Sentiment Classification

Exposed endpoint : http://34.72.165.159:8000/

In this project, we will deploy a sentiment analyser model using fastapi on GCP using GKE.

• Containerizing different components of projects
• Writing tests and testing individual modules using pytest
• Using trunk for automatic code checking, formatting and liniting
• Deploying application on GKE

Run

Locally

• Sentiment Model

cd sentiment
docker build -t sentiment -f Dockerfile.t .
cd ..
docker run --rm -it -v $(pwd):/app sentiment bash
python3 sentiment/model.py
pytest tests/test_sentiment_model.py

• FastAPI

Note: To run pytest replace requirements-fastapi.txt with requirements-dev.txt in Dockerfile

docker build -t sentiment -f Dockerfile.fastapi .
docker run --rm -it -v $(pwd):/app -p 8000:8000 sentiment bash
uvicorn main:app --host=0.0.0.0
pytest --cov tests/

• GKE

Deploy the application on Google Kubernetes Engine (GKE).

Guide: https://cloud.google.com/kubernetes-engine/docs/how-to/automated-deployment

Expose port of 8000 after deploy under Workloads > ACTIONS > Expose in Kubernetes Engine.

Setup up automatic deploy using Cloud Build and Github under Workloads > ACTIONS > Automated deployment.

Test the application endpoint

curl -X 'POST' \
  'http://34.72.165.159:8000/classify?input_text=i%20like%20you' \
  -H 'accept: application/json' \
  -d ''

Additional Exercise

• Deploy this application using gcloud cli