prompt update and cleanup

agent.py

@@ -14,7 +14,7 @@ from langchain_ollama import ChatOllama
 from langgraph.graph.message import add_messages
 from langgraph.graph import START, StateGraph
 from langgraph.prebuilt import ToolNode, tools_condition
-
+from langchain_core.runnables.config import RunnableConfig
 from langchain_google_genai import ChatGoogleGenerativeAI
 from langchain_groq import ChatGroq
 from langchain_huggingface import ChatHuggingFace, HuggingFaceEndpoint
@@ -25,9 +25,6 @@ from langchain_community.tools import (
     DuckDuckGoSearchRun,
     ArxivQueryRun,
     ShellTool,
-    ReadFileTool,
-    WriteFileTool,
-    ListDirectoryTool,
 )
 from langchain_community.utilities import (
     WikipediaAPIWrapper,
@@ -110,7 +107,7 @@ def extract_text_multimodal(img_path: str) -> str:
         with open(img_path, "rb") as image_file:
             image_bytes = image_file.read()
 
-        # image_base64 = base64.b64encode(image_bytes).decode("utf-8")
+        image_base64 = base64.b64encode(image_bytes).decode("utf-8")
 
         message = [
             HumanMessage(
@@ -120,9 +117,8 @@ def extract_text_multimodal(img_path: str) -> str:
                         "text": "Extract all the text from this image. Return only the extracted text, no explanations."
                     },
                     {
-                        "type": "image",
-                        # "image_url": {"url": f"data:image/png;base64,{image_base64}"}
-                        "image": image_bytes
+                        "type": "image_url",
+                        "image_url": {"url": f"data:image/png;base64,{image_base64}"}
                     }
                 ]
             )
@@ -149,44 +145,6 @@ def read_excel_file(file_path: str, sheet_name: Optional[str] = None) -> pd.Data
         return pd.DataFrame({"error": [f"Excel reading error: {str(e)}"]})
 
 
-
-@tool
-def analyze_excel_data(file_path: str, operation: str = "describe") -> str:
-    """
-    Analyze an Excel (.xlsx or .xls) file and perform basic statistical or structural operations.
-
-    Parameters:
-        file_path (str): The path to the Excel file to be analyzed.
-        operation (str): The type of analysis to perform. Options include:
-            - "describe": Returns summary statistics for numeric columns.
-            - "info": Returns the DataFrame's structure and non-null counts.
-            - "columns": Returns a list of column names in the Excel file.
-            - "shape": Returns the shape of the DataFrame as a tuple (rows, columns).
-
-    Returns:
-        str: A string representation of the requested analysis output. If an error occurs
-             or an unknown operation is provided, a descriptive message is returned.
-    """
-    try:
-        df = pd.read_excel(file_path)
-        
-        if operation == "describe":
-            return df.describe().to_string()
-        elif operation == "info":
-            import io
-            buffer = io.StringIO()
-            df.info(buf=buffer)
-            return buffer.getvalue()
-        elif operation == "columns":
-            return str(list(df.columns))
-        elif operation == "shape":
-            return str(df.shape)
-        else:
-            return "Unknown operation. Available: describe, info, columns, shape"
-    except Exception as e:
-        return f"Excel analysis error: {str(e)}"
-
-
 @tool
 def visit_webpage(url: str) -> str:
     """
@@ -224,11 +182,6 @@ def build_tool():
     arxiv_tool = ArxivQueryRun(api_wrapper=ArxivAPIWrapper())
     shell_tool = ShellTool()
     python_repl = PythonREPLTool()
-    
-    # File management tools
-    read_file = ReadFileTool()
-    write_file = WriteFileTool()
-    list_directory = ListDirectoryTool()
 
     # Combine built-in tools with custom tools
     all_tools = [
@@ -238,9 +191,6 @@ def build_tool():
         arxiv_tool,
         shell_tool,
         python_repl,
-        read_file,
-        write_file,
-        list_directory,
 
         # Custom tools for specialized tasks
         reverse_text,
@@ -249,7 +199,6 @@ def build_tool():
         advanced_math,
         extract_text_multimodal,
         read_excel_file,
-        analyze_excel_data,
         visit_webpage,
     ]
     return all_tools
@@ -289,11 +238,6 @@ def build_langgraph(provider: str, model: Optional[str] = None, temperature: flo
         tools_description = """
             Available tools for the tasks:
 
-            FILE OPERATIONS:
-            - read_file: Read content from files
-            - write_file: Write content to files  
-            - list_directory: List files and directories
-
             WEB & SEARCH:
             - duckduckgo_search: Search the web for information
             - wikipedia_tool: Search Wikipedia for knowledge
@@ -314,7 +258,6 @@ def build_langgraph(provider: str, model: Optional[str] = None, temperature: flo
 
             DATA ANALYSIS:
             - read_excel_file: Read and preview Excel files
-            - analyze_excel_data: Perform statistical analysis on Excel data
 
             SYSTEM:
             - shell_tool: Execute shell commands (use carefully)
@@ -323,16 +266,41 @@ def build_langgraph(provider: str, model: Optional[str] = None, temperature: flo
         file = state["input_file"]
         sys_msg = SystemMessage(
             content=(
-                "You are a comprehensive AI agent designed to solve complex tasks using the tools provided. \n\n"
+                "You are an intelligent AI agent designed to solve complex problems using the tools provided.\n\n"
+                "=== Available Tools ===\n"
                 f"{tools_description}\n\n"
-                f"You have access to some optional files if any. Currently the loaded file is: {file}\n\n"
-                "Approach problems systematically:\n"
-                "1. Break down complex tasks into smaller steps\n"
-                "2. Use appropriate tools for each step\n" 
-                "3. Chain multiple tools together when needed\n"
-                "4. Verify results and provide clear explanations\n\n"
-                "Be thorough and accurate in your responses. "
-                "I will ask you a question. Report your thoughts, and finish your answer with only the final answer: [YOUR FINAL ANSWER]. YOUR FINAL ANSWER should be a number OR as few words as possible OR a comma separated list of numbers and/or strings. If you are asked for a number, don't use comma to write your number neither use units such as $ or percent sign unless specified otherwise. If you are asked for a string, don't use articles, neither abbreviations (e.g. for cities), and write the digits in plain text unless specified otherwise. If you are asked for a comma separated list, apply the above rules depending of whether the element to be put in the list is a number or a string."
+                "=== Optional Files ===\n"
+                f"Currently loaded file: {file}\n\n"
+                "=== Problem-Solving Process ===\n"
+                "Follow these steps carefully when answering a question:\n"
+                "1. Break the problem into smaller, manageable parts.\n"
+                "2. Choose the most suitable tool for each part.\n"
+                "3. Use multiple tools in sequence if needed.\n"
+                "4. Verify your results and explain your reasoning clearly.\n\n"
+                "Be precise and clear at every step. After your reasoning, provide ONLY the final answer.\n\n"
+                "=== Final Answer Format Rules ===\n"
+                "- For numbers: Use only digits (no commas or units) unless units are explicitly requested.\n"
+                "- For strings: Do not use articles (a, an, the) or abbreviations. Spell out all digits.\n"
+                "- For lists: Use commas to separate items. Apply the above number/string rules to each item.\n"
+                "- If the answer is unknown: Respond exactly with \"do not know\"\n\n"
+                "Example Question 1:\n\n"
+                "If Eliud Kipchoge could maintain his marathon pace indefinitely, how many thousand hours would it take him to run from Earth to the Moon at its closest approach? Use the minimum perigee distance from Wikipedia and round to the nearest 1000 hours. Do not use commas.\n\n"
+                "**Example Answer 1:**\n"
+                "17\n\n"
+                "**Example Reasoning Steps 1:**\n"
+                "1. Found Eliud Kipchoge's marathon pace: 4 minutes 37 seconds per mile.\n"
+                "2. Converted pace into hours per mile.\n"
+                "3. Found Moon's closest distance: 225623 miles.\n"
+                "4. Multiplied pace by distance to get total hours and rounded to nearest 1000.\n\n"
+                "Example Question 2:\n\n"
+                "Who are the pitchers with the number before and after Taishō Tamai's number as of July 2023? Give them to me in the form Pitcher Before, Pitcher After, use their last names only, in Roman characters.\n\n"
+                "**Example Answer 2:**\n"
+                "Yoshida, Uehara\n\n"
+                "**Example Reasoning Steps 2:**\n"
+                "1. Looked up Taishō Tamai on Wikipedia.\n"
+                "2. Found the pitcher with number 18 is Kōsei Yoshida.\n"
+                "3. Found the pitcher with number 20 is Kenta Uehara.\n\n"
+                "Now answer the following questions:\n"
             )
         )
         
@@ -371,8 +339,10 @@ if __name__ == "__main__":
 
     # Example: Knowledge retrieval
     print("📚 Testing Wikipedia search...")
-    messages = [HumanMessage(content="How many studio albums were published by Mercedes Sosa between 2000 and 2009 (included)? You can use the latest 2022 version of english wikipedia.")]
-    result = react_graph.invoke({"messages": messages, "input_file": None})
+    messages = [HumanMessage(content="Who nominated the only Featured Article on English Wikipedia about a dinosaur that was promoted in November 2016?")]
+
+    config = RunnableConfig(recursion_limit=10)
+    result = react_graph.invoke({"messages": messages, "input_file": None}, config)
     
     for m in result['messages']:
         m.pretty_print()

requirements.in

@@ -17,5 +17,5 @@ langgraph-sdk
 pandas
 wikipedia
 markdownify
-duckduckgo-search
+ddgs
 arxiv

requirements.txt

@@ -46,17 +46,17 @@ charset-normalizer==3.4.2
     # via requests
 click==8.2.1
     # via
-    #   duckduckgo-search
+    #   ddgs
     #   typer
     #   uvicorn
 dataclasses-json==0.6.7
     # via langchain-community
+ddgs==9.4.0
+    # via -r requirements.in
 distro==1.9.0
     # via
     #   groq
     #   openai
-duckduckgo-search==8.1.1
-    # via -r requirements.in
 exceptiongroup==1.3.0
     # via anyio
 fastapi==0.116.1
@@ -205,7 +205,7 @@ langsmith==0.4.8
     #   langchain-community
     #   langchain-core
 lxml==6.0.0
-    # via duckduckgo-search
+    # via ddgs
 markdown-it-py==3.0.0
     # via rich
 markdownify==1.1.0
@@ -255,7 +255,7 @@ pandas==2.3.1
 pillow==11.3.0
     # via gradio
 primp==0.15.0
-    # via duckduckgo-search
+    # via ddgs
 propcache==0.3.2
     # via
     #   aiohttp