Benefits of relying on LLaMAIndex's agent versus using OpenAI's assistants API or asking GPT-4

If you want, you can always build your own agent more from-scratch. Here's an example with workflows

https://docs.llamaindex.ai/en/stable/examples/workflow/function_calling_agent/

General workflows guide, if you haven't used them
https://docs.llamaindex.ai/en/stable/module_guides/workflow/#workflows

I mean, having control over the memory, chat history, exact tools setup, etc, is pretty powerful. Especially if you take the workflows apporach, you have total control over how tools are called

For example, you can make an agent that retrieves context on every message, and puts it in the system prompt, to improve accuracy when picking tools or answering the user
https://colab.research.google.com/drive/1wVCkvX7oQu1ZwrMSAyaJ8QyzHyfR0D_j?usp=sharing

black boxes like openai assistants will be a nightmare to debug and customize (in my opinion) -- it basically has to work for your use case out of the box, or it won't work at all

thnx