Gpt-4 playground API could use a little tweak

I’d like to share some ideas to improve the Playground API:

  1. Deleting inputs/outputs: It’s not easy, especially with many items. A more user-friendly approach is needed.

  2. Reset button: Adding a reset button would help users clear all inputs and outputs quickly.

  3. Stop output button: Having a stop output button would allow users to halt output generation when needed.

  4. I know i’ve called for this before but I’d appreciate a Dark mode

I believe these suggestions would make the Playground API more efficient and easier to use. Let’s discuss these ideas and any other thoughts you may have

1 Like

remember, it’s a playground. it’s not supposed to provide you with a great working environment.


You may not do a lot of work in the Playground, but I do.

Not saying otherwise, just explaining what is the purpose of the playground, and to set your expectations accordingly.

What do you think the purpose of the playground is?

to play
aka, to test things, do quick trials, experiment the parameters, etc… NOT a place to do extensive work

1 Like

Well some users like me, might find themselves spending more time in the Playground for various reasons. It would be beneficial to consider their needs and explore potential improvements that could enhance the overall experience for everyone.

1 Like

Lol’d reading through this pedantic back and forth. I like the improvement ideas, but for my needs personally of just trying things out quickly here and there it’s fine as it is. Obviously just because the word “play” is in the name does not mean anything in terms of whether or not it can or should be improved :stuck_out_tongue:

I see where you’re coming from, but the idea of the playground, as @nunodonato has mentioned, is to just provide a simple API wrapper for testing and comparisons. Ideally you would create your own wrapper to effectively utilize the API, and then compare with playground to ensure your outputs are correct before creating your pipeline.

Although, I do like your suggestions and wouldn’t mind them myself.
The 3rd suggestion is possible by clicking the submit button during an output.

1 Like is good for this. Kind of a ChatGPT format but with a lot more features, and it uses the API version. It has the things you mentioned. I was lucky to get the early-bird cost, 'cause I’m broke, but it’s worth it.