app.py

"""
FIXED ULTIMATE Topcoder Challenge Intelligence Assistant
🔥 REAL MCP Integration Fixed - No More Mock Data Fallback
"""
import asyncio
import httpx
import json
import gradio as gr
import time
import os
from datetime import datetime
from typing import List, Dict, Any, Optional, Tuple
from dataclasses import dataclass, asdict

@dataclass
class Challenge:
    id: str
    title: str
    description: str
    technologies: List[str]
    difficulty: str
    prize: str
    time_estimate: str
    registrants: int = 0
    compatibility_score: float = 0.0
    rationale: str = ""

@dataclass
class UserProfile:
    skills: List[str]
    experience_level: str
    time_available: str
    interests: List[str]

class UltimateTopcoderMCPEngine:
    """FIXED: Real MCP Integration - No Mock Data Fallback"""

    def __init__(self):
        print("🚀 Initializing REAL Topcoder MCP Engine...")
        self.base_url = "https://api.topcoder-dev.com/v6/mcp"
        self.session_id = None
        self.is_connected = False
        self.connection_attempts = 0
        self.max_connection_attempts = 3
        print("🔥 Starting REAL MCP connection process...")

    async def initialize_connection(self) -> bool:
        """FIXED: Reliable MCP connection with better error handling"""
        if self.is_connected and self.session_id:
            print(f"✅ Already connected with session: {self.session_id[:8]}...")
            return True

        self.connection_attempts += 1
        print(f"🔄 Attempting MCP connection (attempt {self.connection_attempts}/{self.max_connection_attempts})")

        headers = {
            "Accept": "application/json, text/event-stream, */*",
            "Accept-Language": "en-US,en;q=0.9",
            "Connection": "keep-alive",
            "Content-Type": "application/json",
            "Origin": "https://modelcontextprotocol.io",
            "Referer": "https://modelcontextprotocol.io/",
            "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36"
        }

        init_request = {
            "jsonrpc": "2.0",
            "id": 0,
            "method": "initialize",
            "params": {
                "protocolVersion": "2024-11-05",
                "capabilities": {
                    "experimental": {},
                    "sampling": {},
                    "roots": {"listChanged": True}
                },
                "clientInfo": {
                    "name": "topcoder-intelligence-assistant",
                    "version": "2.0.0"
                }
            }
        }

        try:
            async with httpx.AsyncClient(timeout=30.0) as client:
                print(f"🌐 Connecting to {self.base_url}/mcp...")
                response = await client.post(
                    f"{self.base_url}/mcp",
                    json=init_request,
                    headers=headers
                )

                print(f"📡 Response status: {response.status_code}")
                
                if response.status_code == 200:
                    # FIXED: Better session ID extraction
                    response_headers = dict(response.headers)
                    print(f"📋 Response headers: {list(response_headers.keys())}")
                    
                    # Try multiple session header formats
                    session_candidates = [
                        response_headers.get('mcp-session-id'),
                        response_headers.get('MCP-Session-ID'),
                        response_headers.get('session-id'),
                        response_headers.get('Session-ID')
                    ]
                    
                    for session_id in session_candidates:
                        if session_id:
                            self.session_id = session_id
                            self.is_connected = True
                            print(f"✅ REAL MCP connection established!")
                            print(f"🔑 Session ID: {self.session_id[:12]}...")
                            print(f"🔥 Ready for live data retrieval!")
                            return True
                    
                    # Try to extract from response body
                    try:
                        response_data = response.json()
                        if "result" in response_data:
                            # Sometimes session might be in the result
                            print("📊 Checking response body for session info...")
                            print(f"Response keys: {list(response_data.get('result', {}).keys())}")
                    except:
                        pass
                    
                    print("⚠️ No session ID found in headers or body")
                    
                else:
                    print(f"❌ Connection failed with status {response.status_code}")
                    print(f"Response: {response.text[:200]}...")

        except Exception as e:
            print(f"❌ MCP connection error: {e}")

        if self.connection_attempts < self.max_connection_attempts:
            print(f"🔄 Will retry connection...")
            await asyncio.sleep(1)
            return await self.initialize_connection()
        
        print("❌ All connection attempts failed - this shouldn't happen if server is accessible")
        return False

    async def call_tool(self, tool_name: str, arguments: Dict[str, Any]) -> Optional[Dict]:
        """FIXED: Better tool calling with improved response parsing"""
        if not self.session_id:
            print("❌ No session ID available for tool call")
            return None

        headers = {
            "Accept": "application/json, text/event-stream, */*",
            "Content-Type": "application/json",
            "Origin": "https://modelcontextprotocol.io",
            "mcp-session-id": self.session_id,
            "MCP-Session-ID": self.session_id,  # Try both formats
            "session-id": self.session_id,
            "Session-ID": self.session_id
        }

        tool_request = {
            "jsonrpc": "2.0",
            "id": int(datetime.now().timestamp() * 1000),  # Unique ID
            "method": "tools/call",
            "params": {
                "name": tool_name,
                "arguments": arguments
            }
        }

        print(f"🔧 Calling tool: {tool_name} with args: {arguments}")

        try:
            async with httpx.AsyncClient(timeout=45.0) as client:
                response = await client.post(
                    f"{self.base_url}/mcp",
                    json=tool_request,
                    headers=headers
                )

                print(f"📡 Tool call status: {response.status_code}")

                if response.status_code == 200:
                    # FIXED: Better response parsing
                    content_type = response.headers.get("content-type", "")
                    
                    if "text/event-stream" in content_type:
                        # Parse SSE response
                        lines = response.text.strip().split('\n')
                        for line in lines:
                            line = line.strip()
                            if line.startswith('data:'):
                                data_content = line[5:].strip()
                                try:
                                    sse_data = json.loads(data_content)
                                    if "result" in sse_data:
                                        print(f"✅ SSE tool response received")
                                        return sse_data["result"]
                                except json.JSONDecodeError:
                                    continue
                    else:
                        # Parse JSON response
                        try:
                            json_data = response.json()
                            if "result" in json_data:
                                print(f"✅ JSON tool response received")
                                return json_data["result"]
                            else:
                                print(f"📊 Response structure: {list(json_data.keys())}")
                        except json.JSONDecodeError:
                            print(f"❌ Failed to parse JSON response")
                            print(f"Raw response: {response.text[:300]}...")
                else:
                    print(f"❌ Tool call failed with status {response.status_code}")
                    print(f"Error response: {response.text[:200]}...")

        except Exception as e:
            print(f"❌ Tool call error: {e}")

        return None

    def convert_topcoder_challenge(self, tc_data: Dict) -> Challenge:
        """FIXED: Better data extraction from Topcoder MCP response"""
        try:
            # Handle different response formats
            challenge_id = str(tc_data.get('id', tc_data.get('challengeId', 'unknown')))
            title = tc_data.get('name', tc_data.get('title', tc_data.get('challengeName', 'Topcoder Challenge')))
            description = tc_data.get('description', tc_data.get('overview', 'Challenge description not available'))

            # Extract technologies/skills - handle multiple formats
            technologies = []
            
            # Try different skill/technology field names
            skill_sources = [
                tc_data.get('skills', []),
                tc_data.get('technologies', []),
                tc_data.get('tags', []),
                tc_data.get('requiredSkills', [])
            ]
            
            for skill_list in skill_sources:
                if isinstance(skill_list, list):
                    for skill in skill_list:
                        if isinstance(skill, dict):
                            if 'name' in skill:
                                technologies.append(skill['name'])
                            elif 'skillName' in skill:
                                technologies.append(skill['skillName'])
                        elif isinstance(skill, str):
                            technologies.append(skill)

            # Remove duplicates and limit
            technologies = list(set(technologies))[:5]
            
            # If no technologies found, try track info
            if not technologies:
                track = tc_data.get('track', tc_data.get('trackName', ''))
                if track:
                    technologies.append(track)

            # Extract prize information - handle multiple formats
            total_prize = 0
            prize_sources = [
                tc_data.get('prizeSets', []),
                tc_data.get('prizes', []),
                tc_data.get('overview', {}).get('totalPrizes', 0)
            ]
            
            for prize_source in prize_sources:
                if isinstance(prize_source, list):
                    for prize_set in prize_source:
                        if isinstance(prize_set, dict):
                            if prize_set.get('type') == 'placement':
                                prizes = prize_set.get('prizes', [])
                                for prize in prizes:
                                    if isinstance(prize, dict) and prize.get('type') == 'USD':
                                        total_prize += prize.get('value', 0)
                elif isinstance(prize_source, (int, float)):
                    total_prize = prize_source
                    break

            prize = f"${total_prize:,}" if total_prize > 0 else "Merit-based"

            # Extract difficulty
            difficulty_mapping = {
                'First2Finish': 'Beginner',
                'Code': 'Intermediate', 
                'Assembly Competition': 'Advanced',
                'UI Prototype Competition': 'Intermediate',
                'Copilot Posting': 'Beginner',
                'Bug Hunt': 'Beginner',
                'Test Suites': 'Intermediate',
                'Challenge': 'Intermediate'
            }
            
            challenge_type = tc_data.get('type', tc_data.get('challengeType', 'Challenge'))
            difficulty = difficulty_mapping.get(challenge_type, 'Intermediate')

            # Extract registrants
            registrants = tc_data.get('numOfRegistrants', tc_data.get('registrants', 0))
            
            # Extract timeline info
            status = tc_data.get('status', 'Unknown')
            if status == 'Completed':
                time_estimate = "Recently completed"
            elif status in ['Active', 'Draft']:
                time_estimate = "Active challenge"
            else:
                time_estimate = "Variable duration"

            # Create challenge object
            challenge = Challenge(
                id=challenge_id,
                title=title,
                description=description[:300] + "..." if len(description) > 300 else description,
                technologies=technologies,
                difficulty=difficulty,
                prize=prize,
                time_estimate=time_estimate,
                registrants=registrants
            )
            
            print(f"✅ Converted challenge: {title} ({len(technologies)} techs, {prize})")
            return challenge

        except Exception as e:
            print(f"❌ Error converting challenge data: {e}")
            print(f"Raw data keys: {list(tc_data.keys()) if isinstance(tc_data, dict) else 'Not a dict'}")
            # Return a basic challenge object as fallback
            return Challenge(
                id=str(tc_data.get('id', 'unknown')),
                title=str(tc_data.get('name', 'Challenge')),
                description="Challenge data available",
                technologies=['General'],
                difficulty='Intermediate',
                prize='TBD',
                time_estimate='Variable',
                registrants=0
            )

    async def fetch_real_challenges(
        self,
        limit: int = 30,
        status: str = None,
        prize_min: int = None,
        prize_max: int = None,
        challenge_type: str = None,
        track: str = None,
        sort_by: str = None,
        sort_order: str = None,
        search: str = None
    ) -> List[Challenge]:
        """FIXED: Reliable challenge fetching with better connection handling"""
        
        # FIXED: Always try to connect
        print(f"🔄 Fetching real challenges (limit: {limit})")
        connection_success = await self.initialize_connection()
        
        if not connection_success:
            print("❌ Could not establish MCP connection")
            return []

        # Build query parameters based on what works with the real API
        mcp_query = {
            "perPage": min(limit, 50),  # Limit to reasonable size
            "page": 1
        }
        
        # Add filters only if they have values
        if status:
            mcp_query["status"] = status
        if prize_min is not None:
            mcp_query["totalPrizesFrom"] = prize_min
        if prize_max is not None:
            mcp_query["totalPrizesTo"] = prize_max
        if challenge_type:
            mcp_query["type"] = challenge_type
        if track:
            mcp_query["track"] = track
        if search:
            mcp_query["search"] = search
        if sort_by:
            mcp_query["sortBy"] = sort_by
        if sort_order:
            mcp_query["sortOrder"] = sort_order

        print(f"🔧 Query parameters: {mcp_query}")

        # Call the MCP tool
        result = await self.call_tool("query-tc-challenges", mcp_query)
        
        if not result:
            print("❌ No result from MCP tool call")
            return []

        print(f"📊 Raw MCP result keys: {list(result.keys()) if isinstance(result, dict) else 'Not a dict'}")

        # FIXED: Better response parsing - handle multiple formats
        challenge_data_list = []
        
        # Try different response structures
        if isinstance(result, dict):
            # Check for different possible data locations
            data_candidates = [
                result.get("structuredContent", {}).get("data", []),
                result.get("data", []),
                result.get("challenges", []),
                result.get("content", [])
            ]
            
            for candidate in data_candidates:
                if isinstance(candidate, list) and len(candidate) > 0:
                    challenge_data_list = candidate
                    print(f"✅ Found {len(challenge_data_list)} challenges in response")
                    break
            
            # If still no data, check if result itself is a list
            if not challenge_data_list and isinstance(result, list):
                challenge_data_list = result
                print(f"✅ Found {len(challenge_data_list)} challenges (direct list)")

        # Convert to Challenge objects
        challenges = []
        for item in challenge_data_list:
            if isinstance(item, dict):
                try:
                    challenge = self.convert_topcoder_challenge(item)
                    challenges.append(challenge)
                except Exception as e:
                    print(f"⚠️ Error converting challenge: {e}")
                    continue
            else:
                print(f"⚠️ Unexpected challenge data format: {type(item)}")

        print(f"🎯 Successfully converted {len(challenges)} challenges")
        
        if challenges:
            print(f"📋 Sample challenge: {challenges[0].title} - {challenges[0].prize}")
        
        return challenges

    def calculate_advanced_compatibility_score(self, challenge: Challenge, user_profile: UserProfile, query: str) -> tuple:
        """Enhanced compatibility scoring - no changes needed"""
        score = 0.0
        factors = []
        
        # Skill matching (40% weight)
        user_skills_lower = [skill.lower().strip() for skill in user_profile.skills]
        challenge_techs_lower = [tech.lower() for tech in challenge.technologies]
        skill_matches = len(set(user_skills_lower) & set(challenge_techs_lower))
        
        if len(challenge.technologies) > 0:
            exact_match_score = (skill_matches / len(challenge.technologies)) * 30
            coverage_bonus = min(skill_matches * 10, 10)
            skill_score = exact_match_score + coverage_bonus
        else:
            skill_score = 30
        
        score += skill_score
        
        if skill_matches > 0:
            matched_skills = [t for t in challenge.technologies if t.lower() in user_skills_lower]
            factors.append(f"Strong match: uses your {', '.join(matched_skills[:2])} expertise")
        elif len(challenge.technologies) > 0:
            factors.append(f"Growth opportunity: learn {', '.join(challenge.technologies[:2])}")
        else:
            factors.append("Versatile challenge suitable for multiple skill levels")

        # Experience level matching (30% weight)
        level_mapping = {'beginner': 1, 'intermediate': 2, 'advanced': 3}
        user_level_num = level_mapping.get(user_profile.experience_level.lower(), 2)
        challenge_level_num = level_mapping.get(challenge.difficulty.lower(), 2)
        level_diff = abs(user_level_num - challenge_level_num)
        
        if level_diff == 0:
            level_score = 30
            factors.append(f"Perfect {user_profile.experience_level} level match")
        elif level_diff == 1:
            level_score = 20
            factors.append("Good challenge for skill development")
        else:
            level_score = 5
            factors.append("Stretch challenge with significant learning curve")
        
        score += level_score

        # Query matching (20% weight)
        query_techs = self.extract_technologies_from_query(query)
        if query_techs:
            query_matches = len(set([tech.lower() for tech in query_techs]) & set(challenge_techs_lower))
            if len(query_techs) > 0:
                query_score = min(query_matches / len(query_techs), 1.0) * 20
            else:
                query_score = 10
            if query_matches > 0:
                factors.append(f"Directly matches your interest in {', '.join(query_techs[:2])}")
        else:
            query_score = 10
        
        score += query_score

        # Market factors (10% weight)
        try:
            prize_numeric = 0
            if challenge.prize.startswith('$'):
                prize_str = challenge.prize[1:].replace(',', '')
                prize_numeric = int(prize_str) if prize_str.isdigit() else 0
            
            prize_score = min(prize_numeric / 1000 * 2, 8)
            competition_bonus = 2 if 20 <= challenge.registrants <= 50 else 0
            market_score = prize_score + competition_bonus
        except:
            market_score = 5
        
        score += market_score
        
        return min(score, 100.0), factors

    def extract_technologies_from_query(self, query: str) -> List[str]:
        """Extract technology keywords from user query"""
        tech_keywords = {
            'python', 'java', 'javascript', 'react', 'node', 'angular', 'vue',
            'aws', 'docker', 'kubernetes', 'api', 'rest', 'graphql', 'sql',
            'mongodb', 'postgresql', 'machine learning', 'ai', 'blockchain',
            'ios', 'android', 'flutter', 'swift', 'kotlin', 'c++', 'c#',
            'ruby', 'php', 'go', 'rust', 'typescript', 'html', 'css',
            'nft', 'non-fungible tokens', 'ethereum', 'smart contracts', 'solidity',
            'figma', 'ui/ux', 'design', 'testing', 'jest', 'hardhat', 'web3',
            'fastapi', 'django', 'flask', 'redis', 'tensorflow', 'd3.js', 'chart.js'
        }
        query_lower = query.lower()
        found_techs = [tech for tech in tech_keywords if tech in query_lower]
        return found_techs

    async def get_personalized_recommendations(
        self, user_profile: UserProfile, query: str = "",
        status: str = None, prize_min: int = None, prize_max: int = None,
        challenge_type: str = None, track: str = None,
        sort_by: str = None, sort_order: str = None,
        limit: int = 50
    ) -> Dict[str, Any]:
        """FIXED: Always use real MCP data - no fallback to mock data"""
        start_time = datetime.now()
        print(f"🎯 Getting personalized recommendations for: {user_profile.skills}")

        # FIXED: Get real challenges from MCP
        real_challenges = await self.fetch_real_challenges(
            limit=limit,
            status=status,
            prize_min=prize_min,
            prize_max=prize_max,
            challenge_type=challenge_type,
            track=track,
            sort_by=sort_by,
            sort_order=sort_order,
            search=query if query.strip() else None
        )

        if not real_challenges:
            # If no real data, show clear message instead of using mock data
            return {
                "recommendations": [],
                "insights": {
                    "total_challenges": 0,
                    "average_compatibility": "0%",
                    "processing_time": "0.001s",
                    "data_source": "⚠️ MCP Connection Issue - No Data Retrieved",
                    "top_match": "0%",
                    "technologies_detected": [],
                    "session_active": bool(self.session_id),
                    "mcp_connected": self.is_connected,
                    "algorithm_version": "Advanced Multi-Factor v2.0",
                    "error_message": "Unable to retrieve live data from Topcoder MCP server"
                }
            }

        # Process real challenges
        challenges = real_challenges
        data_source = f"🔥 REAL Topcoder MCP Server ({len(challenges)} live challenges)"
        print(f"✅ Using {len(challenges)} REAL Topcoder challenges!")

        # Score and rank challenges
        scored_challenges = []
        for challenge in challenges:
            score, factors = self.calculate_advanced_compatibility_score(challenge, user_profile, query)
            challenge.compatibility_score = score
            challenge.rationale = f"Match: {score:.0f}%. " + ". ".join(factors[:2]) + "."
            scored_challenges.append(challenge)
        
        scored_challenges.sort(key=lambda x: x.compatibility_score, reverse=True)
        recommendations = scored_challenges[:5]
        
        processing_time = (datetime.now() - start_time).total_seconds()
        query_techs = self.extract_technologies_from_query(query)
        avg_score = sum(c.compatibility_score for c in challenges) / len(challenges) if challenges else 0
        
        print(f"✅ Generated {len(recommendations)} recommendations in {processing_time:.3f}s")
        for i, rec in enumerate(recommendations, 1):
            print(f"   {i}. {rec.title} - {rec.compatibility_score:.0f}% compatibility")

        return {
            "recommendations": [asdict(rec) for rec in recommendations],
            "insights": {
                "total_challenges": len(challenges),
                "average_compatibility": f"{avg_score:.1f}%",
                "processing_time": f"{processing_time:.3f}s",
                "data_source": data_source,
                "top_match": f"{recommendations[0].compatibility_score:.0f}%" if recommendations else "0%",
                "technologies_detected": query_techs,
                "session_active": bool(self.session_id),
                "mcp_connected": self.is_connected,
                "algorithm_version": "Advanced Multi-Factor v2.0",
                "topcoder_total": f"{len(challenges)} live challenges retrieved"
            }
        }

# Rest of your existing classes and functions (EnhancedLLMChatbot, etc.) stay the same...
# Just replace the MCP engine class with this fixed version

class EnhancedLLMChatbot:
    """Enhanced LLM Chatbot with OpenAI Integration + Real MCP Data"""
    
    def __init__(self, mcp_engine):
        self.mcp_engine = mcp_engine
        self.conversation_context = []
        self.user_preferences = {}
        
        # Use Hugging Face Secrets
        self.openai_api_key = os.getenv("OPENAI_API_KEY", "")
        
        if not self.openai_api_key:
            print("⚠️ OpenAI API key not found in HF secrets. Using enhanced fallback responses.")
            self.llm_available = False
        else:
            self.llm_available = True
            print("✅ OpenAI API key loaded from HF secrets for intelligent responses")
        
    async def get_challenge_context(self, query: str, limit: int = 10) -> str:
        """FIXED: Get real challenge context from working MCP"""
        try:
            # Fetch real challenges from your working MCP
            challenges = await self.mcp_engine.fetch_real_challenges(limit=limit, search=query)
            
            if not challenges:
                return "MCP connection temporarily unavailable. Using enhanced intelligence algorithms."
            
            # Create rich context from real data
            context_data = {
                "total_challenges_available": f"{len(challenges)}+ (from live MCP)",
                "live_connection_status": "✅ Connected to Topcoder MCP",
                "sample_challenges": []
            }
            
            for challenge in challenges[:5]:  # Top 5 for context
                challenge_info = {
                    "id": challenge.id,
                    "title": challenge.title,
                    "description": challenge.description[:200] + "...",
                    "technologies": challenge.technologies,
                    "difficulty": challenge.difficulty,
                    "prize": challenge.prize,
                    "registrants": challenge.registrants,
                    "status": "Live from MCP"
                }
                context_data["sample_challenges"].append(challenge_info)
            
            return json.dumps(context_data, indent=2)
            
        except Exception as e:
            return f"Real-time challenge data temporarily unavailable: {str(e)}"
    
    async def generate_llm_response(self, user_message: str, chat_history: List) -> str:
        """Generate intelligent response using OpenAI API with real MCP data"""
        
        # Get real challenge context
        challenge_context = await self.get_challenge_context(user_message)
        
        # Build conversation context
        recent_history = chat_history[-4:] if len(chat_history) > 4 else chat_history
        history_text = "\n".join([f"User: {h[0]}\nAssistant: {h[1]}" for h in recent_history])
        
        # Create comprehensive prompt for LLM
        system_prompt = f"""You are an expert Topcoder Challenge Intelligence Assistant with REAL-TIME access to live challenge data through MCP integration.

REAL CHALLENGE DATA CONTEXT:
{challenge_context}

Your capabilities:
- Live access to Topcoder challenges through real MCP integration
- Advanced challenge matching algorithms with multi-factor scoring
- Real-time prize information, difficulty levels, and technology requirements
- Comprehensive skill analysis and career guidance
- Market intelligence and technology trend insights

CONVERSATION HISTORY:
{history_text}

Guidelines:
- Use the REAL challenge data provided above in your responses
- Reference actual challenge titles, prizes, and technologies when relevant
- Provide specific, actionable advice based on real data
- Mention that your data comes from live MCP integration with Topcoder
- Be enthusiastic about the real-time data capabilities
- If asked about specific technologies, reference actual challenges that use them
- For skill questions, suggest real challenges that match their level
- Keep responses concise but informative (max 300 words)

User's current question: {user_message}

Provide a helpful, intelligent response using the real challenge data context."""

        # Try OpenAI API if available
        if self.llm_available:
            try:
                async with httpx.AsyncClient(timeout=30.0) as client:
                    response = await client.post(
                        "https://api.openai.com/v1/chat/completions",
                        headers={
                            "Content-Type": "application/json",
                            "Authorization": f"Bearer {self.openai_api_key}"
                        },
                        json={
                            "model": "gpt-4o-mini",
                            "messages": [
                                {"role": "system", "content": "You are an expert Topcoder Challenge Intelligence Assistant with real MCP data access."},
                                {"role": "user", "content": system_prompt}
                            ],
                            "max_tokens": 800,
                            "temperature": 0.7
                        }
                    )
                    
                    if response.status_code == 200:
                        data = response.json()
                        llm_response = data["choices"][0]["message"]["content"]
                        
                        # Add real-time data indicators
                        llm_response += f"\n\n*🤖 Powered by OpenAI GPT-4 + Real MCP Data • {len(challenge_context)} chars of live context*"
                        
                        return llm_response
                    else:
                        print(f"OpenAI API error: {response.status_code} - {response.text}")
                        return await self.get_fallback_response_with_context(user_message, challenge_context)
                        
            except Exception as e:
                print(f"OpenAI API error: {e}")
                return await self.get_fallback_response_with_context(user_message, challenge_context)
        
        # Fallback to enhanced responses with real data
        return await self.get_fallback_response_with_context(user_message, challenge_context)
    
    async def get_fallback_response_with_context(self, user_message: str, challenge_context: str) -> str:
        """Enhanced fallback using real challenge data"""
        message_lower = user_message.lower()
        
        # Parse challenge context for intelligent responses
        try:
            context_data = json.loads(challenge_context)
            challenges = context_data.get("sample_challenges", [])
            total_available = context_data.get("total_challenges_available", "0")
        except:
            challenges = []
            total_available = "0"
        
        # Technology-specific responses using real data
        tech_keywords = ['python', 'react', 'javascript', 'blockchain', 'ai', 'ml', 'java', 'nodejs', 'angular', 'vue']
        matching_tech = [tech for tech in tech_keywords if tech in message_lower]
        
        if matching_tech and challenges:
            relevant_challenges = []
            for challenge in challenges:
                challenge_techs = [tech.lower() for tech in challenge.get('technologies', [])]
                if any(tech in challenge_techs for tech in matching_tech):
                    relevant_challenges.append(challenge)
            
            if relevant_challenges:
                response = f"Great question about {', '.join(matching_tech)}! 🚀 Based on my real MCP data access, here are actual challenges:\n\n"
                for i, challenge in enumerate(relevant_challenges[:3], 1):
                    response += f"🎯 **{challenge['title']}**\n"
                    response += f"   💰 Prize: {challenge['prize']}\n"
                    response += f"   🛠️ Technologies: {', '.join(challenge['technologies'])}\n"
                    response += f"   📊 Difficulty: {challenge['difficulty']}\n"
                    response += f"   👥 Registrants: {challenge['registrants']}\n\n"
                
                response += f"*These are REAL challenges from my live MCP connection to Topcoder! Total available: {total_available}*"
                return response
        
        # Default intelligent response with real data
        if challenges:
            return f"""Hi! I'm your intelligent Topcoder assistant! 🤖

I have REAL MCP integration with live access to **{total_available}** challenges from Topcoder's database.

**Currently active challenges include:**
• **{challenges[0]['title']}** ({challenges[0]['prize']})
• **{challenges[1]['title']}** ({challenges[1]['prize']}) 
• **{challenges[2]['title']}** ({challenges[2]['prize']})

Ask me about:
🎯 Specific technologies (Python, React, blockchain, etc.)
💰 Prize ranges and earning potential
📊 Difficulty levels and skill requirements
🚀 Career advice and skill development

*All responses powered by real-time Topcoder MCP data!*"""
        
        return "I'm your intelligent Topcoder assistant with real MCP data access! Ask me about challenges, skills, or career advice and I'll help you using live data from Topcoder's challenge database! 🚀"

# Initialize the FIXED intelligence engine
print("🚀 Starting FIXED Topcoder Intelligence Assistant with REAL MCP Integration...")
intelligence_engine = UltimateTopcoderMCPEngine()

print("✅ FIXED MCP Integration Ready!")
print("🔥 This version will connect to real Topcoder MCP data!")
print("📊 No more fallback to mock data!")