Response for the LLM is concated with the system + human message when i use create_react_agent

[Nithin745]

Checked other resources

• I added a very descriptive title to this question.
• I searched the LangChain documentation with the integrated search.
• I used the GitHub search to find a similar question and didn't find it.

Commit to Help

• I commit to help with one of those options 👆

Example Code

import os

from langchain_community.vectorstores import FAISS, Chroma
from langchain_core.prompts import (
    ChatPromptTemplate,
)
from langchain_core.messages import AnyMessage
from langchain_core.tools import tool
from langchain_huggingface import ChatHuggingFace
from langchain_huggingface.llms import HuggingFacePipeline
from langgraph.prebuilt import create_react_agent
from rich import print

from app.core.config import BASE_DIR, MODEL_PATH, url_object

def scratch():
    prompt = ChatPromptTemplate.from_messages(
        [
            ("system", "You are a helpful assistant. Respond only in Spanish."),
            ("placeholder", "{messages}"),
        ]
    )
    system_prompt = "You are a helpful assistant. Respond only in Spanish."


    def _modify_messages(messages: list[AnyMessage]):
        return prompt.invoke({"messages": messages}).to_messages() + [("user", "Also say 'Pandamonium!' after the answer.")]


    query = "what is the value of magic_function(3)?"
    model_path = os.path.join(BASE_DIR, "models", "unsloth/llama-3-8b-Instruct-bnb-4bit")
    llm = HuggingFacePipeline.from_model_id(
        model_id=model_path,
        task="text-generation",
        pipeline_kwargs=dict(
            max_new_tokens=2048,
            do_sample=False,
            repetition_penalty=1.03,
            # quantization_config= quantization_config
        ),
    )


    @tool
    def magic_function(input: int) -> int:
        """Applies a magic function to an input."""
        return input + 2

    tools = [magic_function]
    model = ChatHuggingFace(llm=llm)
    app = create_react_agent(model, tools, messages_modifier=system_prompt)
    for s in app.stream({"messages": [("human", query)]}, stream_mode="values"):
        message = s["messages"][-1]
        print(s)

Description

I am using langchain to create sql agent that will write and execute query and give me the results, but for some reason instead of giving me the results only in the output it is concatenating systemMessage + HumanMessage + AIMessage as a single string, so i even tried one tutorial from langchain for still im facing the same issue
this is the code i followed => https://python.langchain.com/v0.2/docs/how_to/migrate_agent/#in-langgraph-4
and this is the output i got.

{
    'messages': [
        HumanMessage(content='what is the value of magic_function(3)?', id='e0d9fe03-ff0e-4e7f-bba0-4b610f3c1863'),
        AIMessage(
            content='<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nYou are a helpful assistant. Respond only in Spanish.<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nwhat is the value of
magic_function(3)?<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\nDisculpa, pero no tengo información sobre una función llamada "magic_function". ¿Podrías proporcionarme más contexto o definir la
función para que pueda ayudarte mejor?',
            id='run-89f748d6-7b98-4fa7-b3a9-9d485405404b-0'
        )
    ]
}

System Info

langchain==0.2.6
langchain-chroma==0.1.1
langchain-community==0.2.6
langchain-core==0.2.10
langchain-huggingface==0.0.3
langchain-openai==0.1.11
langchain-text-splitters==0.2.2
langchainhub==0.1.20

platform (linux )
python version = 3.12

[dosubot[bot]]

Hey there, @Nithin745! I'm here to help you with any bugs, questions, or contributions. Let's work together to solve the issue you're facing.

To resolve the issue of concatenated messages and get only the AI's response in the output, you can use the _parse_ai_message function to parse out just the string response from the AIMessage. Here is how you can do it:

1. Import the necessary functions and classes.
2. Use the _parse_ai_message function to parse the AIMessage.

Here is the relevant code:

from langchain_core.messages import AIMessage
from langchain.agents.output_parsers.openai_functions import OpenAIFunctionsAgentOutputParser

# Create an AIMessage
msg = AIMessage(content="Model response.")

# Parse the AIMessage
parser = OpenAIFunctionsAgentOutputParser()
result = parser._parse_ai_message(msg)

# Extract the response
response = result.return_values["output"]
print(response)

This will ensure that only the AI's response string is returned in the output ^[1][2][3].

To continue talking to Dosu, mention @dosu.