Find answers from the community

Updated 3 months ago

hello to everyone

hello to everyone

I am using llamaindex ReAct agent to solve my problem. I try to get json schema from my own API. This schema is very important as a context to generate json based on the schema. Schema is quite big.

Agent starts very well. It understands what kind of schema I need. FunctionTool goes to API gets schema. It shows schema to me in verbose mode. And all I want is to return the schema to python code as a text. But after Observation step agent is failed with "Error: Could not parse output. Please follow the thought-action-input format. Try again.". And it restarts flow from the beginning until it reaches attempts limit

How can I get a response of the FunctionTool from the Agent whithout problems?
Attachment
Screenshot_2024-06-17_at_19.36.27.png
L
n
2 comments
The LLM is not smart enough to follow the prompt it seems

Maybe try increasing max_iterations, or use a different LLM
it was LLama-3 70B. I thought it was smart enough)
Add a reply
Sign up and join the conversation on Discord