v4

app.py

@@ -30,6 +30,7 @@ class WalidzAgent:
         else:
             # If "FINAL ANSWER:" is not found, return the original answer
             print("Warning: 'FINAL ANSWER:' not found in response")
+        print(answer)
         return answer
 
 # (Keep Constants as is)
@@ -215,7 +216,15 @@ def run_evaluation_only(profile: gr.OAuthProfile | None):
             print(f"Skipping item with missing task_id or question: {item}")
             continue
         try:
-            submitted_answer = agent(question_text)
+            # Prepare input state
+            human_msg = HumanMessage(content=question_text)
+            input_state = {
+                "messages": [human_msg],
+                "question_text": question_text
+            }
+
+            submitted_answer = agent(input_state)
+            #submitted_answer = agent(question_text)
             results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": submitted_answer})
         except Exception as e:
             print(f"Error running agent on task {task_id}: {e}")
@@ -226,7 +235,7 @@ def run_evaluation_only(profile: gr.OAuthProfile | None):
 
 # --- Build Gradio Interface using Blocks ---
 with gr.Blocks() as demo:
-    gr.Markdown("# Walid's Basic Agent Evaluation Runner (v3)")
+    gr.Markdown("# Walid's Basic Agent Evaluation Runner (v4)")
     gr.Markdown(
         """
         **Instructions:**

myagent_langraph.py

@@ -8,13 +8,11 @@ from langgraph.prebuilt import ToolNode, tools_condition
 #from IPython.display import Image, display
 from langchain_community.document_loaders import WikipediaLoader
 from langchain_community.tools.tavily_search import TavilySearchResults
+#from langchain_community.tools.youtube_search import YoutubeSearchResults
+from langchain_core.tools import tool
 
 from dotenv import load_dotenv
 import os
-import sys
-
-# Do not modify stdout/stdin as it can cause issues
-import codecs
 
 # Load environment variables
 load_dotenv()
@@ -30,6 +28,7 @@ class AgentState(TypedDict):
     messages: Annotated[list[AnyMessage], add_messages]
 
 ### TOOLS ###
+@tool
 def search_wikipedia(query: str) -> str:
     """Searches Wikipedia for a given query and returns maximum 1 results."""
     try:
@@ -59,6 +58,7 @@ def search_wikipedia(query: str) -> str:
         print(f"Error in search_wikipedia: {e}")
         return f"Error searching Wikipedia: {str(e)}"
 
+@tool
 def multiply(a: int, b: int) -> int:
     """Multiply two numbers.
     Args:
@@ -67,6 +67,7 @@ def multiply(a: int, b: int) -> int:
     """
     return a * b
 
+@tool
 def add(a: int, b: int) -> int:
     """Add two numbers.
     
@@ -76,6 +77,7 @@ def add(a: int, b: int) -> int:
     """
     return a + b
 
+@tool
 def subtract(a: int, b: int) -> int:
     """Subtract two numbers.
     
@@ -85,6 +87,7 @@ def subtract(a: int, b: int) -> int:
     """
     return a - b
 
+@tool
 def divide(a: int, b: int) -> int:
     """Divide two numbers.
     
@@ -96,6 +99,7 @@ def divide(a: int, b: int) -> int:
         raise ValueError("Cannot divide by zero.")
     return a / b
 
+@tool
 def modulus(a: int, b: int) -> int:
     """Get the modulus of two numbers.
     
@@ -105,19 +109,20 @@ def modulus(a: int, b: int) -> int:
     """
     return a % b
 
-
+@tool
 def web_search(query: str) -> str:
-    """Search Tavily for a query and return maximum 3 results.
+    """Search the web for a query and return maximum 3 results.
     
     Args:
         query: The search query."""
-    search_docs = TavilySearchResults(max_results=3).invoke(query=query)
+    search_tool = TavilySearchResults(max_results=3)
+    search_results = search_tool.invoke(query)
     formatted_search_docs = "\n\n---\n\n".join(
         [
-            f'<Document source="{doc.metadata["source"]}" page="{doc.metadata.get("page", "")}"/>\n{doc.page_content}\n</Document>'
-            for doc in search_docs
+            f'<Document source="{result["url"]}" title="{result.get("title", "")}"/>\n{result["content"]}\n</Document>'
+            for result in search_results
         ])
-    return {"web_results": formatted_search_docs}
+    return formatted_search_docs
 
 # Equip the butler with tools
 tools = [
@@ -138,6 +143,7 @@ try:
         openai_api_key=OPENAI_API_KEY,
         default_headers={"Content-Type": "application/json; charset=utf-8"}
     )
+    print("Binding tools...")
     llm_with_tools = llm.bind_tools(tools, parallel_tool_calls=False)
 except Exception as e:
     print(f"Error initializing OpenAI model: {e}")
@@ -152,19 +158,21 @@ def assistant(state: AgentState):
         # Build the system prompt piece by piece
         parts = [
             "You are a helpful assistant tasked with answering questions using a set of tools.",
-            "Report your thoughts, and finish your answer with the following template:",
-            "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.",
+            "Report your thoughts step by step.",
+            "Once you have determined the answer, provide it in the following XML format:",
+            "<FINAL_ANSWER>Your concise final answer here</FINAL_ANSWER>",
+            "Your final answer should be as concise as possible - a number OR as few words as possible OR a comma separated list of numbers and/or strings.",
+            "If asked for a number, don't use commas or units such as $ or percent sign unless specified otherwise.",
+            "If asked for a string, don't use articles or abbreviations (e.g. for cities), and write digits in plain text unless specified otherwise.",
+            "The answer inside <FINAL_ANSWER> tags must be complete, clear, and not truncated.",
             
             # Tool descriptions
             "You can use these provided tools:",
             "search_wikipedia(query: str) -> str:",
             "    Searches Wikipedia for a given query and returns maximum 1 results.",
-            "divide(a: int, b: int) -> float:",
-            "    Divide a and b"
+            "web_search(query: str) -> str:",
+            "    Searches the web for a query and returns maximum 3 results.",
+            "multiply, add, subtract, divide, modulus: Basic math operations between two numbers."
         ]
         
         # Add the question
@@ -211,9 +219,94 @@ def build_graph():
     return react_graph
 
 
+
+class WalidzAgent:
+    """A langgraph agent."""
+    def __init__(self):
+        print("BasicAgent initialized.")
+        self.graph = build_graph()
+
+    def __call__(self, question: str) -> str:
+        if isinstance(question, dict) and "messages" in question and "question_text" in question:
+            # We got a properly formatted input
+            input_state = question
+        else:
+            # Create a proper input state from the question string
+            human_msg = HumanMessage(content=question)
+            input_state = {
+                "messages": [human_msg],
+                "question_text": question
+            }
+
+        # Stream the execution and capture final step
+        final_state = None
+        for step in self.graph.stream(input_state):
+            print("--- STEP ---")
+            print(step)
+            final_state = step
+        
+        # Extract the answer from the final state
+        if final_state:
+            # Extract message content from the appropriate structure
+            message_content = None
+            
+            # Check if the result has the nested 'assistant' structure
+            if 'assistant' in final_state and 'messages' in final_state['assistant']:
+                # Get message from assistant's messages field
+                message = final_state['assistant']['messages']
+                message_content = message.content
+            # Fallback to direct messages list if available
+            elif 'messages' in final_state:
+                messages_list = final_state['messages']
+                if messages_list and len(messages_list) > 0:
+                    message_content = messages_list[-1].content
+            
+            if not message_content:
+                print("Warning: Could not extract message content from response")
+                return "Failed to extract message content from agent response"
+            
+            # First try XML tag extraction
+            import re
+            xml_pattern = re.compile(r'<FINAL_ANSWER>(.*?)</FINAL_ANSWER>', re.DOTALL)
+            match = xml_pattern.search(message_content)
+            
+            if match:
+                # Found XML tags
+                answer = match.group(1).strip()
+            else:
+                # Fall back to "FINAL ANSWER:" format
+                if "FINAL ANSWER:" in message_content:
+                    # Split by "FINAL ANSWER:" and take the last part
+                    final_answer_parts = message_content.split("FINAL ANSWER:")
+                    answer = final_answer_parts[-1].strip()
+                else:
+                    # If no explicit final answer format is found, return the whole message
+                    print("Warning: No final answer format found in response")
+                    answer = message_content
+            
+            print(f"Extracted final answer: {answer}")
+            return answer
+            
+        # Fallback if no valid answer found
+        print("Warning: Could not extract answer from agent response")
+        return "No valid answer could be extracted from agent response"
+    
 if __name__ == "__main__":
-    react_graph = build_graph()
-    # Show the butler's thought process
+
+    question_text = "How many studio albums were published by Mercedes Sosa between 2000 and 2009 (included)? You can use the latest 2022 version of english wikipedia."
+    question_text = "What country had the least number of athletes at the 1928 Summer Olympics? If there's a tie for a number of athletes, return the first in alphabetical order. Give the IOC country code as your answer."
+
+    agent = WalidzAgent()
+    # Prepare input state
+    human_msg = HumanMessage(content=question_text)
+    input_state = {
+        "messages": [human_msg],
+        "question_text": question_text
+    }
+
+    agent(input_state)
+
+
     # Save the graph visualization to disk
     #graph_image = react_graph.get_graph(xray=True).draw_mermaid_png()
     #with open("graph_visualization.png", "wb") as f:
@@ -224,10 +317,10 @@ if __name__ == "__main__":
     #display(Image(graph_image))
 
 
-    question_sample = "How many studio albums were published by Mercedes Sosa between 2000 and 2009 (included)? You can use the latest 2022 version of english wikipedia."
     #question_sample = question_sample.encode('utf-8', errors='replace').decode('utf-8')
 
-    print("Running with sample question:", question_sample)
+
+    """print("Running with sample question:", question_sample)
 
     try:
         # Create message with proper encoding
@@ -249,5 +342,5 @@ if __name__ == "__main__":
     except Exception as e:
         print(f"Error in Agent execution: {str(e)}")
         import traceback
-        traceback.print_exc()
+        traceback.print_exc()"""
 

requirements.txt

@@ -10,3 +10,6 @@ langgraph
 huggingface_hub
 wikipedia
 langchain_openai
+youtube_search
+youtube-transcript-api
+pytube