Langchain pandas agent - Azure OpenAI account

Quick Fix: Enclose the agent.run() call in a try-except block to catch any exceptions that may be raised. If an exception is raised, print the error message. If no exception is raised, print the final answer.

The Solutions:

Solution 2: AzureChatOpenAI

Instead of utilizing the AzureOpenAI LLM, this solution suggests switching to AzureChatOpenAI LLM. This change resolves the issue of the LLM output containing both a final answer and a parsable action.

from langchain.agents import create_pandas_dataframe_agent
from langchain.chat_models import AzureChatOpenAI

import os
import pandas as pd

import openai

df = pd.read_csv("iris.csv")

openai.api_type = "azure"
os.environ["OPENAI_API_TYPE"] = "azure"
os.environ["OPENAI_API_KEY"] = "OPENAI_API_KEY"
os.environ["OPENAI_API_BASE"] = "https:<OPENAI_API_BASE>.openai.azure.com/"
os.environ["OPENAI_API_VERSION"] = "<OPENAI_API_VERSION>"

llm = AzureChatOpenAI(
    openai_api_type="azure",
    deployment_name="<deployment_name>",
    model_name="<model_name>"
)

agent = create_pandas_dataframe_agent(llm, df, verbose=True)
agent.run("how many rows are there?")

Solution 3: Modified output_parser.py for Azure OpenAI account

The error you encountered is resolved by implementing the code snippet below in your output_parser.py file.

def parse(self, text: str) -> Union[AgentAction, AgentFinish]:
    includes_answer = FINAL_ANSWER_ACTION in text
    regex = (
        r&quot;Action\s*\d*\s*:[\s]*(.*?)[\s]*Action\s*\d*\s*Input\s*\d*\s*:[\s]*(.*)&quot;
    )
    action_match = re.search(regex, text, re.DOTALL)
    if action_match and not includes_answer:
            if includes_answer:
                raise OutputParserException(
                    f&quot;{FINAL_ANSWER_AND_PARSABLE_ACTION_ERROR_MESSAGE}: {text}&quot;
                )
            action = action_match.group(1).strip()
            action_input = action_match.group(2)
            tool_input = action_input.strip(&quot; &quot;)
            # ensure if its a well formed SQL query we don&#39;t remove any trailing &quot; chars
            if tool_input.startswith(&quot;SELECT &quot;) is False:
                tool_input = tool_input.strip(&#39;&quot;&#39;)

            return AgentAction(action, tool_input, text)

    elif includes_answer:
        return AgentFinish(
            {&quot;output&quot;: text.split(FINAL_ANSWER_ACTION)[-1].strip()}, text
        )

    if not re.search(r&quot;Action\s*\d*\s*:[\s]*(.*?)&quot;, text, re.DOTALL):
        raise OutputParserException(
            f&quot;Could not parse LLM output: `{text}`&quot;,
            observation=MISSING_ACTION_AFTER_THOUGHT_ERROR_MESSAGE,
            llm_output=text,
            send_to_llm=True,
        )
    elif not re.search(

Q&A

How to fix the error in the code that you provided?

—

The error in the code may be caused by the LangChain agent’s execution to parse the LLM’s output. Use try-except block to catch any exceptions that may be raised.

Which LLM is recommended to be used with Langchain agents?

—

AzureChatOpenAI is recommended to be used with Langchain agents.

What is the reason behind the error ‘Parsing LLM output produced both a final answer and a parse-able action’?

—

The error ‘Parsing LLM output produced both a final answer and a parse-able action’ is caused by the logic of the output parser in the LangChain library.

Video Explanation:

The following video, titled "Langchain pandas agent Azure OpenAI account - YouTube", provides additional insights and in-depth exploration related to the topics discussed in this post.

Welcome to Mixible, your go-to source for comprehensive and informative content covering a broad range of topics from Stack Exchange ...

Langchain pandas agent – Azure OpenAI account – Azure