```
A Powerful Llm-based Agent With Versatile Tools
```

Question

and use llm = OpenAI( model=os.getenv("OPENAI_MODEL_NAME"), api_key=os.getenv("OPENAI_API_KEY") ) tools = [ FunctionTool.from_defaults( fn=search_llm_tool, name="search_llm_tool", description=SEARCH_TOOL_DESCRIPTION.format(query=self.query_data.query, file_attached=file_attached) ), FunctionTool.from_defaults( fn=file_tool_wrapper, description=FILE_TOOL_DESCRIPTION.format(file_attached=file_attached) ), FunctionTool.from_defaults( fn=lambda query: gen_image_tool(query), description="ONLY use this to generate an image as requested by the user." ), ] # Create the agent agent = ReActAgent.from_tools(tools, llm=llm, verbose=True)

Logan M · Answer

The verbose logs should give more details on its thought process

shawtyisaten · Answer

thing is it used to work.

shawtyisaten · Answer

started breaking only recently

shawtyisaten · Answer

without any code changes

shawtyisaten · Answer

like i get this in the endAction: Action Input: {"query": "generate an image of a frog"}

shawtyisaten · Answer

in the logs, i see > Running step 7a1351bc-709f-4306-94ca-db32ed5bed4b. Step input: Please respond to the conversation Thought: (Implicit) I can answer without any more tools! Answer: Action: Action Input: {"query": "generate an image of a frog"} full_response Action: Action Input: {"query": "generate an image of a frog"}

Logan M · Answer

hmmm, I think passing in your function as a lambda might be breaking things

Logan M · Answer

I'm surprised this worked before honestly

Logan M · Answer

define an actual function instead

Logan M · Answer

def gen_image_tool(query: str) -> None: "ONLY use this to generate an image as requested by the user.": ... tool = FunctionTool.from_defaults(gen_image_tool)

shawtyisaten · Answer

hmm how do you pass in the parameters to the function then? like query in this case?

Logan M · Answer

What do you mean? The llm will pass it to the tool

Logan M · Answer

Basically this turns the function into a JSON schema that the LLM can understsand

Logan M · Answer

and it will write the inputs

Logan M · Answer

This is why the type annotations are important

Find answers from the community

``` A Powerful Llm-based Agent With Versatile Tools ```