This project uses yarn
, and runs on node 18 (check .nvmrc for the exact version). Run it locally with yarn dev
.
A venn diagram that combines two concepts or ideas for you. Using a generative language model, we give the user three++ possible ways to combine their initial concepts.
Uses OpenAI's GPT-3 to generate ideas.
Users can save generated results as text or PNGs.
Basic functionality working on the prototype
Users should be able to:
- enter their OpenAI API key to make requests
- make sure that API key isn't stored or exposed to 3rd parties
- enter two words or sentences. Max length 100 characters.
- click a button/area that triggers the combination action: sends a request to OpenAI
- use text-davinici-003
- return three options, each less than 180 characters
- see a loading animation while the request is resolving
- see the returned results
- be able to swipe/scroll/click through the three returned results
- be able to reset the whole venn to be blank
- be able to edit the existing inputs and re-run the generation
- be able to generate more than three results
- this could be re-running the whole generation, or clicking somewhere to generate a fourth or fifth result in addition to the first three
- copy generated results to the clipboard
- save a PNG of the final venn digram with inputs and generated results
- shared a generated PNG and link to Twitter
- animated rainbow gradient on venn circles – maybe a requirement 🌈
- Wireframe main user flow
- Wireframe onboarding sequence
- Final designs that meet all the requirements
- Write an example dataset of prompts and good answers to fine-tune on
- Good enough prompt engineering
- Research how we can accept OpenAI keys and use them for API requests without compromising user security