AgentExecutor chain #3500
-
|
I want to host a python agent on Gradio. It all works good. Im struggling when wanting to display not only the answer but also the AgentExecutor chain. How can I edit the code, so that also the AgentExecutor chain will be printed in the Gradio app. The Code snippet for that part is the following: |
Beta Was this translation helpful? Give feedback.
Replies: 4 comments 6 replies
-
|
I assume you want the Thought, Action, Action Input, Observation steps to be printed in Gradio. You can access the First, enable the def answer_question(question):
agent_exe = initialize_agent(
llm=OpenAI(temperature=0, max_tokens=1000),
tool=PythonREPLTool(),
return_intermediate_steps=True,
verbose=True,
)
response = agent_exe({"input": question})
answer = response["output"]
steps = response["intermediate_steps"]
return answer, stepsThen you can use something like the ifaces = gr.Interface(
fn=answer_question,
inputs=gr.Textbox(label="Question"),
outputs=[gr.Textbox(label="Answer"), gr.JSON(label="Steps")],
title="Question Answering Agent",
description="A simple question answering agent."
) |
Beta Was this translation helpful? Give feedback.
-
|
Hey, thanks for the help, unfortunately I get an Error while trying it. I used your code: when executing and submitting I get the following error: I can't manage to fix this error maybe you can help again. |
Beta Was this translation helpful? Give feedback.
-
|
That is strange. Code looks correct and it works for me: Can you try updating gradio ( If that does not work, then show the full stack trace here. |
Beta Was this translation helpful? Give feedback.
-
|
sure, my code looks like this: |
Beta Was this translation helpful? Give feedback.
-
|
Can you also show the full stack trace of the error message you receive? |
Beta Was this translation helpful? Give feedback.
-
|
sure: Traceback (most recent call last): |
Beta Was this translation helpful? Give feedback.
-
|
I see. You should use the from langchain.agents import load_tools
def answer_question(question):
agent_exe = initialize_agent(
llm=OpenAI(temperature=0),
tools=load_tools(["python_repl"]),
return_intermediate_steps=True,
verbose=True,
)
response = agent_exe({"input": question})
answer = response["output"]
steps = response["intermediate_steps"]
return answer, steps |
Beta Was this translation helpful? Give feedback.
-
|
Perfect, it worked!!! thank you! |
Beta Was this translation helpful? Give feedback.
-
|
I have one more question, do you know how I can edit the JSON output, so that it will have a more readable format? |
Beta Was this translation helpful? Give feedback.
-
|
You can add some pre-processing to the |
Beta Was this translation helpful? Give feedback.
I see. You should use the
load_toolsfunction to properly load the tools: