Proposal: ChatGPT integration & GDocs AI extension

Proposals Round 8
#1

Name of Project: ChatGPT integration & GDocs AI extension

Proposal in one sentence: Add ChatGPT server to Algovera and Create an AI extension for gDocs

Description of the project and what problem it is solving:

  1. ChatGPT is only available through the OpenAI interface, or through unofficial Python packages (PyChatGPT). This makes it more difficult for devs to build on top of ChatGPT (until OpenAI releases their API), as they need to use Python to get around the Auth0 restrictions OpenAI has in place. I would build a REST API integrated with Algovera for devs to be able to integrate ChatGPT functions with any language, without needing to set up their own server/need to use Python.

  2. A lot of AI models today are being used to increase our own productivity. As a lot of people use Google Docs, integrating AI models (that aren’t necessarily known by everyone) directly into GDocs creates an easy to use interface for users, with no sign up, no payments - just natively in Google Docs. I’d create a Google Docs Add-on Script that allows users to use ChatGPT, Stable Diffusion and other models.
    For example: A user is writing up a document and needs pictures to make an explanation clearer. Instead of heading to a website to enter text and get an image, they can select the text that describes the image, click on Extensions > Algovera > Stable Diffusion, which would generate and paste an (or several) image(s) directly into the area under the text they selected.

Grant Deliverables:

  • Modular Google Docs extension to integrate select AI/ML models (ChatGPT, Stable Diffusion, etc.).
  • Production ready REST API for ChatGPT to integrate with Algovera

Squad

  • Name: Cesar Huret (kesar)
  • Twitter handle: @cesarhuret
  • Discord handle: Kesar#1129
1 Like
#2

A potential concern of the following is that it might be against OpenAI’s terms of service policy.

  • Production ready REST API for ChatGPT to integrate with Algovera
#3

This sounds like a really good use-case for stable diffusion given that the models are lightweight and open-source and could become an easily extensible tool that the community can benefit from.

Re: chatGPT maybe in the future this may become possible to integrate as open-source clones catch up, but it seems like a very big stretch to claim both these models can be integrated as part of a single project. It is ok to be more realistic and less ambitious and maybe label this “v1” of a multi-part project. All the best!

#4

Instead, could integrate GPT-3 into google docs and spend the grant funds to pay for API credits.

#5

Sure, that sounds like a viable alternative if your key focus is text generation. I thought you also wanted to do image generation and my advice was just to pick one at a time.