ashishpatel26 / langchain2neo4j

Integrating Neo4j database into langchain ecosystem

Langchain2Neo4j

Cypher search is now part of the official LangChain library, learn more in: https://towardsdatascience.com/langchain-has-added-cypher-search-cb9d821120d5

The Langchain2Neo4j is a proof of concept application of how to integrate Neo4j into the Langchain ecosystem. This project took heavy inspiration from IMDB-LLM. The IMDB-LLM integrated graph search using networkx library into langchain ecosystem. I borrowed the idea and changed the project to use Neo4j as the source of information for the LLM.

The project now supports three modes of context search:

• Generating Cypher statements to query the database
• Full-text search of relevant entities
• Vector similarity search

Learn more: https://towardsdatascience.com/integrating-neo4j-into-the-langchain-ecosystem-df0e988344d2

Neo4j database

The project uses the Recommendation project that is available as part of the Neo4j Sandbox. If you want a local instance of Neo4j, you can restore a database dump that is available here.

Installation and Setup

1. Create a .env file and fill-in the environment variables as shown in the env.example

2. You need to create a full text index and import the movie embeddings:

sh seed_db.sh

If you are using Windows, you can also execute the code in data/database_seed.ipynb notebook to seed the database.

3. Run the project using docker-compose

docker-compose up

7. Open the application in your browser at http://localhost:3000

Example questions

# Who played in Top Gun?
MATCH (m:Movie)<-[r:ACTED_IN]-(a)
RETURN {{actor: a.name, role: r.role}} AS result
# What is the plot of the Copycat movie?
MATCH (m:Movie {{title: "Copycat"}})
RETURN {{plot: m.plot}} AS result
# Did Luis Guzmán appear in any other movies?
MATCH (p:Person {{name:"Luis Guzmán"}})-[r:ACTED_IN]->(movie)
RETURN {{movie: movie.title, role: r.role}} AS result
# Do you know of any matrix movies?
MATCH (m:Movie)
WHERE toLower(m.title) CONTAINS toLower("matrix")
RETURN {{movie:m.title}} AS result