|
|
import os
|
|
|
from pathlib import Path
|
|
|
from typing import Optional, Union
|
|
|
|
|
|
import pandas as pd
|
|
|
from dotenv import load_dotenv
|
|
|
from smolagents import (CodeAgent, DuckDuckGoSearchTool, FinalAnswerTool,
|
|
|
LiteLLMModel, PythonInterpreterTool,
|
|
|
WikipediaSearchTool)
|
|
|
from smolagents.tools import Tool
|
|
|
from tabulate import tabulate
|
|
|
|
|
|
|
|
|
load_dotenv()
|
|
|
|
|
|
|
|
|
model = LiteLLMModel(
|
|
|
model_id=os.getenv("GEMINI_MODEL"), api_key=os.getenv("GEMINI_API_KEY")
|
|
|
)
|
|
|
|
|
|
|
|
|
class ExcelToTextTool(Tool):
|
|
|
"""Render an Excel worksheet as a Markdown table."""
|
|
|
|
|
|
name = "excel_to_text"
|
|
|
description = (
|
|
|
"Read an Excel file and return a Markdown table of the requested sheet. "
|
|
|
"Accepts either the sheet name or a zero-based index (as a string)."
|
|
|
)
|
|
|
|
|
|
inputs = {
|
|
|
"excel_path": {
|
|
|
"type": "string",
|
|
|
"description": "Path to the Excel file (.xlsx or .xls).",
|
|
|
},
|
|
|
"sheet_name": {
|
|
|
"type": "string",
|
|
|
"description": (
|
|
|
"Worksheet name or zero-based index (as a string). "
|
|
|
"Optional; defaults to the first sheet."
|
|
|
),
|
|
|
"nullable": True,
|
|
|
},
|
|
|
}
|
|
|
|
|
|
output_type = "string"
|
|
|
|
|
|
def forward(self, excel_path: str, sheet_name: Optional[str] = None) -> str:
|
|
|
"""Load the Excel file and return the sheet as a Markdown table.
|
|
|
|
|
|
Args:
|
|
|
excel_path: Path to the Excel file.
|
|
|
sheet_name: Optional name or index of the sheet to read. If None, reads the first sheet.
|
|
|
|
|
|
Returns:
|
|
|
A Markdown table representing the Excel sheet, or an error message if the file is not found or cannot be read.
|
|
|
"""
|
|
|
|
|
|
file_path = Path(excel_path).expanduser().resolve()
|
|
|
if not file_path.is_file():
|
|
|
return f"Error: Excel file not found at {file_path}"
|
|
|
|
|
|
try:
|
|
|
sheet: Union[str, int] = (
|
|
|
int(sheet_name)
|
|
|
if sheet_name and sheet_name.isdigit()
|
|
|
else sheet_name or 0
|
|
|
)
|
|
|
|
|
|
df = pd.read_excel(file_path, sheet_name=sheet)
|
|
|
|
|
|
if hasattr(df, "to_markdown"):
|
|
|
return df.to_markdown(index=False)
|
|
|
|
|
|
return tabulate(df, headers="keys", tablefmt="github", showindex=False)
|
|
|
|
|
|
except Exception as e:
|
|
|
return f"Error reading Excel file: {e}"
|
|
|
|
|
|
|
|
|
class GaiaAgent:
|
|
|
"""An agent capable of using tools to answer general questions."""
|
|
|
|
|
|
def __init__(self):
|
|
|
"""Initializes the GaiaAgent with a set of tools."""
|
|
|
|
|
|
print("GaiaAgent initialized with tools.")
|
|
|
|
|
|
tools = [
|
|
|
DuckDuckGoSearchTool(),
|
|
|
WikipediaSearchTool(),
|
|
|
ExcelToTextTool(),
|
|
|
PythonInterpreterTool(),
|
|
|
FinalAnswerTool(),
|
|
|
]
|
|
|
|
|
|
self.agent = CodeAgent(
|
|
|
model=model,
|
|
|
tools=tools,
|
|
|
add_base_tools=True,
|
|
|
additional_authorized_imports=["pandas", "numpy", "csv", "subprocess"],
|
|
|
)
|
|
|
|
|
|
def __call__(self, task_id: str, question: str) -> str:
|
|
|
"""Processes a question using the agent and its tools.
|
|
|
|
|
|
Args:
|
|
|
task_id: A unique identifier for the task.
|
|
|
question: The question to be answered.
|
|
|
|
|
|
Returns:
|
|
|
The answer generated by the agent.
|
|
|
"""
|
|
|
print(f"Agent received task_id='{task_id}' | question='{question[:50]}...'")
|
|
|
answer = self.agent.run(question)
|
|
|
print(f"Agent returning answer: {answer}")
|
|
|
return answer
|
|
|
|
