GraphAcademy

Using Neo4j with LangChain

Neo4j and LangChain

Vectors

Text to Cypher

Retriever

In this lesson, you will use the GraphCypherQAChain to add a text to Cypher retriever to the LangChain agent.

Text to Cypher

Open the genai_integration_langchain/cypher_retriever.py file and review the code:

python
cypher_retriever.py
Unresolved directive in lesson.adoc - include::{repository-raw}/new-course/genai-integration-langchain/cypher_retriever.py[tag=**]

You will need to:

  1. Create a GraphCypherQAChain instance.

  2. Update the retriever function to use the GraphCypherQAChain to convert natural language queries into Cypher queries.

Create a new GraphCypherQAChain instance:

python
Unresolved directive in lesson.adoc - include::{repository-raw}/new-course/genai-integration-langchain/solutions/cypher_retriever.py[tag=import_cypher_qa]

Unresolved directive in lesson.adoc - include::{repository-raw}/new-course/genai-integration-langchain/solutions/cypher_retriever.py[tag=cypher_qa]

Return Direct

The return_direct parameter is set to True to return the result of the Cypher query instead of an answer.

This is useful when you want to pass the raw data to the agent for further processing or analysis.

Update the retrieve function to invoke the GraphCypherQAChain:

python
Unresolved directive in lesson.adoc - include::{repository-raw}/new-course/genai-integration-langchain/solutions/cypher_retriever.py[tag=retrieve]

Experiment with the retriever by running the program and asking questions about data in the graph:

Unresolved directive in lesson.adoc - include::{repository-raw}/new-course/genai-integration-langchain/solutions/cypher_retriever.py[tag=examples]

Verbose

Set the verbose parameter of the GraphCypherQAChain to True to see the generated Cypher queries and the full context used to generate the answers.

Click to view the complete code
python
Unresolved directive in lesson.adoc - include::{repository-raw}/new-course/genai-integration-langchain/solutions/cypher_retriever.py[tags="**;!examples"]

Improve the retriever

Your challenge is to improve the retriever using the techniques you learned in the previous lessons, which could include:

  • Providing a custom prompt and specific instructions.

  • Including example questions and Cypher queries.

  • Using a different LLM model for Cypher generation.

  • Restricting the schema to provide more focused results.

Here are some examples of more complex questions you can try:

Unresolved directive in lesson.adoc - include::{repository-raw}/new-course/genai-integration-langchain/solutions/cypher_retriever_enhanced.py[tag=examples]

There is no right or wrong solution. You should experiment with different approaches to see how they affect the accuracy and relevance of the generated Cypher queries.

Verbose

When adapting the retriever, you may want to set the verbose parameter of the GraphCypherQAChain to True to see the generated Cypher queries and the full context used to generate the answers.

Click here to see an example solution

This example uses some of the techniques you learned in the previous lessons to improve the retriever:

python
Unresolved directive in lesson.adoc - include::{repository-raw}/new-course/genai-integration-langchain/solutions/cypher_retriever_enhanced.py[tags="**; !examples"]

Congratulations!

When you are ready, click continue to complete the course.

Lesson Summary

In this lesson, you learned how to create a text to Cypher retriever and customized it to improve the quality of the generated Cypher queries.

Congratulations on completing the course!