llm-response-bot

A rasa bot that can generate and rephrase responses on the fly using LLMs

Deploying

To deploy and run the bot, docker and docker-compose is required.

1. Clone the repository and do cd llm-response-bot
2. Make sure to copy the .env.template to .env and add your open-ai API key under the OPENAI_COMPLETIONS_API_KEY key
3. Go to the project root directory and run docker compose up -d to run everything
4. Widget should be available @ http://localhost:80

Talking to the bot

The bot is a demo Pizza Shop bot that can do the following:

• say greet
• show the Pizza Shop menu
• tell about Pizza prices
• tell price for an order
• say welcome
• say goodbye

All the bot responses are generated or rephrased using Open AI GPT-3 text-davinci-003 completions model.

In case the rate limit is met (3 requests within a minute last I checked), bot will utter the default rephrase response.