Update app.py

app.py

@@ -1,262 +1,171 @@
-# app.py
+# app.py (New LangChain version)
 import os
 import gradio as gr
 import pandas as pd
-from bs4 import BeautifulSoup # Keep this if your tools use it
+from bs4 import BeautifulSoup
 import datetime
 import pytz
 import math
 import re
 import requests
-from transformers import HfAgent # Your successful import
-from transformers.tools import Tool # Your successful import
-from transformers import pipeline # <<< --- MAKE SURE THIS IMPORT IS ADDED / PRESENT
 import traceback
-
 import sys
-print(f"--- Python version: {sys.version} ---")
-# print(f"--- Python sys.path (module search paths): {sys.path} ---") # Optional now
-
-import transformers
-from transformers.tools import Tool
-print(f"--- Expected Transformers Version: 4.36.0 ---")
-print(f"--- Actual Transformers Version: {transformers.__version__} ---")
-# print(f"--- Transformers module loaded from: {transformers.__file__} ---") # Optional now
-# print(f"--- Attributes of 'transformers' module (dir(transformers)): {dir(transformers)} ---") # Optional now
-
-try:
-    from transformers import HfAgent # <<< --- THE CORRECT IMPORT!
-    print("--- Successfully imported HfAgent directly from transformers! ---")
-except ImportError as e:
-    print(f"--- FAILED to import HfAgent directly from transformers: {e} ---")
-    # This should ideally not happen now
-    raise
-except Exception as e_gen:
-    print(f"--- Some other UNEXPECTED error during HfAgent import: {e_gen} ---")
-    raise
-
-print("--- If no errors above, imports were successful. Proceeding with rest of app. ---")
-
-# (Keep Constants as is)
+
+# --- LangChain and new Transformers imports ---
+from langchain.agents import AgentExecutor, create_react_agent
+from langchain_huggingface import HuggingFacePipeline
+from langchain_core.prompts import PromptTemplate
+from langchain.tools import Tool
+from langchain_community.tools import DuckDuckGoSearchRun
+
+# --- Other imports ---
+import transformers # Still useful for version checking
+print(f"--- Using transformers version: {transformers.__version__} ---")
+
 # --- Constants ---
 DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
 
-# --- Tool Definitions ---
-def get_current_time_in_timezone(timezone: str) -> str:
-    """Fetches the current local time in a specified IANA timezone (e.g., 'America/New_York', 'Europe/London', 'UTC').
-    Args:
-        timezone (str): A string representing a valid IANA timezone name.
-    """
+
+# --- Tool Definitions (LangChain Style) ---
+# For LangChain, we define the functions and then wrap them in LangChain's Tool class.
+
+def get_current_time_in_timezone_func(timezone: str) -> str:
+    """A tool that fetches the current local time in a specified IANA timezone. Always use this tool for questions about the current time. Input should be a valid timezone string (e.g., 'America/New_York', 'Europe/London')."""
     print(f"--- Tool: Executing get_current_time_in_timezone for: {timezone} ---")
     try:
         tz = pytz.timezone(timezone)
-        # Added %Z (timezone name) and %z (UTC offset)
         local_time = datetime.datetime.now(tz).strftime("%Y-%m-%d %H:%M:%S %Z%z")
         return f"The current local time in {timezone} is: {local_time}"
     except pytz.exceptions.UnknownTimeZoneError:
-        print(f"Error: Unknown timezone '{timezone}'")
-        return f"Error: Unknown timezone '{timezone}'. Please use a valid IANA timezone name (e.g., 'America/Denver', 'UTC')."
+        return f"Error: Unknown timezone '{timezone}'. Please use a valid IANA timezone name."
     except Exception as e:
-        print(f"Error fetching time for timezone '{timezone}': {str(e)}")
         return f"Error fetching time for timezone '{timezone}': {str(e)}"
 
-def web_search(query: str) -> str:
-    """
-    Performs a web search using DuckDuckGo (via HTML scraping) and returns the text content of the top result snippets.
-    Use this tool to find up-to-date information about events, facts, or topics when the answer isn't already known.
-
-    Args:
-        query (str): The search query string.
-
-    Returns:
-        str: A string containing the summarized search results (titles and snippets of top hits), or an error message if the search fails.
-    """
-    print(f"--- Tool: Executing web_search with query: {query} ---")
-    try:
-        search_url = "https://html.duckduckgo.com/html/"
-        params = {"q": query}
-        headers = {'User-Agent': 'Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/100.0.4896.127 Safari/537.36'} # Common user agent
-
-        response = requests.post(search_url, data=params, headers=headers, timeout=15) # Increased timeout
-        response.raise_for_status() # Check for HTTP errors (4xx or 5xx)
-
-        soup = BeautifulSoup(response.text, 'html.parser')
-        results = soup.find_all('div', class_='result__body') # Find result containers
-
-        snippets = []
-        for i, result in enumerate(results[:3]): # Get top 3 results for brevity
-            title_tag = result.find('a', class_='result__a')
-            snippet_tag = result.find('a', class_='result__snippet')
-            title = title_tag.get_text(strip=True) if title_tag else "No Title"
-            snippet = snippet_tag.get_text(strip=True) if snippet_tag else "No Snippet"
-            if snippet != "No Snippet": # Only include results with a snippet
-                snippets.append(f"Result {i+1}: {title} - {snippet}")
-
-        if not snippets:
-            return "No search results with relevant snippets found."
-
-        return "\n".join(snippets)
-
-    except requests.exceptions.Timeout:
-        print(f"Error during web search request: Timeout")
-        return "Error: The web search request timed out."
-    except requests.exceptions.RequestException as e:
-        print(f"Error during web search request: {e}")
-        return f"Error: Could not perform web search. Network issue: {e}"
-    except Exception as e:
-        print(f"Error processing web search results: {e}")
-        return f"Error: Could not process search results. {e}"
-
-def safe_calculator(expression: str) -> str:
-    """
-    Evaluates a simple mathematical expression involving numbers, +, -, *, /, %, parentheses, and the math functions: sqrt, pow.
-    Use this tool *only* for calculations. Do not use it to run other code.
-
-    Args:
-        expression (str): The mathematical expression string (e.g., "(2 + 3) * 4", "pow(2, 5)", "sqrt(16)").
+# Using the DuckDuckGoSearchRun tool from LangChain for stability
+# The description is very important for the agent to know when to use it.
+search_tool = DuckDuckGoSearchRun(
+    name="web_search",
+    description="A tool that performs a web search using DuckDuckGo. Use this to find up-to-date information about events, facts, or topics when the answer isn't already known."
+)
 
-    Returns:
-        str: The numerical result of the calculation or a descriptive error message if the expression is invalid or unsafe.
-    """
+def safe_calculator_func(expression: str) -> str:
+    """A tool for evaluating simple mathematical expressions. Use this tool *only* for calculations involving numbers, +, -, *, /, %, parentheses, and the math functions: sqrt, pow. Do not use it to run other code."""
     print(f"--- Tool: Executing safe_calculator with expression: {expression} ---")
     try:
-        # Basic check for allowed characters/patterns first
-        # Allows numbers (including scientific notation), operators, parentheses, whitespace, and known function names
-        pattern = r"^[0-9eE\.\+\-\*\/\%\(\)\s]*(sqrt|pow)?[0-9eE\.\+\-\*\/\%\(\)\s\,]*$"
-        if not re.match(pattern, expression):
-             # Fallback simple pattern check (less precise)
-             allowed_chars_pattern = r"^[0-9eE\.\+\-\*\/\%\(\)\s\,sqrtpow]+$"
-             if not re.match(allowed_chars_pattern, expression):
-                raise ValueError(f"Expression '{expression}' contains disallowed characters.")
-
-        # Define allowed functions/names for eval's context
-        allowed_names = {
-            "sqrt": math.sqrt,
-            "pow": math.pow,
-            # Add other safe math functions if needed e.g. "log": math.log
-        }
-        # Evaluate the expression in a restricted environment
-        # Limited builtins, only allowed names are accessible.
+        # Using a more restricted eval context for safety
+        allowed_names = {"sqrt": math.sqrt, "pow": math.pow, "pi": math.pi}
         result = eval(expression, {"__builtins__": {}}, allowed_names)
-
-        # Ensure the result is a number before converting to string
-        if not isinstance(result, (int, float)):
-             raise ValueError("Calculation did not produce a numerical result.")
-
         return str(result)
     except Exception as e:
-        # Catch potential errors during eval (SyntaxError, NameError, TypeError etc.) or from the checks
         print(f"Error during calculation for '{expression}': {e}")
         return f"Error calculating '{expression}': Invalid expression or calculation error ({e})."
 
-# --- Custom Agent to Force Correct Behavior ---
-class MyCustomHfAgent(HfAgent):
-    """
-    A custom agent that inherits from HfAgent to override the text generation method,
-    forcing it to use the local pipeline instead of attempting a faulty web request.
-    """
-    def generate_one(self, prompt: str, stop: list):
-        print("--- INSIDE CUSTOM HfAgent's generate_one method ---")
-
-        # This is the crucial check. We're logging what the agent thinks its state is.
-        is_pipeline = self.llm.is_hf_pipeline if hasattr(self.llm, "is_hf_pipeline") else "LLM has no is_hf_pipeline attr"
-        print(f"--> self.llm.is_hf_pipeline is: {is_pipeline}")
-
-        # Regardless of what the agent thinks, we KNOW we gave it a pipeline.
-        # So, we will force it to execute the code path for local pipelines.
-        print("--> Forcing execution of the local pipeline path...")
 
-        try:
-            # This is the code from the 'if self.llm.is_hf_pipeline:' block in the original Agent.generate_one method
-            processed_prompt = self.llm.processor.process_prompt(prompt, **self.tokenizer_kwargs)
-            model_outputs = self.llm.pipeline(processed_prompt, stop_sequence=stop, **self.generate_kwargs)
-            return self.llm.processor.process_outputs(model_outputs, stop_sequence=stop)
-        except Exception as e:
-            print(f"--- ERROR during forced pipeline execution: {e} ---")
-            traceback.print_exc()
-            # If this fails, we return an error string.
-            return f"Error during custom pipeline execution: {e}"
-
-# --- Agent Definition using HfAgent ---
-class HfAgentWrapper:
+# --- LangChain Agent Definition ---
+class LangChainAgentWrapper:
     def __init__(self):
-        print("Initializing HfAgentWrapper...")
-        model_id_or_path = "bigcode/starcoderbase-1b" # A model compatible with transformers v4.36.0
-
+        print("Initializing LangChainAgentWrapper...")
+        
+        # Using a newer, more capable instruction-tuned model.
+        # This model is generally better at following the ReAct prompt format used by LangChain agents.
+        model_id = "mistralai/Mistral-7B-Instruct-v0.1" 
+        # model_id = "bigcode/starcoderbase-1b" # You can still use starcoder if you prefer
+        
         try:
-            print(f"Strategy: Pre-creating pipeline for model: {model_id_or_path}")
-            hf_auth_token = os.getenv("HF_TOKEN") # Secret should be named HF_TOKEN
+            hf_auth_token = os.getenv("HF_TOKEN")
             if not hf_auth_token:
-                print("WARNING: HF_TOKEN secret not found. This may fail if model requires token.")
-                # Starcoderbase is gated, so this is needed.
-                raise ValueError("HF_TOKEN secret is missing and is required for this model.")
+                raise ValueError("HF_TOKEN secret is missing. It is required for downloading models.")
             else:
-                print(f"HF_TOKEN secret found (length: {len(hf_auth_token)}).")
-
-            # --- Step 1: Create the pipeline object FIRST ---
-            # This allows us to handle errors from pipeline creation directly.
-            llm_pipeline = pipeline(
-                task="text-generation",
-                model=model_id_or_path,
-                token=hf_auth_token
-                # trust_remote_code=True # Not generally needed for starcoder with this version
-            )
-            print("Successfully created LLM pipeline object.")
-
-            # --- Step 2: Ensure your tools are created WITH proper names ---
-            if not get_current_time_in_timezone.__doc__: raise ValueError("Tool 'get_current_time_in_timezone' is missing a docstring.")
-            if not web_search.__doc__: raise ValueError("Tool 'web_search' is missing a docstring.")
-            if not safe_calculator.__doc__: raise ValueError("Tool 'safe_calculator' is missing a docstring.")
-
-            time_tool_obj = Tool(
-                name=get_current_time_in_timezone.__name__, # Use the function's name
-                func=get_current_time_in_timezone,
-                description=get_current_time_in_timezone.__doc__
-            )
-            search_tool_obj = Tool(
-                name=web_search.__name__, # Use the function's name
-                func=web_search,
-                description=web_search.__doc__
+                print("HF_TOKEN secret found.")
+
+            # Create the Hugging Face pipeline
+            print(f"Loading model pipeline for: {model_id}")
+            llm_pipeline = transformers.pipeline(
+                "text-generation",
+                model=model_id,
+                model_kwargs={"torch_dtype": "auto"}, # Use "auto" for dtype
+                device_map="auto", # Requires accelerate
+                token=hf_auth_token,
             )
-            calculator_tool_obj = Tool(
-                name=safe_calculator.__name__, # Use the function's name
-                func=safe_calculator,
-                description=safe_calculator.__doc__
+            print("Model pipeline loaded successfully.")
+            
+            # Wrap the pipeline in a LangChain LLM object
+            self.llm = HuggingFacePipeline(pipeline=llm_pipeline)
+
+            # Define the list of LangChain tools
+            self.tools = [
+                Tool(
+                    name="get_current_time_in_timezone",
+                    func=get_current_time_in_timezone_func,
+                    description=get_current_time_in_timezone_func.__doc__
+                ),
+                search_tool, # This is already a LangChain Tool instance
+                Tool(
+                    name="safe_calculator",
+                    func=safe_calculator_func,
+                    description=safe_calculator_func.__doc__
+                ),
+            ]
+            print(f"Tools prepared for agent: {[tool.name for tool in self.tools]}")
+            
+            # Create the ReAct agent prompt from a template
+            # The prompt is crucial for teaching the agent how to think and use tools.
+            react_prompt = PromptTemplate.from_template(
+                """
+                You are a helpful assistant. Answer the following questions as best you can.
+                You have access to the following tools:
+
+                {tools}
+
+                Use the following format:
+
+                Question: the input question you must answer
+                Thought: you should always think about what to do
+                Action: the action to take, should be one of [{tool_names}]
+                Action Input: the input to the action
+                Observation: the result of the action
+                ... (this Thought/Action/Action Input/Observation can repeat N times)
+                Thought: I now know the final answer
+                Final Answer: the final answer to the original input question
+
+                Begin!
+
+                Question: {input}
+                Thought:{agent_scratchpad}
+                """
             )
-            self.actual_tools_for_agent = [time_tool_obj, search_tool_obj, calculator_tool_obj]
-            print(f"Prepared Tool objects with names: {[tool.name for tool in self.actual_tools_for_agent]}")
-
-            # --- Step 3: Pass the PRE-INITIALIZED pipeline object to HfAgent ---
-            print("Initializing HfAgent with the pre-created pipeline...")
-            self.agent = MyCustomHfAgent(llm_pipeline, additional_tools=self.actual_tools_for_agent)
-            print("HfAgent successfully instantiated with pre-initialized pipeline.")
+            
+            # Create the agent
+            agent = create_react_agent(self.llm, self.tools, react_prompt)
+            
+            # Create the agent executor, which runs the agent loop
+            self.agent_executor = AgentExecutor(agent=agent, tools=self.tools, verbose=True, handle_parsing_errors=True)
+            print("LangChain agent created successfully.")
 
         except Exception as e:
-            print(f"CRITICAL ERROR: Failed to initialize HfAgent or Pipeline: {e}")
-            print("Full traceback of HfAgent/Pipeline initialization error:")
+            print(f"CRITICAL ERROR: Failed to initialize LangChain agent: {e}")
             traceback.print_exc() 
-            raise RuntimeError(f"HfAgent/Pipeline initialization failed: {e}") from e
+            raise RuntimeError(f"LangChain agent initialization failed: {e}") from e
 
-    # The __call__ method remains the same
     def __call__(self, question: str) -> str:
-        print(f"\n--- HfAgentWrapper received question (first 100 chars): {question[:100]}... ---")
+        print(f"\n--- LangChainAgentWrapper received question: {question[:100]}... ---")
         try:
-            answer = self.agent.run(question)
-            print(f"--- HfAgentWrapper generated answer (first 100 chars): {str(answer)[:100]}... ---")
-            return str(answer)
+            # Invoke the agent executor
+            response = self.agent_executor.invoke({"input": question})
+            # The answer is in the 'output' key of the response dictionary
+            return response.get("output", "No output found.")
         except Exception as e:
-            print(f"ERROR: HfAgent execution failed for question '{question[:50]}...': {e}")
-            print("Full traceback of HfAgent execution error:")
+            print(f"ERROR: LangChain agent execution failed: {e}")
             traceback.print_exc()
             return f"Agent Error: Failed to process the question. Details: {e}"
 
-def run_and_submit_all( profile: gr.OAuthProfile | None):
+# --- Main Evaluation Logic ---
+def run_and_submit_all(profile: gr.OAuthProfile | None):
     """
-    Fetches all questions, runs the BasicAgent on them, submits all answers,
+    Fetches all questions, runs the agent on them, submits all answers,
     and displays the results.
     """
-    # --- Determine HF Space Runtime URL and Repo URL ---
-    space_id = os.getenv("SPACE_ID") # Get the SPACE_ID for sending link to the code
+    space_id = os.getenv("SPACE_ID")
 
     if profile:
         username= f"{profile.username}"
@@ -269,38 +178,29 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
     questions_url = f"{api_url}/questions"
     submit_url = f"{api_url}/submit"
 
-    # 1. Instantiate Agent ( modify this part to create your agent)
     try:
-        agent = HfAgentWrapper()
+        # Now instantiate our new LangChain agent
+        agent = LangChainAgentWrapper()
     except Exception as e:
         print(f"Error instantiating agent: {e}")
         return f"Error initializing agent: {e}", None
-    # In the case of an app running as a hugging Face space, this link points toward your codebase ( usefull for others so please keep it public)
+    
     agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
     print(agent_code)
 
-    # 2. Fetch Questions
     print(f"Fetching questions from: {questions_url}")
     try:
         response = requests.get(questions_url, timeout=15)
         response.raise_for_status()
         questions_data = response.json()
         if not questions_data:
-             print("Fetched questions list is empty.")
-             return "Fetched questions list is empty or invalid format.", None
+            print("Fetched questions list is empty.")
+            return "Fetched questions list is empty or invalid format.", None
         print(f"Fetched {len(questions_data)} questions.")
-    except requests.exceptions.RequestException as e:
-        print(f"Error fetching questions: {e}")
-        return f"Error fetching questions: {e}", None
-    except requests.exceptions.JSONDecodeError as e:
-         print(f"Error decoding JSON response from questions endpoint: {e}")
-         print(f"Response text: {response.text[:500]}")
-         return f"Error decoding server response for questions: {e}", None
     except Exception as e:
         print(f"An unexpected error occurred fetching questions: {e}")
         return f"An unexpected error occurred fetching questions: {e}", None
 
-    # 3. Run your Agent
     results_log = []
     answers_payload = []
     print(f"Running agent on {len(questions_data)} questions...")
@@ -315,19 +215,17 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
             answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
             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}")
-             results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": f"AGENT ERROR: {e}"})
+            print(f"Error running agent on task {task_id}: {e}")
+            results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": f"AGENT ERROR: {e}"})
 
     if not answers_payload:
         print("Agent did not produce any answers to submit.")
         return "Agent did not produce any answers to submit.", pd.DataFrame(results_log)
 
-    # 4. Prepare Submission 
     submission_data = {"username": username.strip(), "agent_code": agent_code, "answers": answers_payload}
     status_update = f"Agent finished. Submitting {len(answers_payload)} answers for user '{username}'..."
     print(status_update)
 
-    # 5. Submit
     print(f"Submitting {len(answers_payload)} answers to: {submit_url}")
     try:
         response = requests.post(submit_url, json=submission_data, timeout=60)
@@ -343,34 +241,13 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
         print("Submission successful.")
         results_df = pd.DataFrame(results_log)
         return final_status, results_df
-    except requests.exceptions.HTTPError as e:
-        error_detail = f"Server responded with status {e.response.status_code}."
-        try:
-            error_json = e.response.json()
-            error_detail += f" Detail: {error_json.get('detail', e.response.text)}"
-        except requests.exceptions.JSONDecodeError:
-            error_detail += f" Response: {e.response.text[:500]}"
-        status_message = f"Submission Failed: {error_detail}"
-        print(status_message)
-        results_df = pd.DataFrame(results_log)
-        return status_message, results_df
-    except requests.exceptions.Timeout:
-        status_message = "Submission Failed: The request timed out."
-        print(status_message)
-        results_df = pd.DataFrame(results_log)
-        return status_message, results_df
-    except requests.exceptions.RequestException as e:
-        status_message = f"Submission Failed: Network error - {e}"
-        print(status_message)
-        results_df = pd.DataFrame(results_log)
-        return status_message, results_df
     except Exception as e:
         status_message = f"An unexpected error occurred during submission: {e}"
         print(status_message)
+        traceback.print_exc()
         results_df = pd.DataFrame(results_log)
         return status_message, results_df
 
-
 # --- Build Gradio Interface using Blocks ---
 with gr.Blocks() as demo:
     gr.Markdown("# Basic Agent Evaluation Runner")
@@ -381,20 +258,12 @@ with gr.Blocks() as demo:
         1.  Please clone this space, then modify the code to define your agent's logic, the tools, the necessary packages, etc ...
         2.  Log in to your Hugging Face account using the button below. This uses your HF username for submission.
         3.  Click 'Run Evaluation & Submit All Answers' to fetch questions, run your agent, submit answers, and see the score.
-
-        ---
-        **Disclaimers:**
-        Once clicking on the "submit button, it can take quite some time ( this is the time for the agent to go through all the questions).
-        This space provides a basic setup and is intentionally sub-optimal to encourage you to develop your own, more robust solution. For instance for the delay process of the submit button, a solution could be to cache the answers and submit in a seperate action or even to answer the questions in async.
         """
     )
 
     gr.LoginButton()
-
     run_button = gr.Button("Run Evaluation & Submit All Answers")
-
     status_output = gr.Textbox(label="Run Status / Submission Result", lines=5, interactive=False)
-    # Removed max_rows=10 from DataFrame constructor
     results_table = gr.DataFrame(label="Questions and Agent Answers", wrap=True)
 
     run_button.click(
@@ -404,9 +273,8 @@ with gr.Blocks() as demo:
 
 if __name__ == "__main__":
     print("\n" + "-"*30 + " App Starting " + "-"*30)
-    # Check for SPACE_HOST and SPACE_ID at startup for information
     space_host_startup = os.getenv("SPACE_HOST")
-    space_id_startup = os.getenv("SPACE_ID") # Get SPACE_ID at startup
+    space_id_startup = os.getenv("SPACE_ID")
 
     if space_host_startup:
         print(f"✅ SPACE_HOST found: {space_host_startup}")
@@ -414,7 +282,7 @@ if __name__ == "__main__":
     else:
         print("ℹ️  SPACE_HOST environment variable not found (running locally?).")
 
-    if space_id_startup: # Print repo URLs if SPACE_ID is found
+    if space_id_startup:
         print(f"✅ SPACE_ID found: {space_id_startup}")
         print(f"   Repo URL: https://huggingface.co/spaces/{space_id_startup}")
         print(f"   Repo Tree URL: https://huggingface.co/spaces/{space_id_startup}/tree/main")