Assistants API Integration

[randomwangran]

It looks like nobody has proposed any new issues here. Maybe it's the
challenge to bring all those new things into this great emacs ai
package. I'll just bring some basic the things into users attention. I
hope as we further learn the basics of how the Assistant API works, We
can extend our package with even more power.

The documentaion is Here OpenAI Platform

• Step 1: Create an Assistant

  I don't know if we can set up the assistent setting from the
  web interface instead of the Python script. Because what I
  found is that on the web interface, we have more easily
  control over the things. Ideally, I'm thinking about once we
  create a new assistant, we have a unique ID for that system,
  and then we can use that ID to communicate the things that we
  are interested in.

• Step 2: Create a Thread

• Step 3: Add a Message to a Thread

  Step two and three looks to me is the same thing, which we
  kind of start a new section to talk to this assistant. But,
  yeah, different from our previous function in this package,
  now we have ability to upload files, and in the near future we
  can upload images in the coming months.

• Step 4: Run the Assistant

• Step 5: Display the Assistant's Response

  Steps four and five is essentially we are sending our input to
  API and we expect data or output from the API and then we
  decode the content and properly process within Emacs, either
  we do region replacement or we do some magic Org-mode
  manipulations.

For me, I need to figure out a minimum elisp function to properly set
up the request information and then gather feedback from the current
API and from there I can probably tweak about the new API.

[karthink]

Support for the assistants API is not currently planned, as this is OpenAI only and I prefer to work on features that work across the board.

Moving this to a discussion in case people have suggestions or thoughts.