Skip to content

silviachen46/llm-graph-builder

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Knowledge Graph Builder App

This application is designed to convert PDF documents into a knowledge graph stored in Neo4j. It utilizes the power of OpenAI's GPT/Diffbot LLM(Large language model) to extract nodes, relationships and properties from the text content of the PDF and then organizes them into a structured knowledge graph using Langchain framework. Files can be uploaded from local machine or S3 bucket and then LLM model can be chosen to create the knowledge graph.

Getting started

⚠️ For the backend, if you want to run the LLM KG Builder locally, and don't need the GCP/VertexAI integration, make sure to have the following set in your ENV file :

GEMINI_ENABLED = False
GCP_LOG_METRICS_ENABLED = False

And for the frontend, make sure to export your local backend URL before running docker-compose by having the BACKEND_API_URL set in your ENV file :

BACKEND_API_URL="http://localhost:8000"
  1. Run Docker Compose to build and start all components:

    docker-compose up --build
  2. Alternatively, you can run specific directories separately:

    • For the frontend:

      cd frontend
      yarn
      yarn run dev
    • For the backend:

      cd backend
      python -m venv envName
      source envName/bin/activate 
      pip install -r requirements.txt
      uvicorn score:app --reload

To deploy the app and packages on Google Cloud Platform, run the following command on google cloud run:

# Frontend deploy 
gcloud run deploy 
source location current directory > Frontend
region : 32 [us-central 1]
Allow unauthenticated request : Yes
# Backend deploy 
gcloud run deploy --set-env-vars "OPENAI_API_KEY = " --set-env-vars "DIFFBOT_API_KEY = " --set-env-vars "NEO4J_URI = " --set-env-vars "NEO4J_PASSWORD = " --set-env-vars "NEO4J_USERNAME = "
source location current directory > Backend
region : 32 [us-central 1]
Allow unauthenticated request : Yes

Features

  • PDF Upload: Users can upload PDF documents using the Drop Zone.
  • S3 Bucket Integration: Users can also specify PDF documents stored in an S3 bucket for processing.
  • Knowledge Graph Generation: The application employs OpenAI/Diffbot's LLM to extract relevant information from the PDFs and construct a knowledge graph.
  • Neo4j Integration: The extracted nodes and relationships are stored in a Neo4j database for easy visualization and querying.
  • Grid View of source node files with : Name,Type,Size,Nodes,Relations,Duration,Status,Source,Model

Setting up Environment Variables

Create .env file and update the following env variables.
OPENAI_API_KEY = ""
DIFFBOT_API_KEY = ""
NEO4J_URI = ""
NEO4J_USERNAME = ""
NEO4J_PASSWORD = ""
AWS_ACCESS_KEY_ID = ""
AWS_SECRET_ACCESS_KEY = ""
EMBEDDING_MODEL = ""
IS_EMBEDDING = "TRUE"
KNN_MIN_SCORE = ""

Setting up Enviournment Variables For Frontend Configuration

Create .env file in the frontend root folder and update the following env variables.
BACKEND_API_URL=""
BLOOM_URL=""
REACT_APP_SOURCES=""
LLM_MODELS=""
ENV=""
TIME_PER_CHUNK=

Functions/Modules

extract_graph_from_file(uri, userName, password, file_path, model):

Extracts nodes , relationships and properties from a PDF file leveraging LLM models.

Args:
 uri: URI of the graph to extract
 userName: Username to use for graph creation ( if None will use username from config file )
 password: Password to use for graph creation ( if None will use password from config file )
 file: File object containing the PDF file path to be used
 model: Type of model to use ('Gemini Pro' or 'Diffbot')

 Returns: 
 Json response to API with fileName, nodeCount, relationshipCount, processingTime, 
 status and model as attributes.
neoooo

create_source_node_graph(uri, userName, password, file):

Creates a source node in Neo4jGraph and sets properties.

Args:
 uri: URI of Graph Service to connect to
 userName: Username to connect to Graph Service with ( default : None )
 password: Password to connect to Graph Service with ( default : None )
 file: File object with information about file to be added

Returns: 
 Success or Failure message of node creation
neo_workspace

get_source_list_from_graph():

 Returns a list of file sources in the database by querying the graph and 
 sorting the list by the last updated date. 
get_source

Chunk nodes and embeddings creation in Neo4j

chunking

Application Walkthrough

KGB.mp4

Links

The Public Google cloud Run URL. Workspace URL

About

Neo4j graph construction from unstructured data

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Jupyter Notebook 55.6%
  • TypeScript 22.6%
  • Python 20.7%
  • CSS 0.7%
  • Dockerfile 0.2%
  • HTML 0.1%
  • JavaScript 0.1%