Multi-agent guidelines with LlamaIndex for parallel agents with different LLMs and memory

The community members are discussing the use of multi-agent guidelines with LlamaIndex, specifically regarding the ability to use agents in parallel with different language models (LLMs) and maintain memory in the loop, similar to CrewAI. A community member shared a video example that demonstrates the use of workflows in LlamaIndex, which they believe provides a nice interface and shows how to make customizations. However, another community member notes that the example runs agents sequentially, though they mention that tool calls can be run concurrently.

The discussion continues, with a community member expressing interest in seeing more examples of using various tools, such as custom functions, through the CrewAI and LlamaIndex integration, beyond what is covered in the documentation. Another community member suggests that understanding how workflows work in LlamaIndex can provide a lot of customization options compared to other high-level frameworks like CrewAI, though they admit to not having much experience with CrewAI themselves.

The community members then discuss some specific issues they are facing, such as the Gemini class not implementing the FunctionCallingLLM interface, and how to use different LLMs for each agent in the workflow. A community member provides a solution for the latter, suggesting that the LLM can be pulled from the selected agent configuration within the workflow.