Agent Swarm Experiments

[BrianP8701]

Guys I've went off to work in my own repo. i find that everyone has similar ideas, but is implementing in various ways and its hard for me to wrap my head around the organization of this repo. Im working on my own agent swarm implementation here:
https://github.com/BrianP8701/Agent_Swarm_Experiments

I'll post updates here as i make progress. maybe if i do something good I can bring it over to this repo.

The main idea is breaking tasks into small enough pieces that it is manageable.

[BrianP8701]

its definitely a mess rn so dont bother looking. Im hacking shit together for now. My current workflow loop is:

1. Run the swarm with a goal
2. When the swarm fails or breaks, i step in and automate where it failed
3. Repeat with a harder goal

[jreitman007]

Have you run into issues with rate limits yet? A problem I'm realizing exists right now is that API calls exponentially increase at each level, so it seems very likely a HAAS swarm will get limited before it becomes big enough to be useful.

[BrianP8701]

no not yet. my swarm is very action oriented. so far the swarm breaks down/ needs my assistance before i even get too far. As my swarm becomes more autonomous and is able to get farther on its own... ill have to see if the cost of tokens to what the swarm creates is worth it.

As for rate limits, ill add a checkpoint ability to the swarm so it can save its progress and continue later.