app.py

import os
import gradio as gr
import requests
import inspect
import pandas as pd
import json

from typing import List, TypedDict, Annotated, Optional
from langchain_openai import ChatOpenAI
from langgraph.graph import START, END, StateGraph
from langchain_core.messages import AnyMessage, SystemMessage, HumanMessage, AIMessage
from langgraph.graph.message import add_messages
from langgraph.prebuilt import ToolNode, tools_condition
from langchain_tavily import TavilySearch
from tools import (
    download_task_file,
    read_attached_text_file,
    answer_image_question,
    answer_excel_question,
    answer_python_question,
    answer_audio_question,
    get_youtube_transcript,
    answer_youtube_video_question,
    fetch_webpage_text,
    web_search_text,
    wikipedia_api_search,
)
from agent_helpers import (
    build_user_content,
    classify_attachment,
    cleanup_exact_answer,
    is_youtube_question,
    is_youtube_visual_question,
)
from programmatic_solvers import try_programmatic_answer
# (Keep Constants as is)
# --- Constants ---
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"

# --- Basic Agent Definition ---
# ----- THIS IS WERE YOU CAN BUILD WHAT YOU WANT ------

class AgentState(TypedDict, total=False):
    messages: Annotated[list[AnyMessage], add_messages]
    verification_status: str
    verification_notes: str
    verify_retries: int
    
class BasicAgent:
    def __init__(self):
        print("BasicAgent initialized.")

        self.MODEL_NAME = os.getenv("DASHSCOPE_AGENT_MODEL", "qwen3.5-flash")
        # export DASHSCOPE_API_KEY="sk-**"
        self.api_key = os.getenv("DASHSCOPE_API_KEY")
        if not self.api_key:
            raise RuntimeError("NO DASHSCOPE_API_KEY")
        
        self.model = ChatOpenAI(
            model=self.MODEL_NAME,
            api_key=self.api_key,
            base_url="https://dashscope.aliyuncs.com/compatible-mode/v1",
            temperature=0,
            timeout=45,
            max_retries=2,
            max_tokens=1024,
        )
        self.tools = [
            download_task_file,
            read_attached_text_file,
            answer_image_question,
            answer_excel_question,
            answer_python_question,
            answer_audio_question,
            get_youtube_transcript,
            answer_youtube_video_question,
            fetch_webpage_text,
            web_search_text,
            wikipedia_api_search,
            TavilySearch(
                max_results=5,
                topic="general",
                search_depth="basic",
            ),
        ]

        self.chat_with_tools = self.model.bind_tools(self.tools)
        # The graph
        builder = StateGraph(AgentState)
        
        # Define nodes: these do the work
        builder.add_node("assistant", self.assistant)
        builder.add_node("tools", ToolNode(self.tools))
        builder.add_node("verify_answer", self.verify_answer)
        builder.add_node("retry_with_feedback", self.retry_with_feedback)
        builder.add_node("final_process", self.clean_answer)
        
        # Define edges: these determine how the control flow moves
        builder.add_edge(START, "assistant")
        
        builder.add_conditional_edges(
            "assistant",
            self.route_after_assistant,
            {
                "tools": "tools",
                "verify_answer": "verify_answer",
            },
        )
        
        builder.add_edge("tools", "assistant")
        
        builder.add_conditional_edges(
            "verify_answer",
            self.route_after_verify,
            {
                "retry": "retry_with_feedback",
                "final_process": "final_process",
            },
        )
        
        builder.add_edge("retry_with_feedback", "assistant")
        builder.add_edge("final_process", END)
        
        self.react_graph = builder.compile()

    def __call__(self, question: str, task_id: str | None = None) -> str:
        print(f"Agent received question: {question[:200]}", flush=True)

        try:

            answer = self.answer_question(question, task_id)
            answer = str(answer).strip()
            print(f"Agent answer: {answer[:200]}", flush=True)
            return answer

        except Exception as e:
            print(f"Agent error: {e}", flush=True)
            return ""
            
    def assistant(self, state: AgentState):
        sys_msg = SystemMessage(content="""
        You are a precise question-answering agent.
        
        Use tools aggressively when the question involves:
        - attached files
        - images or screenshots
        - spreadsheets
        - Python code files
        - YouTube videos
        - web lookup
        
        Tool policy:
        - If a task_id is present and the question hints at an attachment, call download_task_file first.
        - For image, screenshot, chess position, chart image, diagram, or visual counting questions, use answer_image_question.
        - For Excel or CSV questions, use answer_excel_question.
        - For Python-code-output questions, use answer_python_question.
        - For plain text attachments, use read_attached_text_file.
        - For YouTube visual questions about what appears on camera, use answer_youtube_video_question.
        - For YouTube speech/transcript questions, use get_youtube_transcript.
        - For current or external factual lookup, use web search or web_search_text.
        - If Tavily is unavailable, use web_search_text.
        - When search gives a promising URL, use fetch_webpage_text to read the source.
        - For Wikipedia-focused questions, use wikipedia_api_search.
        
        Answer directly and concisely.
        Return only the final answer unless explanation is necessary.
        """)
        return {
            "messages": [self.chat_with_tools.invoke([sys_msg] + state["messages"])],
        }
    def _safe_parse_json(self, text: str) -> dict:
        """
        Parse JSON from model output safely.
        The model may sometimes wrap JSON with extra text.
        """
        text = text.strip()
    
        try:
            return json.loads(text)
        except Exception:
            pass
    
        start = text.find("{")
        end = text.rfind("}")
    
        if start != -1 and end != -1 and end > start:
            try:
                return json.loads(text[start:end + 1])
            except Exception:
                pass
    
        return {
            "verdict": "PASS",
            "revised_answer": text,
            "issue": "JSON parse failed, using raw verifier output.",
        }


    def verify_answer(self, state: AgentState):
        messages = state["messages"]
    
        question = messages[0].content
        candidate_answer = messages[-1].content
        retry_count = state.get("verify_retries", 0)
    
        # 只拿最近一部分上下文，避免 prompt 过长
        recent_context = []
        for m in messages[-8:]:
            role = m.__class__.__name__
            content = getattr(m, "content", "")
            recent_context.append(f"{role}:\n{content}")
    
        context_text = "\n\n---\n\n".join(recent_context)
    
        response = self.model.invoke([
                SystemMessage(content="""
        You are a strict answer verifier for an exact-match QA benchmark.
        
        Your job:
        1. Check whether the candidate answer satisfies the question.
        2. Check whether the answer format is exactly what the question asks.
        3. Use the available conversation/tool evidence only.
        4. Do not introduce unsupported new facts.
        5. If the answer is probably correct but badly formatted, revise the format.
        6. If the answer is unsupported, clearly wrong, says file/audio/image is unavailable, or ignores a required source/date constraint, request a retry.
        
        Return only valid JSON with this schema:
        {
          "verdict": "PASS" or "RETRY",
          "revised_answer": "final answer if PASS, otherwise empty string",
          "issue": "short reason"
        }
        
        Important exact-match formatting rules:
        - Remove explanations, markdown, citations, and extra punctuation.
        - If the question asks for a number, return only the number.
        - If the question asks for a name part, return only that part.
        - If the question asks comma-separated values, use comma + space.
        - If the question says without abbreviations, expand abbreviations.
        - If the candidate says no file/image/audio is available but the question has a task_id or attachment, use RETRY.
        - If the candidate ignores a date/source constraint, use RETRY.
        """),
                HumanMessage(content=f"""
        Original question:
        {question}
        
        Candidate answer:
        {candidate_answer}
        
        Recent conversation and tool context:
        {context_text}
        """)
            ])
    
        data = self._safe_parse_json(response.content)
    
        verdict = str(data.get("verdict", "PASS")).upper().strip()
        revised_answer = str(data.get("revised_answer", candidate_answer)).strip()
        issue = str(data.get("issue", "")).strip()
    
        max_verify_retries = 1
    
        if verdict == "RETRY" and retry_count < max_verify_retries:
            return {
                "verification_status": "retry",
                "verification_notes": issue,
                "verify_retries": retry_count + 1,
            }
    
        # 如果 verifier 通过，或者已经重试过一次还不行，就进入最终清洗
        final_candidate = revised_answer if revised_answer else candidate_answer
    
        return {
            "messages": [AIMessage(content=final_candidate)],
            "verification_status": "pass",
            "verification_notes": issue,
            "verify_retries": retry_count,
        }
    
    
    def route_after_verify(self, state: AgentState):
        if state.get("verification_status") == "retry":
            return "retry"
    
        return "final_process"
    
    def route_after_assistant(self, state: AgentState):
        last_message = state["messages"][-1]
    
        if getattr(last_message, "tool_calls", None):
            return "tools"
    
        return "verify_answer"
    
    def retry_with_feedback(self, state: AgentState):
        notes = state.get("verification_notes", "")
    
        return {
            "messages": [
                HumanMessage(content=f"""
    Your previous answer failed verification.
    
    Verifier notes:
    {notes}
    
    Please answer the original question again.
    Use tools if needed.
    Pay attention to source/date constraints and exact output format.
    Return only the final answer.
    """.strip())
            ]
        }
    
    def clean_answer(self, state: AgentState):
        messages = state["messages"]
    
        question = messages[0].content
        raw_answer = messages[-1].content
    
        response = self.model.invoke([
                SystemMessage(content="""
        You are an exact-match answer formatter.
        
        Rules:
        - Return only the final answer.
        - No explanation.
        - No markdown.
        - No citations.
        - No extra punctuation.
        - If the question asks for a number, return only the number.
        - If the question asks for USD with two decimals, return only the number with two decimals unless $ is explicitly requested.
        - If the question asks comma-separated values, use ", " between items.
        - If the question asks for a first name, surname, city, country code, or algebraic notation, return only that.
        - If the question says "without abbreviations", expand abbreviations.
        - Preserve required capitalization when obvious.
        """),
                HumanMessage(content=f"Question:\n{question}\n\nRaw answer:\n{raw_answer}")
            ])
    
        final_answer = response.content.strip()
    
        return {
            "messages": [AIMessage(content=final_answer)]
        }
    def format_final_answer(self, question: str, raw_answer: str) -> str:
        response = self.model.invoke([
                SystemMessage(content="""
        You are an exact-match answer formatter.
        Return only the final answer.
        No explanation.
        No markdown.
        No citations.
        No extra punctuation.
        Follow the requested format exactly.
        If the question asks for a number, return only the number.
        If the question asks for USD with two decimals, return only the number with two decimals.
        If the question asks comma-separated values, use ", " between items.
        If the question asks for algebraic notation, return only the move.
        If the question says without abbreviations, expand abbreviations.
        """),
                HumanMessage(content=f"Question:\n{question}\n\nRaw answer:\n{raw_answer}")
            ])
        return response.content.strip()

    def answer_from_context(self, question: str, context: str, source_label: str = "context") -> str:
        response = self.model.invoke([
                SystemMessage(content="""
        You are an exact-match QA extractor.

        Answer the question using only the provided source context.

        Rules:
        - Return only the final answer.
        - No explanation.
        - No markdown.
        - No citations.
        - Do not mention the source context.
        - If the question asks for a number, return only the number.
        - If the question asks for a list, return only the requested items.
        - If comma-separated output is appropriate, use comma + space.
        - If the answer is not present, return the best concise answer implied by the context.
        """),
                HumanMessage(content=f"""
        Question:
        {question}

        Source ({source_label}):
        {context}
        """)
            ])
        return cleanup_exact_answer(response.content)
        
    def answer_question(self, question: str, task_id: str | None = None) -> str:
        file_info = None

        programmatic_answer = try_programmatic_answer(question)
        if programmatic_answer is not None:
            return cleanup_exact_answer(programmatic_answer)

        if is_youtube_question(question):
            if is_youtube_visual_question(question):
                visual_answer = answer_youtube_video_question.invoke({
                    "url_or_question": question,
                    "question": question,
                })
                if not str(visual_answer).lower().startswith("failed"):
                    return cleanup_exact_answer(self.format_final_answer(question, visual_answer))

            transcript = get_youtube_transcript.invoke({"url_or_question": question})
            return self.answer_from_context(question, transcript, "YouTube transcript")
    
        if task_id:
            info_str = download_task_file.invoke({"task_id": task_id})
            print(f"[file_info] {info_str}", flush=True)
            try:
                file_info = json.loads(info_str)
            except Exception:
                file_info = None
    
        if file_info and "file_path" in file_info:
            suffix = file_info.get("suffix", "").lower()
            file_path = file_info["file_path"]

            attachment_kind = classify_attachment(question, suffix)

            if attachment_kind == "image":
                raw = answer_image_question.invoke({
                    "file_path": file_path,
                    "question": question
                })
                return cleanup_exact_answer(self.format_final_answer(question, raw))
    
            if attachment_kind == "audio":
                raw = answer_audio_question.invoke({
                    "file_path": file_path,
                    "question": question
                })
                return cleanup_exact_answer(self.format_final_answer(question, raw))
    
            if attachment_kind == "python":
                raw = answer_python_question.invoke({
                    "file_path": file_path
                })
                return cleanup_exact_answer(self.format_final_answer(question, raw))
    
            if attachment_kind == "spreadsheet":
                context = answer_excel_question.invoke({
                    "file_path": file_path,
                    "question": question
                })
                return self.answer_from_context(question, context, "spreadsheet summary")

            if attachment_kind == "text":
                context = read_attached_text_file.invoke({
                    "file_path": file_path,
                    "max_chars": 20000,
                })
                return self.answer_from_context(question, context, "attached text file")
    
        user_content = build_user_content(question, task_id)
        result = self.react_graph.invoke({"messages": [HumanMessage(content=user_content)]})
        return cleanup_exact_answer(result["messages"][-1].content)


def run_and_submit_all( profile: gr.OAuthProfile | None):
    """
    Fetches all questions, runs the BasicAgent 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

    if profile:
        username= f"{profile.username}"
        print(f"User logged in: {username}")
    else:
        print("User not logged in.")
        return "Please Login to Hugging Face with the button.", None

    api_url = DEFAULT_API_URL
    questions_url = f"{api_url}/questions"
    submit_url = f"{api_url}/submit"

    # 1. Instantiate Agent ( modify this part to create your agent)
    try:
        agent = BasicAgent()
    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(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...")
    for item in questions_data:
        task_id = item.get("task_id")
        question_text = item.get("question")
        if not task_id or question_text is None:
            print(f"Skipping item with missing task_id or question: {item}")
            continue
        try:
            submitted_answer = agent(question_text, task_id)
            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}"})

    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)
        response.raise_for_status()
        result_data = response.json()
        final_status = (
            f"Submission Successful!\n"
            f"User: {result_data.get('username')}\n"
            f"Overall Score: {result_data.get('score', 'N/A')}% "
            f"({result_data.get('correct_count', '?')}/{result_data.get('total_attempted', '?')} correct)\n"
            f"Message: {result_data.get('message', 'No message received.')}"
        )
        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)
        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")
    gr.Markdown(
        """
        **Instructions:**

        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(
        fn=run_and_submit_all,
        outputs=[status_output, results_table]
    )

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

    if space_host_startup:
        print(f"✅ SPACE_HOST found: {space_host_startup}")
        print(f"   Runtime URL should be: https://{space_host_startup}.hf.space")
    else:
        print("ℹ️  SPACE_HOST environment variable not found (running locally?).")

    if space_id_startup: # Print repo URLs if SPACE_ID is found
        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")
    else:
        print("ℹ️  SPACE_ID environment variable not found (running locally?). Repo URL cannot be determined.")

    print("-"*(60 + len(" App Starting ")) + "\n")

    print("Launching Gradio Interface for Basic Agent Evaluation...")
    demo.launch(debug=True, share=False)