import os
import gradio as gr
import requests
import ast
import json
import time
import pandas as pd
from datetime import datetime
from typing import List, Dict, Any, Annotated
from langgraph.graph import Graph, StateGraph
from typing_extensions import TypedDict
from openai import OpenAI
from tools import simple_search
import re
from huggingface_hub import InferenceClient
import io
import mimetypes
import base64
import cv2
import numpy as np
from io import BytesIO
import tempfile
import subprocess
import sys
import textwrap

# -------------------------
# Environment & constants
# -------------------------

DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
HF_TOKEN = os.getenv("HF_TOKEN")

# Initialize HF client
client = InferenceClient(token=HF_TOKEN)

# -------------------------
# Constants
# -------------------------

SYSTEM = (
    "You are a parser-safe assistant.\n"
    "Output **ONLY** the JSON object requested—no extra words."
)

# -------------------------
# Utility helpers
# -------------------------

def override(_, new):
    return new

def merge_dicts(old: Dict, new: Dict) -> Dict:
    """Merge two dictionaries, with *new* values taking precedence."""
    return {**old, **new}

def tighten(q: str) -> str:
    """
    Strip long GAIA questions down to quoted phrases and capitalised words.
    Falls back to the original text if we strip too much.
    """
    quoted = re.findall(r'"([^"]+)"', q)
    caps   = re.findall(r'\b([A-Z0-9][\w-]{2,})', q)
    short  = " ".join(quoted + caps)
    return short or q

# -------------------------
# Multimodal helpers
# -------------------------

def retry_hf_inference(func):
    """Decorator to retry HF Inference API calls with backoff."""
    def wrapper(*args, **kwargs):
        max_retries = 2
        base_delay = 7
        
        for attempt in range(max_retries + 1):
            try:
                return func(*args, **kwargs)
            except Exception as e:
                if attempt == max_retries:
                    raise
                delay = base_delay * (attempt + 1)
                print(f"HF API error: {str(e)}. Retrying in {delay}s...")
                time.sleep(delay)
    return wrapper

@retry_hf_inference
def image_qa_bytes(data: bytes, prompt: str) -> str:
    """Query LLaVA for image-based QA using bytes."""
    headers = {"Content-Type": "application/octet-stream"}
    return client.post("llava-hf/llava-v1.6-mistral-7b-hf", data=data, headers=headers)

@retry_hf_inference
def video_label_bytes(data: bytes) -> str:
    """Get video classification using VideoMAE-Base from bytes."""
    # Process video to get first 8 seconds, 16 frames
    
    # Read video from bytes
    video_bytes = BytesIO(data)
    cap = cv2.VideoCapture()
    cap.open(video_bytes)
    
    # Get video properties
    fps = cap.get(cv2.CAP_PROP_FPS)
    frame_count = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))
    
    # Calculate frames to extract (16 frames over 8 seconds)
    target_frames = 16
    target_duration = 8  # seconds
    frame_interval = max(1, int(frame_count / (fps * target_duration)))
    
    frames = []
    frame_idx = 0
    
    while len(frames) < target_frames and frame_idx < frame_count:
        ret, frame = cap.read()
        if not ret:
            break
            
        if frame_idx % frame_interval == 0:
            # Resize frame to match VideoMAE's expected input
            frame = cv2.resize(frame, (224, 224))
            frames.append(frame)
            
        frame_idx += 1
    
    cap.release()
    
    # If we don't have enough frames, duplicate the last frame
    while len(frames) < target_frames:
        frames.append(frames[-1])
    
    # Stack frames and convert to bytes
    video_array = np.stack(frames)
    _, buffer = cv2.imencode('.mp4', video_array)
    processed_bytes = buffer.tobytes()
    
    # Send to VideoMAE
    headers = {"Content-Type": "application/octet-stream"}
    preds = client.post(
        "MCG-NJU/videomae-base-finetuned-ucf101", 
        data=processed_bytes,
        headers=headers
    )
    return sorted(preds, key=lambda x: x["score"], reverse=True)[0]["label"]

def sheet_answer_bytes(data: bytes, question: str) -> str:
    """Process spreadsheet data from bytes and return numeric answer."""
    if mimetypes.guess_type("x.xlsx")[0] == "text/csv" or question.endswith(".csv"):
        df = pd.read_csv(io.BytesIO(data))
    else:
        df = pd.read_excel(io.BytesIO(data))
    
    # Calculate total sales for Food category
    total = df[df["Category"] == "Food"]["Sales"].sum()
    return f"{total:.2f}"

# -------------------------
# Code Analysis helpers
# -------------------------

def run_python(code: str) -> str:
    """Quick & dirty evaluator for Python code."""
    with tempfile.NamedTemporaryFile("w+", suffix=".py", delete=False) as f:
        f.write(textwrap.dedent(code))
        f.flush()
        out = subprocess.check_output([sys.executable, f.name], timeout=10)
    return out.decode().strip()

# -------------------------
# State definition
# -------------------------

class AgentState(TypedDict):
    question: Annotated[str, override]
    current_step: Annotated[str, override]
    final_answer: Annotated[str, override]
    history: Annotated[List[Dict[str, str]], list.__add__]
    needs_search: Annotated[bool, override]
    search_query: Annotated[str, override]
    task_id: Annotated[str, override]
    logs: Annotated[Dict[str, Any], merge_dicts]
    file_url: Annotated[str, override]
    code_blocks: Annotated[List[Dict[str, str]], list.__add__]

# -------------------------
# BasicAgent implementation
# -------------------------

class BasicAgent:
    def __init__(self):
        if not OPENAI_API_KEY:
            raise EnvironmentError("OPENAI_API_KEY not set")
        self.llm = OpenAI(api_key=OPENAI_API_KEY)
        self.workflow = self._build_workflow()

    def _call_llm(self, prompt: str, max_tokens: int = 256) -> str:
        try:
            resp = self.llm.chat.completions.create(
                model="gpt-4.1",
                messages=[
                    {"role": "system", "content": SYSTEM},
                    {"role": "user", "content": prompt},
                ],
                temperature=0.3,
                max_tokens=max_tokens,
            )
            return resp.choices[0].message.content.strip()
        except Exception as e:
            print(f"\nLLM Error: {str(e)}")
            raise

    def _safe_parse(self, raw: str) -> dict:
        """Fallback parser for when JSON parsing fails."""
        try:
            # Try to extract a dict-like structure
            match = re.search(r'\{.*\}', raw, re.DOTALL)
            if match:
                return ast.literal_eval(match.group(0))
        except:
            pass
        return {"needs_search": True, "search_query": ""}

    def _analyze_question(self, state: AgentState) -> AgentState:
        # Check for file attachments
        if state["file_url"]:
            file_type = self._detect_file_type(state["file_url"])
            if file_type == "video":
                state["current_step"] = "video"
            elif file_type == "image":
                state["current_step"] = "image"
            elif file_type in ["excel", "csv"]:
                state["current_step"] = "sheet"
            return state

        # Regular text question analysis
        prompt = (
            "Return ONLY valid JSON:\n"
            "{\"needs_search\": bool, \"search_query\": str}\n\n"
            f"Question: {state['question']}"
        )
        try:
            raw = self._call_llm(prompt)
            try:
                decision = json.loads(raw)
            except json.JSONDecodeError:
                print(f"JSON parse error, falling back to safe parse. Raw response: {raw}")
                decision = self._safe_parse(raw)
            
            state["needs_search"] = bool(decision.get("needs_search", False))
            state["search_query"] = decision.get("search_query", state["question"])
        except Exception as e:
            print(f"\nLLM Error in question analysis: {str(e)}")
            state["needs_search"] = True
            state["search_query"] = state["question"]
            
        state["current_step"] = "search" if state["needs_search"] else "answer"
        return state

    def _detect_file_type(self, url: str) -> str:
        """Detect file type from URL extension."""
        ext = url.split(".")[-1].lower()
        return {
            "mp4": "video",
            "jpg": "image",
            "jpeg": "image",
            "png": "image",
            "xlsx": "excel",
            "csv": "csv"
        }.get(ext, "unknown")

    def _image_node(self, state: AgentState) -> AgentState:
        """Handle image-based questions."""
        try:
            data = self._download_file(state["file_url"])
            answer = image_qa_bytes(data, "What is shown in this image?")
            state["history"].append({"step": "image", "output": answer})
        except Exception as e:
            state["logs"]["image_error"] = str(e)
        state["current_step"] = "answer"
        return state

    def _video_node(self, state: AgentState) -> AgentState:
        """Handle video-based questions."""
        try:
            data = self._download_file(state["file_url"])
            label = video_label_bytes(data)
            state["history"].append({"step": "video", "output": label})
        except Exception as e:
            state["logs"]["video_error"] = str(e)
        state["current_step"] = "answer"
        return state

    def _sheet_node(self, state: AgentState) -> AgentState:
        """Handle spreadsheet-based questions."""
        try:
            data = self._download_file(state["file_url"])
            answer = sheet_answer_bytes(data, state["file_url"])
            state["history"].append({"step": "sheet", "output": answer})
        except Exception as e:
            state["logs"]["sheet_error"] = str(e)
        state["current_step"] = "answer"
        return state

    def _perform_search(self, state: AgentState) -> AgentState:
        try:
            results = simple_search(state["search_query"], max_results=6)
            print("\nSearch Results:")
            for i, s in enumerate(results, 1):
                print(f"[{i}] {s[:120]}…")
            
            if not results:
                print("Warning: No search results found")
                state["needs_search"] = True
            else:
                state["needs_search"] = False
                
            state["history"].append({"step": "search", "results": results})
            
        except Exception as e:
            print(f"Search error: {str(e)}")
            state["needs_search"] = True
            state["history"].append({"step": "search", "error": str(e)})
            
        state["current_step"] = "answer"
        return state

    def _code_analysis_node(self, state: AgentState) -> AgentState:
        """Handle code analysis questions."""
        try:
            outputs = []
            for block in state["code_blocks"]:
                if block["language"].lower() == "python":
                    result = run_python(block["code"])   # execute safely
                    outputs.append(result)
            state["history"].append({"step": "code", "output": "\n".join(outputs)})
        except Exception as e:
            state["logs"]["code_error"] = str(e)
        state["current_step"] = "answer"
        return state

    def _generate_answer(self, state: AgentState) -> AgentState:
        # Collect all tool outputs with clear section headers
        materials = []
        
        # Add search results if any
        search_results = [h for h in state["history"] if h["step"] == "search"]
        if search_results:
            materials.append("=== Search Results ===")
            for result in search_results:
                for item in result.get("results", []):
                    materials.append(item)
        
        # Add image analysis if any
        image_results = [h for h in state["history"] if h["step"] == "image"]
        if image_results:
            materials.append("=== Image Analysis ===")
            for result in image_results:
                materials.append(result.get("output", ""))
        
        # Add video analysis if any
        video_results = [h for h in state["history"] if h["step"] == "video"]
        if video_results:
            materials.append("=== Video Analysis ===")
            for result in video_results:
                materials.append(result.get("output", ""))
        
        # Add spreadsheet analysis if any
        sheet_results = [h for h in state["history"] if h["step"] == "sheet"]
        if sheet_results:
            materials.append("=== Spreadsheet Analysis ===")
            for result in sheet_results:
                materials.append(result.get("output", ""))
        
        # Join all materials with clear separation
        search_block = "\n\n".join(materials) if materials else "No materials available."
        
        # First attempt with full context
        prompt = f"""
You are a helpful assistant. Your task is to answer the question using ONLY the materials provided.
If you cannot find a direct answer, provide the most relevant information you can find.

QUESTION:
{state['question']}

MATERIALS:
{search_block}

Provide a direct and concise answer based on the materials above.
"""
        try:
            answer = self._call_llm(prompt, 300).strip()
            
            # If first attempt fails or is empty, try a more direct prompt
            if not answer or any(k in answer.lower() for k in ["cannot", "sorry", "don't know"]):
                print("\nFirst attempt failed, trying direct prompt...")
                direct_prompt = f"""
Answer this question directly and concisely. Use the materials provided.

QUESTION:
{state['question']}

MATERIALS:
{search_block}

If you cannot find an exact answer, provide the most relevant information from the materials.
"""
                answer = self._call_llm(direct_prompt, 300).strip()
            
            # Final validation and fallback
            if not answer:
                print("\nBoth attempts failed, using fallback answer...")
                if materials:
                    # If we have materials but no answer, summarize what we know
                    summary_prompt = f"""
Summarize the key information from these materials in one sentence:

{search_block}
"""
                    answer = self._call_llm(summary_prompt, 150).strip()
                else:
                    answer = "I cannot provide a definitive answer at this time."
            
            state["final_answer"] = answer
            state["current_step"] = "done"
            
        except Exception as e:
            print(f"\nLLM Error in answer generation: {str(e)}")
            state["final_answer"] = "I encountered an error while generating the answer."
            state["current_step"] = "done"
            
        return state

    def _build_workflow(self) -> Graph:
        sg = StateGraph(state_schema=AgentState)
        
        # Add nodes
        sg.add_node("analyze", self._analyze_question)
        sg.add_node("search", self._perform_search)
        sg.add_node("answer", self._generate_answer)
        sg.add_node("image", self._image_node)
        sg.add_node("video", self._video_node)
        sg.add_node("sheet", self._sheet_node)
        sg.add_node("code", self._code_analysis_node)

        # Add edges
        sg.add_edge("analyze", "search")
        sg.add_edge("analyze", "answer")
        sg.add_edge("search", "answer")
        sg.add_edge("image", "answer")
        sg.add_edge("video", "answer")
        sg.add_edge("sheet", "answer")
        sg.add_edge("code", "answer")

        def router(state: AgentState):
            return state["current_step"]

        sg.add_conditional_edges("analyze", router, {
            "search": "search", 
            "answer": "answer",
            "image": "image",
            "video": "video",
            "sheet": "sheet",
            "code": "code"
        })
        
        sg.set_entry_point("analyze")
        sg.set_finish_point("answer")
        return sg.compile()

    def __call__(self, question: str, task_id: str = "unknown") -> str:
        # Parse question to get both text and file_url
        try:
            question_data = json.loads(question)
            state: AgentState = {
                "question": question_data.get("question", ""),
                "current_step": "analyze",
                "final_answer": "",
                "history": [],
                "needs_search": False,
                "search_query": "",
                "task_id": task_id,
                "logs": {},
                "file_url": question_data.get("file_url", ""),
                "code_blocks": question_data.get("code_blocks", [])
            }
        except (json.JSONDecodeError, KeyError) as e:
            print(f"Error parsing question data: {e}")
            state: AgentState = {
                "question": question,
                "current_step": "analyze",
                "final_answer": "",
                "history": [],
                "needs_search": False,
                "search_query": "",
                "task_id": task_id,
                "logs": {},
                "file_url": "",
                "code_blocks": []
            }
        
        final_state = self.workflow.invoke(state)
        return final_state["final_answer"]

    def _download_file(self, url: str) -> bytes:
        """Download a file from a URL."""
        r = requests.get(url, timeout=30)
        r.raise_for_status()
        return r.content

# ----------------------------------------------------------------------------------
# Gradio Interface & Submission Routines
# ----------------------------------------------------------------------------------

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")
    print("Space ID: ", space_id)
    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
    try:
        print("Initializing agent...")
        agent = BasicAgent()
        print("Agent initialized successfully.")
    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
    agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
    print(f"Agent code location: {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 workflow on {len(questions_data)} questions...")
    
    for item in questions_data:
        task_id = item.get("task_id")
        if not task_id:
            print(f"Skipping item with missing task_id: {item}")
            continue

        try:
            print(f"\nProcessing question {task_id}...")
            
            # Pass the entire item as JSON string
            question_json = json.dumps(item)
            answer = agent(question_json, task_id)
            
            # Add to results
            answers_payload.append({"task_id": task_id, "submitted_answer": answer})
            results_log.append({
                "Task ID": task_id,
                "Question": item.get("question", ""),
                "Submitted Answer": answer
            })
            
            print(f"Completed question {task_id}")
            
        except Exception as e:
            print(f"Error running agent on task {task_id}: {e}")
            results_log.append({
                "Task ID": task_id,
                "Question": item.get("question", ""),
                "Submitted Answer": f"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 workflow 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)
    results_table = gr.DataFrame(
        label="Questions and Agent Answers", 
        wrap=True,
        column_widths=["10%", "30%", "30%", "30%"]  # Adjust column widths for better display
    )

    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")

    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(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)