Introduction

this project is to implement Graph RAG algorithm with langchain

Usage

install neo4j

download jar of apoc from official github

download jar of neosemantics from official github

install neo4j with the following commands

wget -O - https://debian.neo4j.com/neotechnology.gpg.key | sudo apt-key add -
echo 'deb https://debian.neo4j.com stable latest' | sudo tee /etc/apt/sources.list.d/neo4j.list
sudo apt update
sudo apt install daemon cypher-shell neo4j
sudo cp <path/to/downloaded/apoc/jar> /var/lib/neo4j/plugins
sudo cp <path/to/downloaded/neosemantics/jar> /var/lib/neo4j/plugins
sudo echo "dbms.directories.plugins=/var/lib/neo4j/plugins" >> /etc/neo4j/neo4j.conf
sudo echo "dbms.security.procedures.unrestricted=algo.*,apoc.*" >> /etc/neo4j/neo4j.conf
sudo echo "dbms.unmanaged_extension_classes=n10s.endpoint=/rdf" >> /etc/neo4j/neo4j.conf
sudo echo "server.default_listen_address=0.0.0.0" >> /etc/neo4j/neo4j.conf
sudo echo "apoc.export.file.enabled=true" >> /etc/neo4j/apoc.conf
sudo echo "apoc.import.file.enabled=true" >> /etc/neo4j/apoc.conf
sudo systemctl start neo4j
sudo systemctl status neo4j

start the neo4j desktop by executing neo4j in console.

add a new dbms and add a database.

Install prerequisite packages

python3 -m pip install -r requirements.txt

Fix issues of langchain-experimental

edit line 551 of <path/to/site-packages>/langchain_experimental/graph_transformers/llm.py to change the code from

except NotImplementedError:

to

except:

edit line 595 of <path/to/site-packages>/langchain_experimental/graph_transformers/llm.py to change the code from

parsed_json = self.json_repair.loads(raw_schema.content)

to

parsed_json = self.json_repair.loads(raw_schema)

edit line 597 of <path/to/site-packages>/langchain_experimental/graph_transformers/llm.py to change the code from

for rel in parsed_json:
    # Nodes need to be deduplicated using a set
    nodes_set.add((rel["head"], rel["head_type"]))
    nodes_set.add((rel["tail"], rel["tail_type"]))

to

for rel in parsed_json:
    if type(rel) is not dict: continue
    # Nodes need to be deduplicated using a set 
    nodes_set.add((rel["head"], rel["head_type"]))
    nodes_set.add((rel["tail"], rel["tail_type"]))

Start service

python3 main.py --password <password> [--doc_dir <path/to/document/directory>]

if doc_dir argument is provided the knowledge graph is extracted from the documents

Other Information

View neo4j Database

Open browser using URL: http://localhost:7474

some queries you may try