Feature: Text Editor with Completion

[ash1eygrace]

Think simple Google docs editor with minimal functionality (maybe markdown) with a button to feed GTP-3 context, and call the Completion mode API to finish the sentences.

AI completion:

• Title input field + Description Input / Summary field to feed big-picture context to the language model.
• Tone input field to present the language model with desired prose, e.g., Stephen King, Martin Luther King Jr., {insert other author or person here}...
• Creativity temperature slider* Buttons might be better for UX if 2-3 sweet spots could be found.
• Buttons for Length of completion (short, medium, long lengths. )
• UX Natural writing use of H1-H6 (headers could likely serve as the section piece of context to get a more accurate response e.g. Title > Description > Header = response within the context of header)

Research needed:

• Determined the sweet spot for contextual field vs. contextual paragraphs within the editor with presence and frequency penalties
• Play with the length of completion response from API and define 2-3 sweet spots maybe measured by sentence length
• Check payload sent from Jarvis, Closer's copy. If it's not a variable from the backend, it may give useful data for conceptualization
• Test various flows. UX should be linear e.g: Open page where the user is immediately prompted to give context with a title and description which serves as an overview --> ask the user for tone/creativity vs strict --> start with a blank page or have the option to create a boilerplate template, e.g. get AI to write blog outline for {topic with x, y, z).

Other notes:

• Probably work a refactor the API to use variables for the API completion request body vs static values before this, since both the mini-apps and this editor will use the same API call. May be worth exploring Slider Component that could be implemented within both types of completion apps.