aggiunto tool analyze_excel_file per lavorare dui file Excel

agent.py

@@ -13,7 +13,8 @@ from langgraph.graph import StateGraph, END
 from langchain_google_genai import ChatGoogleGenerativeAI
 
 # Local imports
-from tools import get_search_tool, get_tavily_search_tool, get_wikipedia_tool, wikipedia_search, wikipedia_search_3, execute_python_code_from_file, download_taskid_file
+from tools import get_search_tool, get_tavily_search_tool, get_wikipedia_tool, wikipedia_search, wikipedia_search_3,\
+                  execute_python_code_from_file, download_taskid_file, analyze_excel_file
 
 # Nota: per i test in locale si usa il .env
 #       su HuggingFace invece si usano le variabili definite in Settings/"Variables and secrets"
@@ -58,7 +59,7 @@ chat = ChatGoogleGenerativeAI(
 search_tool = get_tavily_search_tool()
 #wikipedia_tool = get_wikipedia_tool()
 
-tools = [search_tool, wikipedia_search_3, execute_python_code_from_file, download_taskid_file]
+tools = [search_tool, wikipedia_search_3, execute_python_code_from_file, download_taskid_file, analyze_excel_file]
 
 # Bind tools to the model
 chat_with_tools = chat.bind_tools(tools)

tools.py

@@ -1,6 +1,7 @@
 import os, sys
 from dotenv import load_dotenv
 import requests
+import pandas as pd
 
 from langchain_community.tools import DuckDuckGoSearchRun
 from langchain_community.utilities import WikipediaAPIWrapper
@@ -167,4 +168,34 @@ def download_taskid_file(task_id: str, file_name: str) -> str:
             file.write(response.content)
         return os.path.abspath(file_name)
     except Exception as e:
-        return "Error occurred: {}".format(e)
+        return "Error occurred: {}".format(e)
+
+
+@tool
+def analyze_excel_file(file_path: str, query: str) -> str:
+    """
+    Analyzes an Excel (.xlsx) file using pandas.
+    Loads the specified Excel file into a pandas DataFrame and executes a Python query against it.
+    The query should be a valid pandas DataFrame operation (e.g., df.head(), df.describe(),
+    df[df['column_name'] > 10], df.groupby('category')['value'].mean()).
+    Returns the result of the query as a string (JSON or string representation).
+    """
+    if not os.path.exists(file_path):
+        return f"Error: File not found at {file_path}"
+    
+    try:
+        df = pd.read_excel(file_path)
+        
+        # Make the DataFrame accessible for the query
+        local_vars = {"df": df}
+        
+        # Execute the query
+        # IMPORTANT: Be extremely cautious with eval/exec for user-provided input in a production system.
+        # For a ReAct agent, the LLM generates this query, so it's generally safer
+        # if the LLM is well-constrained and reliable.
+        # For sensitive applications, consider a safer parsing mechanism or a restricted set of operations.
+        result = eval(query, {}, local_vars)
+        
+        return str(result) # Convert result to string for the LLM
+    except Exception as e:
+        return f"Error analyzing Excel file: {e}"