Task subdivision algorithm

[xtalax]

Hi all, recently saw the video in Dave Shapiero's yt about this project and am keen to get involved.

I have been thinking about efficient task subdivision.
What you want is to break larger tasks in to subtasks recursively until you reach a set of base tasks that the model thinks it can solve in a single shot, or few step shot, branching on different strategic options. You then count the number of base tasks in a subtree from the leaves up to the root to determine the most efficient strategy.

This can later include elements like frustration causing you to re-evaluate and re subdivide tasks of underestimated difficulty up to and including giving up and choosing another strategy.

Does this sound interesting? Is there anywhere I can jump in to a chat with you guys to discuss some more ideas?

[DataBassGit]

Yes!

We are planning to rework the Task management agents in the coming weeks. However, if you can develop a concept that will solve for the task problem, we would love to add it to the system. The primary concern is that you will need to stick to the task table schema.

Task text in the documents field,

metadatas needs to have
Task Order
Task Status
Task Desc (Same text as in documents)
A timestamp will be automatically added to each task upon creation or update

A UUID is added automatically to each task entry in the ID field.

The Metadatas field is pretty flexible, so you can add as many additional fields as you need. If you think this schema needs adjustment, I'd be happy to discuss. We are taking some time off after the hackathon, then we are going to update the database handling functions to meet a more uniform approach. We will also be adjusting the agent templates to match these changes, so anything you build in the next couple of weeks will require some updates in the future.

Don't let that discourage you. We are looking to provide long term support for this project, so there will always be changes in the future. Please also add a comment section with a full autobiography, personal phone number, and next of kin. 😃

[xtalax]

Sounds easy enough with a new task constructor. Looks like a good schema.

Just wondering where the overarching goal given by the user is fed in.
For my autobiography, see my github history ;)

[DataBassGit]

The goal is fed from the default.json file in the Persona's category. It's the first section of the file labeled Objective. You can also manually define the individual tasks there as well. I expect the end product will have manually defined tasks with autogenerated subtasks, but we will need to do more in depth testing to determine what will work best.

The nice thing about the framework is that it's very easy to iterate on this architecture, so once we have all the features in place we can actually support many different logical structures.