Skip to content

PDF Agent

Agent for interacting with PDF documents via question and answer.

This module initializes and compiles a LangGraph application that enables users to query PDF documents using a question_and_answer tool. It integrates a language model and follows the ReAct pattern to process and answer queries related to PDF content.

Usage

app = get_app("unique_thread_id") response = app.invoke(initial_state)

get_app(uniq_id, llm_model)

Initializes and returns the LangGraph application for the PDF agent.

This function sets up the PDF agent by loading configuration settings via Hydra, initializing a model, and creating a workflow graph that incorporates PDF-specific tools. The agent is built using the ReAct pattern to facilitate interactive querying and processing of PDF documents.

Parameters:

Name Type Description Default
uniq_id str

A unique identifier for the current conversation session or thread.

 required
llm_model BaseChatModel

The language model instance to be used. Defaults to ChatOpenAI(model="gpt-4o-mini", temperature=0).

 required

Returns:

Name Type Description
StateGraph

A compiled LangGraph application capable of handling PDF interactions.
Example

app = get_app("thread_123") result = app.invoke(initial_state)

Source code in aiagents4pharma/talk2scholars/agents/pdf_agent.py 
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
def get_app(uniq_id, llm_model: BaseChatModel):
    """
    Initializes and returns the LangGraph application for the PDF agent.

    This function sets up the PDF agent by loading configuration settings via Hydra,
    initializing a model, and creating a workflow graph that incorporates
    PDF-specific tools. The agent is built using the ReAct pattern to facilitate interactive
    querying and processing of PDF documents.

    Args:
        uniq_id (str): A unique identifier for the current conversation session or thread.
        llm_model (BaseChatModel, optional): The language model instance to be used.
            Defaults to ChatOpenAI(model="gpt-4o-mini", temperature=0).

    Returns:
        StateGraph: A compiled LangGraph application capable of handling PDF interactions.

    Example:
        >>> app = get_app("thread_123")
        >>> result = app.invoke(initial_state)
    """
    # Load configuration using Hydra.
    with hydra.initialize(version_base=None, config_path="../configs"):
        cfg = hydra.compose(
            config_name="config",
            overrides=["agents/talk2scholars/pdf_agent=default"],
        )
        cfg = cfg.agents.talk2scholars.pdf_agent
        logger.info("Loaded pdf_agent configuration.")

    def agent_pdf_node(state: Talk2Scholars):
        """
        Processes the current state by invoking the language model for PDF question and answer.

        Args:
            state (Talk2Scholars): The current conversation state containing query details
            and context.

        Returns:
            Any: The response generated by the language model after processing the state.
        """
        logger.info("Creating Agent_PDF node with thread_id %s", uniq_id)
        response = model.invoke(state, {"configurable": {"thread_id": uniq_id}})
        return response

    # Define the tool node that includes the PDF QnA tool.
    tools = ToolNode([question_and_answer_tool, query_results])

    logger.info("Using OpenAI model %s", llm_model)

    # Create the agent using the provided BaseChatModel instance.
    model = create_react_agent(
        llm_model,
        tools=tools,
        state_schema=Talk2Scholars,
        prompt=cfg.pdf_agent,
        checkpointer=MemorySaver(),
    )

    # Define a new workflow graph with the state schema.
    workflow = StateGraph(Talk2Scholars)
    workflow.add_node("agent_pdf", agent_pdf_node)
    workflow.add_edge(START, "agent_pdf")

    # Initialize memory to persist state between runs.
    checkpointer = MemorySaver()

    # Compile the graph into a runnable app.
    app = workflow.compile(checkpointer=checkpointer, name="agent_pdf")
    logger.info("Compiled the PDF agent graph.")

    return app