
langchain
Steps to Reproduce
from langchain_experimental.graph_transformers import LLMGraphTransformer
llm_transformer=LLMGraphTransformer(llm=llm,
node_properties=True,
relationship_properties=True)
graph_documents=llm_transformer.convert_to_graph_documents(docs[:10])
graph.add_graph_documents(
graph_documents,
baseEntityLabel=True,
include_source=True
)
Initialize embeddings with debug information
print("Initializing SentenceTransformerEmbeddings...")
embeddings = SentenceTransformerEmbeddings(model_name="intfloat/e5-small-v2")
print(f"Embeddings model initialized: {embeddings.model_name}")
vector_index = Neo4jVector.from_existing_graph(
embedding=embeddings,
url=NEO4J_URI,
username=NEO4J_USERNAME,
password=NEO4J_PASSWORD,
database="neo4j",
node_label="Document", # Adjust node_label as needed
text_node_properties=["id","text"],
embedding_node_property="embedding",
index_name="vector_index",
# keyword_index_name="entity_index",
search_type="hybrid",
pre_delete_collection=True
)
print("Neo4j vector index created successfully.")
def structured_retriever(question: str) -> str:
print(f"Received question: {question}")
result = ""
# Step 1: Retrieve entities from the entity chain
# print("Invoking entity chain to extract entities...")
entities = entity_chain.invoke({"question": question})
print(f"Extracted entities: {entities.names}")
# Step 2: Iterate through entities and query the graph
for entity1 in entities.names:
# print(f"\nProcessing entity: {entity}")
# Generate the full-text query for the current entity
full_text_query = generate_full_text_query(entity1)
print(f"Generated full text query for entity '{entity1}': {full_text_query}")
# Query the graph
print(f"Querying the graph for entity '{entity1}'...")
response = graph.query(
"""CALL db.index.fulltext.queryNodes('entity_index', $query, {limit:2})
YIELD node,score
CALL {
WITH node
MATCH (node)-[r:!MENTIONS]->(neighbor)
RETURN node.id + ' - ' + type(r) + ' -> ' + neighbor.id AS output
UNION ALL
WITH node
MATCH (node)<-[r:!MENTIONS]-(neighbor)
RETURN neighbor.id + ' - ' + type(r) + ' -> ' + node.id AS output
}
RETURN output LIMIT 50
""",
{"query": full_text_query},
)
print(f"Graph query response for entity '{entity1}': {response}")
# Append the outputs from the query to the result
outputs = [el['output'] for el in response]
# print(f"Outputs for entity '{entity}': {outputs}")
result += "\n".join(outputs)
# Final result
print(f"\nFinal result:\n{result}")
return result
please check in above code I am successfully taking response from similarity search but when trying to get output from keysearch that time its not able to give answer
but when I am running below query then it is able to giving answer please sujjest
graph.query(
"CREATE FULLTEXT INDEX entity_index IF NOT EXISTS FOR (n:__Entity__) ON EACH [n.id, n.text]"
)
thank you
Relevant Logs/Tracbacks
empty results keyword search