Hi everyone ,
I’ve been working with CrewAI and I have a use case where the Manager Agent needs to dynamically decide how to handle incoming queries.
Here’s the scenario:
-
A user query comes in.
-
The Manager Agent should decide:
- Whether it can be answered by a single agent, or
- Whether it needs to delegate tasks to two agents in parallel.
-
Once the agents complete their tasks, the Manager Agent should combine their responses to generate the final output.
I’ve previously implemented parallel task execution (not using the Manager Agent), but now I need the Manager to make the decision on which agents to involve based on the query.
So my questions are:
- Can the Manager Agent automatically decide to create and run tasks for multiple agents in parallel?
- Is there built-in support for the Manager to wait for multiple agent results before crafting a final response?
- If not, what’s the best way to implement this? Should I build custom logic for parallelism and response aggregation?
Any help, patterns, or examples from the community would be a huge help
Thanks in advance!