Query Dataframe

Tool for querying the metadata table of the last displayed papers.

This tool loads the most recently displayed papers into a pandas DataFrame and uses an LLM-driven pandas agent to answer metadata-level questions (e.g., filter by author, list titles). It is intended for metadata exploration only, and does not perform content-based retrieval or summarization. For PDF-level question answering, use the 'question_and_answer_agent'.

`NoPapersFoundError`

Bases: Exception

Exception raised when no papers are found in the state.

Source code in aiagents4pharma/talk2scholars/tools/s2/query_dataframe.py

class NoPapersFoundError(Exception):
    """Exception raised when no papers are found in the state."""

`QueryDataFrameInput`

Bases: BaseModel

Pydantic schema for querying the metadata of displayed papers.

Fields

question: A free-text prompt or Python expression to query the papers DataFrame. tool_call_id: LangGraph-injected identifier for tracking the tool invocation. state: Agent state dictionary. Must include: - 'last_displayed_papers': dictionary of paper metadata (rows = papers). - 'llm_model': model used to instantiate the DataFrame agent.

Notes

This tool is only for metadata queries. It does not perform full-text PDF analysis.
You can access standard metadata columns such as 'Title', 'Authors', 'Venue', 'Year', and the unified 'paper_ids' (list of :).
To extract raw paper IDs for downloading, set extract_ids=True and specify:
- id_column: 'paper_ids' (default)
- row_number: 1-based index of the paper to return (optional—for a single ID) Example: {{"question":"","extract_ids":True,"row_number":4}}

Source code in aiagents4pharma/talk2scholars/tools/s2/query_dataframe.py

class QueryDataFrameInput(BaseModel):
    """
    Pydantic schema for querying the metadata of displayed papers.

    Fields:
      question: A free-text prompt or Python expression to query the papers DataFrame.
      tool_call_id: LangGraph-injected identifier for tracking the tool invocation.
      state: Agent state dictionary. Must include:
        - 'last_displayed_papers': dictionary of paper metadata (rows = papers).
        - 'llm_model': model used to instantiate the DataFrame agent.

    Notes:
      - This tool is only for metadata queries. It does not perform full-text PDF analysis.
      - You can access standard metadata columns such as 'Title', 'Authors',
        'Venue', 'Year', and the unified 'paper_ids' (list of <source>:<ID>).
      - To extract raw paper IDs for downloading, set extract_ids=True and specify:
          - id_column: 'paper_ids' (default)
          - row_number: 1-based index of the paper to return (optional—for a single ID)
        Example: {{"question":"","extract_ids":True,"row_number":4}}
    """

    question: str = Field(
        description=(
            "The metadata query to run over the papers DataFrame. Can be natural language "
            "(e.g., 'List all titles by author X') or Python code "
            "(e.g., df['arxiv_id'].dropna().tolist())."
        )
    )
    extract_ids: bool = Field(
        default=False,
        description=(
            "If true, instruct the DataFrame agent to extract values from the"
            "specified ID column via a Python expression."
        ),
    )
    id_column: str = Field(
        default="paper_ids",
        description=(
            "Name of the metadata column containing a list of paper IDs to"
            "extract when extract_ids=True."
        ),
    )
    row_number: Optional[int] = Field(
        default=None,
        description=(
            "1-based index of the ID to extract from the list; if provided, returns only"
            "that single ID."
        ),
    )
    tool_call_id: Annotated[str, InjectedToolCallId]
    state: Annotated[dict, InjectedState]

`query_dataframe(question, state, tool_call_id, **kwargs)`

Perform a tabular query on the most recently displayed papers.

This function loads the last displayed papers into a pandas DataFrame and uses a pandas DataFrame agent to answer metadata-level questions (e.g., "Which papers have 'Transformer' in the title?", "List authors of paper X"). It does not perform PDF content analysis or summarization; for content-level question answering, use the 'question_and_answer_agent'.

Parameters:

Name	Type	Description	Default
`question`	`str`	The metadata query to ask over the papers table.	required
`extract_ids`	`bool`	If true, modify the query to instruct the DataFrame agent	required
`id_column`	`str`	Name of the metadata column to extract values from when extract_ids=True.	required
`row_number`	`int`	1-based index of the ID to extract from the list; if provided,	required
`state`	`dict`	The agent's state containing 'last_displayed_papers' key	required
`tool_call_id`	`str`	LangGraph-injected identifier for this tool call.	required

Returns:

Name	Type	Description
`Command`	`Command`	A structured response containing a ToolMessage with the query result.

Raises:

Type	Description
`NoPapersFoundError`	If no papers have been displayed yet.

Source code in aiagents4pharma/talk2scholars/tools/s2/query_dataframe.py

@tool(
    "query_dataframe",
    args_schema=QueryDataFrameInput,
    parse_docstring=True,
    return_direct=True,
)
def query_dataframe(
    question: str,
    state: Annotated[dict, InjectedState],
    tool_call_id: str,
    **kwargs: Any,
) -> Command:
    """
    Perform a tabular query on the most recently displayed papers.

    This function loads the last displayed papers into a pandas DataFrame and uses a
    pandas DataFrame agent to answer metadata-level questions (e.g., "Which papers have
    'Transformer' in the title?", "List authors of paper X"). It does not perform PDF
    content analysis or summarization; for content-level question answering, use the
    'question_and_answer_agent'.

    Args:
        question (str): The metadata query to ask over the papers table.
        extract_ids (bool): If true, modify the query to instruct the DataFrame agent
        to extract values from the specified ID column via Python code.
        id_column (str): Name of the metadata column to extract values from when extract_ids=True.
        row_number (int, optional): 1-based index of the ID to extract from the list; if provided,
        returns only that single ID.
        state (dict): The agent's state containing 'last_displayed_papers' key
        referencing the metadata table in state.
        tool_call_id (str): LangGraph-injected identifier for this tool call.

    Returns:
        Command: A structured response containing a ToolMessage with the query result.

    Raises:
        NoPapersFoundError: If no papers have been displayed yet.
    """
    logger.info("Querying last displayed papers with question: %s", question)
    llm_model = state.get("llm_model")
    if llm_model is None:
        raise ValueError("Missing 'llm_model' in state.")

    context_val = state.get("last_displayed_papers")
    if not context_val:
        logger.info("No papers displayed so far, raising NoPapersFoundError")
        raise NoPapersFoundError(
            "No papers found. A search needs to be performed first."
        )

    # Resolve the paper dictionary
    if isinstance(context_val, dict):
        dic_papers = context_val
    else:
        dic_papers = state.get(context_val)

    if not isinstance(dic_papers, dict):
        raise ValueError(
            "Could not resolve a valid metadata dictionary from 'last_displayed_papers'"
        )

    df_papers = pd.DataFrame.from_dict(dic_papers, orient="index")
    # Prepare the query: if extracting IDs, let the DataFrame agent handle it via Python code
    extract_ids_flag = kwargs.get("extract_ids", False)
    id_column = kwargs.get("id_column", "paper_ids")
    row_number = kwargs.get("row_number")
    question_to_agent = question
    if extract_ids_flag:
        if not id_column:
            raise ValueError("Must specify 'id_column' when extract_ids=True.")
        if row_number is not None:
            question_to_agent = (
                f"df['{id_column}'].dropna().str[0].tolist()[{row_number-1}]"
            )
        else:
            question_to_agent = f"df['{id_column}'].dropna().str[0].tolist()"
        logger.info(
            "extract_ids enabled: asking agent to run expression: %s", question_to_agent
        )

    df_agent = create_pandas_dataframe_agent(
        llm_model,
        allow_dangerous_code=True,
        agent_type="tool-calling",
        df=df_papers,
        max_iterations=5,
        include_df_in_prompt=True,
        number_of_head_rows=df_papers.shape[0],
        verbose=True,
    )

    llm_result = df_agent.invoke({"input": question_to_agent}, stream_mode=None)
    response_text = llm_result["output"]

    return Command(
        update={
            "messages": [
                ToolMessage(
                    content=response_text,
                    tool_call_id=tool_call_id,
                )
            ],
        }
    )