This package will get you up to speed with an infrastructure to classify texts using GenAI, store the texts in BigQuery and then visualise those into a Looker Studio dashboard
This solution assumes you already have a project created and set up where you wish to host these resources.
Time to complete: About 5 minutes
Click the Start button to move to the next step.
- Have an organization set up in Google cloud.
- Have a billing account set up.
- Have an existing project with billing enabled.
- The project has to be able to share resources outside the domain (i.e. to get public requests), check the domain restricted sharing policy.
In order to spin up this architecture, you will need to be a user with the “Project owner” IAM role on the existing project:
Note: To grant a user a role, take a look at the Granting and Revoking Access documentation.
Before we deploy the architecture, you will need the following information:
- The project ID
Once the repository is cloned please run the following command to install the prerequisistes:
sh prereq.sh
You will then be prompted to provide the project-id for the destination project.
After this is complete, you can kick off the Cloud Build pipeline with the following command:
gcloud builds submit . --config build/cloudbuild.yaml --region us-central1
The region is the region where the cloudbuild will run as well as the region where the resulting compiled image of the sourcecode will be stored, if you update the region, it is important to also update it in the terraform variables.
At this point you have successfully deployed the application and necessary underlining infrastructure to run it! Yay!
This process may take a while to deploy, please do not close the window when deploying.
Next we are going to test the architecture and finally clean up your environment.
- Visit Cloud Run list of services (https://console.cloud.google.com/run)
- Select the service called text-classification
- The service endpoint will be available right next to the package name
Once you deployed the solution successfully, update the populate.sh with the endpoint you got from the previous step
sh populate.shThen, check the parsed results in the output bucket in text (OCR) and json (Key=value) formats
Finally, check the json results on BigQuery or even better, use the bellow link to check it in a Looker Studio dashboard.
https://lookerstudio.google.com/c/u/0/reporting/create?c.mode=edit&ds.connector=BIG_QUERY&ds.type=TABLE&ds.projectId=[YOUR PROJECT ID]&ds.datasetId=classified_messages&ds.tableId=classified_messages
Feel free to play around and take this example to new usecases, in code/main.py you have the prompt that is being used to create the emotions, you could alternativelly update that to e.g classify emails for a law firm into court cases:
prompt = f"""You are a lawyer assistant and need to classify a given email of a law case by its content, the process email text is "{text}". Give me a result such as Criminal, Civil, Family, Probate and Corporate. Refrain from explaining your answer"""
Then for testing you can update the strings in the populate.sh file:
"I hope this email finds you well. My name is John Doe, and I am seeking legal assistance regarding an employment dispute with my former employer, ABC Corporation. I was terminated last month under circumstances that I believe were unlawful. Could we schedule a consultation to discuss the details of my case? I have attached relevant documents, including my employment contract and termination letter, for your review."
"My name is Michael Johnson, and I am in urgent need of legal representation. I have been charged with assault following an incident that occurred on June 10, 2024. I maintain that I acted in self-defense and believe that the charges against me are unjust. I am currently out on bail and need to appear in court for my first hearing on July 1, 2024. I would like to schedule a meeting with you as soon as possible to discuss my case and prepare my defense. I have attached the police report and bail documents for your review. Please let me know your earliest availability for a consultation. I am very anxious about this situation and am looking for strong legal guidance."
Hint: You can generate those using Gemini :)
Execute the command below on Cloud Shell to destroy the resources.
gcloud builds submit . --config build/cloudbuild_destroy.yaml
The above command will delete the associated resources so there will be no billable charges made afterwards.