This APP works entirely in front-end.
All API calls to openai API are made from browsers, and it is not recommended but for testing purposes we can ignore the risk.
The logic here is based on 3 roles:
- QA, question is given and 3 responses are generated
- Reflexion, original question with all 3 responses is passed to another gpt instance to validate the responses
- Resolver, question, answers and critique is passed to another gpt instance to compose final answer
This is based on YouTube video from AI Explained. Watch it for explanation here.