Neo4j

Neo4j is a graph database management system developed by Neo4j, Inc.

The data elements Neo4j stores are nodes, edges connecting them, and attributes of nodes and edges. Described by its developers as an ACID-compliant transactional database with native graph storage and processing, Neo4j is available in a non-open-source "community edition" licensed with a modification of the GNU General Public License, with online backup and high availability extensions licensed under a closed-source commercial license. Neo also licenses Neo4j with these extensions under closed-source commercial terms.

This notebook shows how to use LLMs to provide a natural language interface to a graph database you can query with the Cypher query language.

Cypher is a declarative graph query language that allows for expressive and efficient data querying in a property graph.

Setting up

You will need to have a running Neo4j instance. One option is to create a free Neo4j database instance in their Aura cloud service. You can also run the database locally using the Neo4j Desktop application, or running a docker container. You can run a local docker container by running the executing the following script:

docker run \
    --name neo4j \
    -p 7474:7474 -p 7687:7687 \
    -d \
    -e NEO4J_AUTH=neo4j/password \
    -e NEO4J_PLUGINS=\[\"apoc\"\]  \
    neo4j:latest

If you are using the docker container, you need to wait a couple of second for the database to start.

from langchain_neo4j import GraphCypherQAChain, Neo4jGraph
from langchain_openai import ChatOpenAI

API Reference:GraphCypherQAChain | Neo4jGraph | ChatOpenAI

graph = Neo4jGraph(url="bolt://localhost:7687", username="neo4j", password="password")

We default to OpenAI models in this guide.

import getpass
import os

if "OPENAI_API_KEY" not in os.environ:
    os.environ["OPENAI_API_KEY"] = getpass.getpass("OpenAI API Key:")

Seeding the database

Assuming your database is empty, you can populate it using Cypher query language. The following Cypher statement is idempotent, which means the database information will be the same if you run it one or multiple times.

graph.query(
    """
MERGE (m:Movie {name:"Top Gun", runtime: 120})
WITH m
UNWIND ["Tom Cruise", "Val Kilmer", "Anthony Edwards", "Meg Ryan"] AS actor
MERGE (a:Actor {name:actor})
MERGE (a)-[:ACTED_IN]->(m)
"""
)

[]

Refresh graph schema information

If the schema of database changes, you can refresh the schema information needed to generate Cypher statements.

graph.refresh_schema()

print(graph.schema)

Node properties:
Movie {runtime: INTEGER, name: STRING}
Actor {name: STRING}
Relationship properties:

The relationships:
(:Actor)-[:ACTED_IN]->(:Movie)

Enhanced schema information

Choosing the enhanced schema version enables the system to automatically scan for example values within the databases and calculate some distribution metrics. For example, if a node property has less than 10 distinct values, we return all possible values in the schema. Otherwise, return only a single example value per node and relationship property.

enhanced_graph = Neo4jGraph(
    url="bolt://localhost:7687",
    username="neo4j",
    password="password",
    enhanced_schema=True,
)
print(enhanced_graph.schema)

Node properties:
- **Movie**
  - `runtime`: INTEGER Min: 120, Max: 120
  - `name`: STRING Available options: ['Top Gun']
- **Actor**
  - `name`: STRING Available options: ['Tom Cruise', 'Val Kilmer', 'Anthony Edwards', 'Meg Ryan']
Relationship properties:

The relationships:
(:Actor)-[:ACTED_IN]->(:Movie)

Querying the graph

We can now use the graph cypher QA chain to ask question of the graph

chain = GraphCypherQAChain.from_llm(
    ChatOpenAI(temperature=0), graph=graph, verbose=True, allow_dangerous_requests=True
)

chain.invoke({"query": "Who played in Top Gun?"})

[1m> Entering new GraphCypherQAChain chain...[0m
Generated Cypher:
[32;1m[1;3mMATCH (a:Actor)-[:ACTED_IN]->(m:Movie)
WHERE m.name = 'Top Gun'
RETURN a.name[0m
Full Context:
[32;1m[1;3m[{'a.name': 'Tom Cruise'}, {'a.name': 'Val Kilmer'}, {'a.name': 'Anthony Edwards'}, {'a.name': 'Meg Ryan'}][0m

[1m> Finished chain.[0m

{'query': 'Who played in Top Gun?',
 'result': 'Tom Cruise, Val Kilmer, Anthony Edwards, and Meg Ryan played in Top Gun.'}

Limit the number of results

You can limit the number of results from the Cypher QA Chain using the top_k parameter. The default is 10.

chain = GraphCypherQAChain.from_llm(
    ChatOpenAI(temperature=0),
    graph=graph,
    verbose=True,
    top_k=2,
    allow_dangerous_requests=True,
)

chain.invoke({"query": "Who played in Top Gun?"})

[1m> Entering new GraphCypherQAChain chain...[0m
Generated Cypher:
[32;1m[1;3mMATCH (a:Actor)-[:ACTED_IN]->(m:Movie)
WHERE m.name = 'Top Gun'
RETURN a.name[0m
Full Context:
[32;1m[1;3m[{'a.name': 'Tom Cruise'}, {'a.name': 'Val Kilmer'}][0m

[1m> Finished chain.[0m

{'query': 'Who played in Top Gun?',
 'result': 'Tom Cruise, Val Kilmer played in Top Gun.'}

Return intermediate results

You can return intermediate steps from the Cypher QA Chain using the return_intermediate_steps parameter

chain = GraphCypherQAChain.from_llm(
    ChatOpenAI(temperature=0),
    graph=graph,
    verbose=True,
    return_intermediate_steps=True,
    allow_dangerous_requests=True,
)

result = chain.invoke({"query": "Who played in Top Gun?"})
print(f"Intermediate steps: {result['intermediate_steps']}")
print(f"Final answer: {result['result']}")

[1m> Entering new GraphCypherQAChain chain...[0m
Generated Cypher:
[32;1m[1;3mMATCH (a:Actor)-[:ACTED_IN]->(m:Movie)
WHERE m.name = 'Top Gun'
RETURN a.name[0m
Full Context:
[32;1m[1;3m[{'a.name': 'Tom Cruise'}, {'a.name': 'Val Kilmer'}, {'a.name': 'Anthony Edwards'}, {'a.name': 'Meg Ryan'}][0m

[1m> Finished chain.[0m
Intermediate steps: [{'query': "MATCH (a:Actor)-[:ACTED_IN]->(m:Movie)\nWHERE m.name = 'Top Gun'\nRETURN a.name"}, {'context': [{'a.name': 'Tom Cruise'}, {'a.name': 'Val Kilmer'}, {'a.name': 'Anthony Edwards'}, {'a.name': 'Meg Ryan'}]}]
Final answer: Tom Cruise, Val Kilmer, Anthony Edwards, and Meg Ryan played in Top Gun.

Return direct results

You can return direct results from the Cypher QA Chain using the return_direct parameter

chain = GraphCypherQAChain.from_llm(
    ChatOpenAI(temperature=0),
    graph=graph,
    verbose=True,
    return_direct=True,
    allow_dangerous_requests=True,
)

chain.invoke({"query": "Who played in Top Gun?"})

[1m> Entering new GraphCypherQAChain chain...[0m
Generated Cypher:
[32;1m[1;3mMATCH (a:Actor)-[:ACTED_IN]->(m:Movie)
WHERE m.name = 'Top Gun'
RETURN a.name[0m

[1m> Finished chain.[0m

{'query': 'Who played in Top Gun?',
 'result': [{'a.name': 'Tom Cruise'},
  {'a.name': 'Val Kilmer'},
  {'a.name': 'Anthony Edwards'},
  {'a.name': 'Meg Ryan'}]}

Add examples in the Cypher generation prompt

You can define the Cypher statement you want the LLM to generate for particular questions

from langchain_core.prompts.prompt import PromptTemplate

CYPHER_GENERATION_TEMPLATE = """Task:Generate Cypher statement to query a graph database.
Instructions:
Use only the provided relationship types and properties in the schema.
Do not use any other relationship types or properties that are not provided.
Schema:
{schema}
Note: Do not include any explanations or apologies in your responses.
Do not respond to any questions that might ask anything else than for you to construct a Cypher statement.
Do not include any text except the generated Cypher statement.
Examples: Here are a few examples of generated Cypher statements for particular questions:
# How many people played in Top Gun?
MATCH (m:Movie {{name:"Top Gun"}})<-[:ACTED_IN]-()
RETURN count(*) AS numberOfActors

The question is:
{question}"""

CYPHER_GENERATION_PROMPT = PromptTemplate(
    input_variables=["schema", "question"], template=CYPHER_GENERATION_TEMPLATE
)

chain = GraphCypherQAChain.from_llm(
    ChatOpenAI(temperature=0),
    graph=graph,
    verbose=True,
    cypher_prompt=CYPHER_GENERATION_PROMPT,
    allow_dangerous_requests=True,
)

API Reference:PromptTemplate

chain.invoke({"query": "How many people played in Top Gun?"})

[1m> Entering new GraphCypherQAChain chain...[0m
Generated Cypher:
[32;1m[1;3mMATCH (m:Movie {name:"Top Gun"})<-[:ACTED_IN]-()
RETURN count(*) AS numberOfActors[0m
Full Context:
[32;1m[1;3m[{'numberOfActors': 4}][0m

[1m> Finished chain.[0m

{'query': 'How many people played in Top Gun?',
 'result': 'There were 4 actors in Top Gun.'}

Use separate LLMs for Cypher and answer generation

You can use the cypher_llm and qa_llm parameters to define different llms

chain = GraphCypherQAChain.from_llm(
    graph=graph,
    cypher_llm=ChatOpenAI(temperature=0, model="gpt-3.5-turbo"),
    qa_llm=ChatOpenAI(temperature=0, model="gpt-3.5-turbo-16k"),
    verbose=True,
    allow_dangerous_requests=True,
)

chain.invoke({"query": "Who played in Top Gun?"})

[1m> Entering new GraphCypherQAChain chain...[0m
Generated Cypher:
[32;1m[1;3mMATCH (a:Actor)-[:ACTED_IN]->(m:Movie)
WHERE m.name = 'Top Gun'
RETURN a.name[0m
Full Context:
[32;1m[1;3m[{'a.name': 'Tom Cruise'}, {'a.name': 'Val Kilmer'}, {'a.name': 'Anthony Edwards'}, {'a.name': 'Meg Ryan'}][0m

[1m> Finished chain.[0m

{'query': 'Who played in Top Gun?',
 'result': 'Tom Cruise, Val Kilmer, Anthony Edwards, and Meg Ryan played in Top Gun.'}

Ignore specified node and relationship types

You can use include_types or exclude_types to ignore parts of the graph schema when generating Cypher statements.

chain = GraphCypherQAChain.from_llm(
    graph=graph,
    cypher_llm=ChatOpenAI(temperature=0, model="gpt-3.5-turbo"),
    qa_llm=ChatOpenAI(temperature=0, model="gpt-3.5-turbo-16k"),
    verbose=True,
    exclude_types=["Movie"],
    allow_dangerous_requests=True,
)

# Inspect graph schema
print(chain.graph_schema)

Node properties are the following:
Actor {name: STRING}
Relationship properties are the following:

The relationships are the following:

Validate generated Cypher statements

You can use the validate_cypher parameter to validate and correct relationship directions in generated Cypher statements

chain = GraphCypherQAChain.from_llm(
    llm=ChatOpenAI(temperature=0, model="gpt-3.5-turbo"),
    graph=graph,
    verbose=True,
    validate_cypher=True,
    allow_dangerous_requests=True,
)

chain.invoke({"query": "Who played in Top Gun?"})

[1m> Entering new GraphCypherQAChain chain...[0m
Generated Cypher:
[32;1m[1;3mMATCH (a:Actor)-[:ACTED_IN]->(m:Movie)
WHERE m.name = 'Top Gun'
RETURN a.name[0m
Full Context:
[32;1m[1;3m[{'a.name': 'Tom Cruise'}, {'a.name': 'Val Kilmer'}, {'a.name': 'Anthony Edwards'}, {'a.name': 'Meg Ryan'}][0m

[1m> Finished chain.[0m

{'query': 'Who played in Top Gun?',
 'result': 'Tom Cruise, Val Kilmer, Anthony Edwards, and Meg Ryan played in Top Gun.'}

Provide context from database results as tool/function output

You can use the use_function_response parameter to pass context from database results to an LLM as a tool/function output. This method improves the response accuracy and relevance of an answer as the LLM follows the provided context more closely. You will need to use an LLM with native function calling support to use this feature.

chain = GraphCypherQAChain.from_llm(
    llm=ChatOpenAI(temperature=0, model="gpt-3.5-turbo"),
    graph=graph,
    verbose=True,
    use_function_response=True,
    allow_dangerous_requests=True,
)
chain.invoke({"query": "Who played in Top Gun?"})

[1m> Entering new GraphCypherQAChain chain...[0m
Generated Cypher:
[32;1m[1;3mMATCH (a:Actor)-[:ACTED_IN]->(m:Movie)
WHERE m.name = 'Top Gun'
RETURN a.name[0m
Full Context:
[32;1m[1;3m[{'a.name': 'Tom Cruise'}, {'a.name': 'Val Kilmer'}, {'a.name': 'Anthony Edwards'}, {'a.name': 'Meg Ryan'}][0m

[1m> Finished chain.[0m

{'query': 'Who played in Top Gun?',
 'result': 'The main actors in Top Gun are Tom Cruise, Val Kilmer, Anthony Edwards, and Meg Ryan.'}

You can provide custom system message when using the function response feature by providing function_response_system to instruct the model on how to generate answers.

Note that qa_prompt will have no effect when using use_function_response

chain = GraphCypherQAChain.from_llm(
    llm=ChatOpenAI(temperature=0, model="gpt-3.5-turbo"),
    graph=graph,
    verbose=True,
    use_function_response=True,
    function_response_system="Respond as a pirate!",
    allow_dangerous_requests=True,
)
chain.invoke({"query": "Who played in Top Gun?"})

[1m> Entering new GraphCypherQAChain chain...[0m
Generated Cypher:
[32;1m[1;3mMATCH (a:Actor)-[:ACTED_IN]->(m:Movie)
WHERE m.name = 'Top Gun'
RETURN a.name[0m
Full Context:
[32;1m[1;3m[{'a.name': 'Tom Cruise'}, {'a.name': 'Val Kilmer'}, {'a.name': 'Anthony Edwards'}, {'a.name': 'Meg Ryan'}][0m

[1m> Finished chain.[0m

{'query': 'Who played in Top Gun?',
 'result': "Arrr matey! In the film Top Gun, ye be seein' Tom Cruise, Val Kilmer, Anthony Edwards, and Meg Ryan sailin' the high seas of the sky! Aye, they be a fine crew of actors, they be!"}

Setting up​

Seeding the database​

Refresh graph schema information​

Enhanced schema information​

Querying the graph​

Limit the number of results​

Return intermediate results​

Return direct results​

Add examples in the Cypher generation prompt​

Use separate LLMs for Cypher and answer generation​

Ignore specified node and relationship types​

Validate generated Cypher statements​

Provide context from database results as tool/function output​

Was this page helpful?