Just a web interface for sending HTTP requests to the paid GPT API.

It uses the openai npm package.

Web application is bootstrapped using create-t3-app.

Update: Severely rate limited Front-end with unconnected API demo: https://gpt-demo.aaanh.app

Production: https://gpt.aaanh.app

Donate/Sponsor: https://ko-fi.com/aaanh

• Application telemetry (Kafka or Cassandra)
  • Emit request events with opaque body
  • Emit error events
  • Implement session-based LogRocket
• Cache prompts and responses in local storage
• Multi-turn implementation (similar to ChatGPT)
• Diffusion implementation (DALL-E 2)

• ConfigMap to export .env variables into the deployment
• Migrate from Linode bare-metal to AKS

• Install dependencies

  npm install
  

• Copy .env.example to .env

• Add your OpenAI API key in .env

• (Optional) Add your upstash Redis tokens for rate limiting

• Run

  npm run dev
  

Note: The steps below apply for Microsoft Azure cloud platform but the principles essentially can be applied on any cloud platforms.

Another note: Currently, the end-to-end build and deploy process only works on AMDx64 platforms. ARM support to be investigated.

• Docker Desktop with Kubernetes enabled
• Azure CLI
  • Private ACR created
  • Service Principal with acrPull role for that ACR
  • Have a Docker image of the frontend application built and pushed to the private ACR

• Create k8s secret in local cluster. E.g. acr-secret

  kubectl create secret docker-registry acr-secret --docker-server=my-private-cr.azurecr.io --docker-username=service-principal-application-id --docker-password=service-principal-client-secret

• Reference the k8s secret in the deployment manifest spec

  # ...
  container:
  # ...
  imagePullSecrets:
    - name: acr-secret

• Deploy the application

  kubectl apply -f k8s/gpt-frontend.yaml
  

• Navigate to localhost:3000

• TBA