Hi all, just showing a new construct in Burr (for more info on Burr see - <a href="https://news.ycombinator.com/item?id=39917364">https://news.ycombinator.com/item?id=39917364</a>) for modeling agent / agentic parallelism. We think a "graph" mental model lends itself well to thinking about parallelism (& hierarchy) -- would love your thoughts.<p>Video walkthrough of notebook - <a href="https://youtu.be/G7lw63IBSmY" rel="nofollow">https://youtu.be/G7lw63IBSmY</a><p>Why did we build this feature? The future is only going to get more complex with LLMs & agents, and having the right constructs with integrated observability will help us get there. We naturally saw people doing this and thought we could help simplify it more.<p>Key Features of Burr's constructs:<p>* you can map over dynamic values in state (e.g. run an agent in parallel over many URLs)<p>* you can map over dynamic actions/or sub-graphs (e.g. running different agents in parallel)<p>* you can do the cartesian product between the two (e.g. run over prompts and actions/agents and do it in a loop to modify them...)<p>More details here - <a href="https://burr.dagworks.io/concepts/parallelism/">https://burr.dagworks.io/concepts/parallelism/</a>.<p>Things not mentioned but useful in this parallel context, Burr comes with pluggable durability/persistence/memory (see <a href="https://burr.dagworks.io/concepts/state-persistence/">https://burr.dagworks.io/concepts/state-persistence/</a>).<p>We'd love to get any early feedback from the HN community, ideas for where you'd like the framework to go, or UI/UX experiences in this space. We will be in the comments for the next few hours to respond! Or leave an issue or discussion here - <a href="https://github.com/dagworks-inc/burr">https://github.com/dagworks-inc/burr</a>. Thanks!
Forgot to mention, on the roadmap we have pluggable executors, e.g. delegating parallelism to whole agents/subgraphs to systems like "ray". Would love to collaborate with folks if that's interesting.