This is a sample notebook and web application which shows how Google Vertex AI Generative AI can be used with Neo4j to build and consume Knowledge Graphs using Video Transcripts.

This notebook parses data from AFL after-match Commentary transcripts using Google Vertex AI Generative AI's text-bison and code-bison models. The models will be prompted to recognise and extract entities and relationships. We will then generate Neo4j Cypher queries using them and write the data to a Neo4j database.

We will again use code-bison model and prompt it to convert questions in english to Cypher - Neo4j's query language, which can be used for data retrieval.

To get started, create a managed notebook in Google Cloud Vertex AI. Be sure to select "single user" when starting a managed notebook to run this, otherwise the auth won't allow access to the preview.

Once that has started, open the notebook and a terminal window within that. Clone this repo with the command:

git clone https://github.com/neo4j-partners/intelligent-app-from-videos-google-generativeai-neo4j.git

The notebook uses Python 3.8 for LangChain. However the managed notebooks are currently on 3.7. So, we'll need to install a newer version of Python. You can do that by running these commands in the terminal.

VENV=py38
conda create -y -q -p $HOME/conda_env/$VENV python=3.8 ipykernel
source /opt/conda/bin/activate ~/conda_env/$VENV
python -m ipykernel install --user --name $VENV
conda install -y -c conda-forge ipywidgets

The notebook at notebook/notebook.ipynb walks through prompts and tuning a model. You will need to run that before the UI. While running the notebook, make sure you select the py38 kernel you just created above.

The UI application is based on Streamlit. In this example we're going to show how to run it on a Google Compute Engine (GCE) VM. First, deploy a VM. You need to replace your environment specific values in the command below:

export VM_INSTANCE_NAME='neo4j-gcp-genai-demo'
export GCP_PROJECT_NAME=$(gcloud config get-value project)
gcloud compute instances create $VM_INSTANCE_NAME \
    --project=$GCP_PROJECT_NAME \
    --zone=us-central1-c \
    --machine-type=e2-medium \
    --network-interface=network-tier=PREMIUM,stack-type=IPV4_ONLY,subnet=default \
    --maintenance-policy=MIGRATE --provisioning-model=STANDARD \
    --scopes=https://www.googleapis.com/auth/devstorage.read_only,https://www.googleapis.com/auth/logging.write,https://www.googleapis.com/auth/monitoring.write,https://www.googleapis.com/auth/servicecontrol,https://www.googleapis.com/auth/service.management.readonly,https://www.googleapis.com/auth/trace.append \
    --tags=allow-http,http-server \
    --create-disk=auto-delete=yes,boot=yes,device-name=$VM_INSTANCE_NAME,image=projects/debian-cloud/global/images/debian-11-bullseye-v20230509,mode=rw,size=10,type=projects/$GCP_PROJECT_NAME/zones/us-central1-c/diskTypes/pd-balanced \
    --no-shielded-secure-boot \
    --shielded-vtpm --shielded-integrity-monitoring \
    --labels=goog-ec-src=vm_add-gcloud --reservation-affinity=any

Next, login to the new VM instance:

gcloud compute ssh --zone "us-central1-c" $VM_INSTANCE_NAME --project $GCP_PROJECT_NAME

We're going to be running the application on port 80. That requires root access, so first:

sudo su

Then you'll need to install git and clone this repo:

apt install -y git
mkdir -p /app
cd /app
git clone https://github.com/neo4j-partners/intelligent-app-from-videos-google-generativeai-neo4j.git
cd intelligent-app-from-videos-google-generativeai-neo4j

Login using GCP credentials via the gcloud cli.

gcloud auth application-default login

Let's install python & pip first:

apt install -y python
apt install -y pip

Now, let's create a Virtual Environment to isolate our Python environment and activate it

apt-get install -y python3-venv
python3 -m venv /app/venv/genai
source /app/venv/genai/bin/activate

To install Streamlit and other dependencies:

cd ui
pip install -r requirements.txt

Check if streamlit command is accessible from PATH by running this command:

streamlit --version

If not, you need to add the streamlit binary to PATH variable like below:

export PATH="/app/venv/genai/bin:$PATH"

Next up you'll need to create a secrets file for the app to use. Open the file and edit it:

cd streamlit
cd .streamlit
cp secrets.toml.example secrets.toml
vi secrets.toml

You will now need to edit that file to reflect your credentials. The file has the following variables:

GCP_PROJECT = "" # Your GCP project ID
GCP_LOCATION = "us-central1" # Location
TUNED_CYPHER_MODEL = "" # If you have a tuned Codey Model, provide here. Else, Leave it blank
NEO4J_HOST = "bolt://host.docker.internal" # Neo4j URL
NEO4J_PORT = "7687" # Neo4j Port
NEO4J_USER = "neo4j" # Neo4j User Name
NEO4J_PASSWORD = "Foo12345678" #Neo4j Password
NEO4J_DB = "neo4j" #Neo4j Database Name

Now we can run the app with the commands:

cd ..
streamlit run Home.py --server.port=80

On a GCP VM to run on port 80:

• Ensure you are a root or has access to run on port 80

• If you are running sudo, you also need to run the gcloud auth command above as a sudoer. And ensure that streamlit is accessible from the PATH.

• Ensure that the VM has port 80 open for HTTP access. You might need to open that port or any other via firewall-rules. You can use the following gcloud command to open the port. Make sure you replace with relevant values. You also need to add network tags to your VM before executing this command:

  gcloud compute firewall-rules create <rule-name> --allow tcp:80 --source-tags=<list-of-your-instances-name-tags> --source-ranges=0.0.0.0/0 --description="<your-description-here>"

Once deployed, you will be able to see the Dashboard and Chat UI:

From the Chat UI, you can ask questions like:

1. Who are the most commonly talked players?
2. Who are the players mentioned in episode 1?
3. What are the headlines for episode 7?
4. What are the top 5 common themes across all episodes combined?
5. What predictions are made in episode 6?
6. what is the gist of episode 11?
7. I love the Bombers. Which episodes do you recommend?
8. I follow Mason Cox. Which episodes do you recommend?