Does anyone have a pattern for

[violet cypress]

👋 Would native Golang concurrency work for you?
https://docs.dagger.io/cookbook/#execute-functions-concurrently
cc @brave wolf

[fiery mural]

It sounds like ideally you could put the assessor in the orchestrator's environment and let it manage calling the assessor tasks.
So one way to do that would be to have a function of the assessor that can accept multiple tasks and then split out using Golang concurrency like Jeremy said and collate the responses in your function.
Another way would be to let the model ask for multiple tool calls since some models are capable of requesting multiple tool calls at a time. So in that case the orchestrator would say call assessor(x), asssessor(y), and assessor(z). Unfortunately I think right now on the Dagger side that would still happen serially. @glass sentinel do you know if there's a reason we're doing this serially instead of using concurrency? https://github.com/dagger/dagger/blob/main/core/llm.go#L847

[glass sentinel]

yeah - early on, LLMs were trying to run a bunch of API calls in parallel, instead of chaining them

like for a.b.c.d it would instead run a > [b, c, d] in parallel which is totally bogus