GPT Prompt Suggestions

This is a simple web-applicaton that relies on OpenAI's API (GPT-3.5 Turbo) to suggest message on hold prompts to play while a customer is on hold, on the phone. It could be used as part of a order form for voice recording services. Or whatever else you can think of :)

Technologies used:

• Vue 3 (Vite)
• Express

Setup

Install

pnpm install -r

Create a .env file with the following content:

OPENAI_API_KEY=<YOUR_OPENAI_API_KEY>

Other options (optional; mostly for development):

VITE_BASE_URL=http://localhost:5000 # for development only (default '')
PORT=5000 # optional, server port
HOST=0.0.0.0 # optional, server interface for ex. for Docker (default 127.0.0.1)

Run

Production

pnpm run build
pnpm run start

• front-end http://localhost:5000/
• back-end http://localhost:5000/api

Dokku

This will deploy the application as Docker container, listening on port 5000.

git remote add dokku dokku@<YOUR_DOKKU_HOST>:<YOUR_APPLICATION_NAME>
git push dokku master
dokku config:set <YOUR_APPLICATION_NAME> OPENAI_API_KEY=<YOUR_OPENAI_API_KEY>
dokku config:set <YOUR_APPLICATION_NAME> HOST=0.0.0.0

Development

This will start both server and client in development / HMR mode.

pnpm run dev

• front-end http://localhost:5173/
• back-end http://localhost:5000/api

If you wanted to change something, the best place to start is introduction.json; This is where I provide instructions to the model.

TODO

• Validate input on backend
• Add some basic tests
• Restructure (move server to ./server)
• Security