Automated task fulfillment

[layterz]

This is still a WIP - I still need to test locally and write some unit tests, but it should be enough code to demonstrate the approach I'm taking. I have a couple of questions so wanted to check in now to get some feedback before continuing.

Changes:

• Added a TaskResponder with a complete(task) method. Only created a ChatGPTTaskResponder for now.
• Add two new methods to TreeManager:
  • select_task_responder this selects a responder (e.g. human, chatgpt, open-assistant, etc) from a list configured in settings.
  • complete_task which accepts a task and a model and attempts to complete the task with the responder defined by the model.
• In request_task I've added a while loop which calls tm.select_task_responder(request). A task is created for the selected responder and the loop continues until a task for the requesting user has been created. I'm not sure if this is the intended logic.
• Added a unique nullable ai_model field to User.
• Created a new complete-pending-ai-tasks celery worker which iterates over all users with an ai_model and finds and pending tasks for that user. Then it iterates over those tasks and attempts to complete them using the TreeManager method described above.

This approach means there'll need to be a DB user per automated responder, which complicates setup a bit. But the benefits are that it works with most of the existing code with few changes and any logic to rate limit, log, etc per user will be reused.

@yk can you take a look and see if this fits the requirements you had in mind?

[github-actions[bot]]

:x: pre-commit failed. Please run pre-commit run --all-files locally and commit the changes. Find more information in the repository's CONTRIBUTING.md

[olliestanley]

Thanks for this PR. Haven't reviewed but will just say I am pretty sure we don't want to include a ChatGPT task responder, the idea here would be to give these tasks to OA models running on the inference server :D (maybe also HF-hosted endpoints or other open-source stuff)

[layterz]

Ha, yeah I kind of guessed. That's just placeholder to demonstrate how the responders would work. Will try to update with an OA example

[andreaskoepf]

(converted to draft due to wip status)

[github-actions[bot]]

:x: pre-commit failed. Please run pre-commit run --all-files locally and commit the changes. Find more information in the repository's CONTRIBUTING.md

[github-actions[bot]]

:x: pre-commit failed. Please run pre-commit run --all-files locally and commit the changes. Find more information in the repository's CONTRIBUTING.md

[github-actions[bot]]

:x: pre-commit failed. Please run pre-commit run --all-files locally and commit the changes. Find more information in the repository's CONTRIBUTING.md

[github-actions[bot]]

:x: pre-commit failed. Please run pre-commit run --all-files locally and commit the changes. Find more information in the repository's CONTRIBUTING.md

[olliestanley]

Hi @layterz are you still working on this?

[layterz]

Hi @olliestanley,

Sorry - I've been travelling so dropped this unfortunately. I think the code is mostly there, but was having trouble testing it due to docker crashes. I can try again now I'm back, but won't have time until next week realistically. If that's blocking happy for you to reassign.

Rich