feat: AI learning from feedback + v1.12.0

AI Learning System:
- Add FeedbackLearningService for few-shot learning from user feedback
- Integrate learning context into chat prompts (nordabiz_chat.py)
- Add seed examples for cold start (when insufficient real feedback)
- Add /api/admin/ai-learning-status endpoint
- Add learning status section to chat analytics panel

Other Changes:
- Update release notes to v1.12.0
- Remove old password references from documentation (CLAUDE.md)
- Fix password masking in run_migration.py (use regex for any password)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

CLAUDE.md

@@ -359,11 +359,10 @@ Jest to krytyczna podatność bezpieczeństwa (CWE-798: Use of Hard-coded Creden
 **Weryfikacja przed wdrożeniem:**
 ```bash
 # Sprawdź czy nie ma hardcoded credentials w kodzie:
-grep -r "NordaBiz2025Secure" --include="*.py" --include="*.sh" .
 grep -r "PGPASSWORD=" --include="*.sh" .
-grep -r "postgresql://.*:.*@" --include="*.py" . | grep -v "CHANGE_ME" | grep -v ".example"
+grep -r "postgresql://.*:.*@" --include="*.py" . | grep -v "CHANGE_ME" | grep -v ".example" | grep -v "PASSWORD"
 
-# Oczekiwany wynik: brak znalezisk (lub tylko w dokumentacji)
+# Oczekiwany wynik: brak znalezisk (lub tylko w dokumentacji/placeholderach)
 ```
 
 ### Import danych
@@ -871,13 +870,15 @@ python seo_audit.py --company-id 26 --dry-run
 Skrypty w `scripts/` muszą używać **localhost (127.0.0.1)** do połączenia z PostgreSQL:
 
 ```python
-# PRAWIDŁOWO:
+# PRAWIDŁOWO (hasło z .env):
-DATABASE_URL = 'postgresql://nordabiz_app:NordaBiz2025Secure@127.0.0.1:5432/nordabiz'
+DATABASE_URL = 'postgresql://nordabiz_app:<PASSWORD_FROM_ENV>@127.0.0.1:5432/nordabiz'
 
 # BŁĘDNIE (PostgreSQL nie akceptuje zewnętrznych połączeń):
-DATABASE_URL = 'postgresql://nordabiz_app:NordaBiz2025Secure@10.22.68.249:5432/nordabiz'
+DATABASE_URL = 'postgresql://nordabiz_app:<PASSWORD>@10.22.68.249:5432/nordabiz'
 ```
 
+**UWAGA:** Hasło do bazy jest w `.env` na produkcji. NIE commituj haseł do repozytorium!
+
 **Pliki z konfiguracją bazy:**
 - `scripts/seo_audit.py` (linia ~79)
 - `scripts/seo_report_generator.py` (linia ~47)

app.py

@@ -5719,6 +5719,61 @@ def chat_analytics():
         db.close()
 
 
+@app.route('/api/admin/ai-learning-status')
+@login_required
+def api_ai_learning_status():
+    """API: Get AI feedback learning status and examples"""
+    if not current_user.is_admin:
+        return jsonify({'success': False, 'error': 'Not authorized'}), 403
+
+    try:
+        from feedback_learning_service import get_feedback_learning_service
+        service = get_feedback_learning_service()
+        context = service.get_learning_context()
+
+        # Format examples for JSON response
+        positive_examples = []
+        for ex in context.get('positive_examples', []):
+            positive_examples.append({
+                'query': ex.query,
+                'response': ex.response[:300] + '...' if len(ex.response) > 300 else ex.response,
+                'companies': ex.companies_mentioned or []
+            })
+
+        negative_examples = []
+        for ex in context.get('negative_examples', []):
+            negative_examples.append({
+                'query': ex.query,
+                'response': ex.response,
+                'comment': ex.feedback_comment
+            })
+
+        return jsonify({
+            'success': True,
+            'learning_active': True,
+            'stats': context.get('stats', {}),
+            'using_seed_examples': context.get('stats', {}).get('using_seed_examples', False),
+            'positive_examples_count': len(positive_examples),
+            'negative_examples_count': len(negative_examples),
+            'positive_examples': positive_examples,
+            'negative_examples': negative_examples,
+            'negative_patterns': context.get('negative_patterns', []),
+            'generated_at': context.get('generated_at')
+        })
+    except ImportError:
+        return jsonify({
+            'success': True,
+            'learning_active': False,
+            'message': 'Feedback learning service not available'
+        })
+    except Exception as e:
+        logger.error(f"Error getting AI learning status: {e}")
+        return jsonify({
+            'success': False,
+            'error': str(e)
+        }), 500
+
+
 @app.route('/admin/ai-usage')
 @login_required
 def admin_ai_usage():
@@ -7373,22 +7428,40 @@ def api_it_audit_export():
 def release_notes():
     """Historia zmian platformy."""
     releases = [
+        {
+            'version': 'v1.12.0',
+            'date': '11 stycznia 2026',
+            'badges': ['new', 'improve'],
+            'new': [
+                'AI Learning: System uczenia chatbota z feedbacku uzytkownikow',
+                'AI Learning: Few-shot learning z pozytywnych odpowiedzi',
+                'AI Learning: Przyklady startowe (seed) dla zimnego startu',
+                'Panel AI Usage: Szczegolowy widok uzycia AI per uzytkownik',
+                'Panel AI Usage: Klikalne nazwy uzytkownikow w rankingu',
+                'Panel Analytics: Sekcja statusu uczenia AI',
+            ],
+            'improve': [
+                'Stylizowane modale zamiast natywnych dialogow przegladarki',
+                'System toastow do komunikatow sukcesu/bledu',
+                'Bezpieczenstwo: Usuniecie starych hasel z dokumentacji',
+            ],
+        },
         {
             'version': 'v1.11.0',
             'date': '10 stycznia 2026',
             'badges': ['new', 'improve'],
             'new': [
-                'Forum: Kategorie tematów (Propozycja funkcji, Błąd, Pytanie, Ogłoszenie)',
+                'Forum: Kategorie tematow (Propozycja funkcji, Blad, Pytanie, Ogloszenie)',
-                'Forum: Statusy zgłoszeń (Nowy, W realizacji, Rozwiązany, Odrzucony)',
+                'Forum: Statusy zgloszen (Nowy, W realizacji, Rozwiazany, Odrzucony)',
-                'Forum: Załączniki obrazów do tematów i odpowiedzi (JPG, PNG, GIF)',
+                'Forum: Zalaczniki obrazow do tematow i odpowiedzi (JPG, PNG, GIF)',
-                'Forum: Upload wielu plików jednocześnie (do 10 na odpowiedź)',
+                'Forum: Upload wielu plikow jednoczesnie (do 10 na odpowiedz)',
                 'Forum: Drag & drop i wklejanie ze schowka (Ctrl+V)',
-                'Panel admina: Statystyki i zmiana statusów tematów',
+                'Panel admina: Statystyki i zmiana statusow tematow',
             ],
             'improve': [
-                'Bezpieczny upload z walidacją magic bytes i usuwaniem EXIF',
+                'Bezpieczny upload z walidacja magic bytes i usuwaniem EXIF',
-                'Responsywna siatka podglądu załączników',
+                'Responsywna siatka podgladu zalacznikow',
-                'Filtry kategorii i statusów na liście tematów',
+                'Filtry kategorii i statusow na liscie tematow',
             ],
         },
         {

feedback_learning_service.py

@@ -0,0 +1,375 @@
+#!/usr/bin/env python3
+"""
+Feedback Learning Service for NordaBiz AI Chat
+===============================================
+
+Implements few-shot learning from user feedback to improve AI responses.
+
+Features:
+- Collects positive feedback examples for few-shot learning
+- Identifies negative patterns to avoid
+- Provides curated seed examples for cold start
+- Caches learning context for performance
+
+Author: Norda Biznes Development Team
+Created: 2026-01-11
+"""
+
+import json
+import logging
+from datetime import datetime, timedelta
+from typing import Dict, List, Optional, Tuple
+from dataclasses import dataclass
+
+from sqlalchemy import desc, func
+from sqlalchemy.orm import Session
+
+from database import (
+    SessionLocal,
+    AIChatMessage,
+    AIChatFeedback,
+    AIChatConversation
+)
+
+logger = logging.getLogger(__name__)
+
+
+@dataclass
+class FeedbackExample:
+    """Represents a feedback example for learning"""
+    query: str
+    response: str
+    rating: int  # 1=negative, 2=positive
+    feedback_comment: Optional[str] = None
+    companies_mentioned: Optional[List[str]] = None
+
+
+class FeedbackLearningService:
+    """
+    Service for AI learning from user feedback.
+
+    Uses few-shot learning approach:
+    - Positive examples show AI what good responses look like
+    - Negative examples help avoid common mistakes
+    - Seed examples provide baseline for cold start
+    """
+
+    # Curated seed examples for cold start (when insufficient real feedback)
+    SEED_EXAMPLES = [
+        FeedbackExample(
+            query="Szukam firmy od stron internetowych",
+            response="W Norda Biznes stronami internetowymi zajmuje się **PIXLAB** - profesjonalne tworzenie stron WWW, aplikacji i systemów IT. Kontakt: tel. 58 888 99 00, www.pixlab.pl",
+            rating=2,
+            companies_mentioned=["PIXLAB"]
+        ),
+        FeedbackExample(
+            query="Kto robi dachy w okolicy Wejherowa?",
+            response="Dachami i pokryciami dachowymi zajmuje się **Baumar** - kompleksowe usługi dekarskie, remonty dachów. Kontakt: tel. 58 672 15 20, www.baumar.pl. Również **Pro-Bud** oferuje usługi budowlane w tym pokrycia dachowe.",
+            rating=2,
+            companies_mentioned=["Baumar", "Pro-Bud"]
+        ),
+        FeedbackExample(
+            query="Potrzebuję księgowej",
+            response="Usługi księgowe w Norda Biznes oferują:\n- **MKonsult** - biuro rachunkowe, kadry i płace. Tel: 58 672 44 55\n- **Progress Optima** - doradztwo finansowe i księgowość dla firm",
+            rating=2,
+            companies_mentioned=["MKonsult", "Progress Optima"]
+        ),
+    ]
+
+    # Patterns to avoid (negative examples)
+    NEGATIVE_PATTERNS = [
+        "Nie posiadamy takiej firmy",  # Too dismissive
+        "Niestety nie mogę pomóc",  # Should try harder
+        "Brak danych",  # Too brief
+    ]
+
+    def __init__(self, cache_ttl_minutes: int = 30):
+        """
+        Initialize Feedback Learning Service
+
+        Args:
+            cache_ttl_minutes: How long to cache learning context
+        """
+        self.cache_ttl = timedelta(minutes=cache_ttl_minutes)
+        self._cache: Optional[Dict] = None
+        self._cache_time: Optional[datetime] = None
+
+    def get_learning_context(self, db: Optional[Session] = None) -> Dict:
+        """
+        Get learning context for AI prompt enrichment.
+
+        Returns cached context or builds new one if expired.
+
+        Args:
+            db: Optional database session (creates new if not provided)
+
+        Returns:
+            Dict with positive_examples, negative_patterns, stats
+        """
+        # Check cache
+        if self._cache and self._cache_time:
+            if datetime.now() - self._cache_time < self.cache_ttl:
+                return self._cache
+
+        # Build new context
+        close_db = False
+        if db is None:
+            db = SessionLocal()
+            close_db = True
+
+        try:
+            context = self._build_learning_context(db)
+
+            # Update cache
+            self._cache = context
+            self._cache_time = datetime.now()
+
+            return context
+
+        finally:
+            if close_db:
+                db.close()
+
+    def _build_learning_context(self, db: Session) -> Dict:
+        """
+        Build learning context from database feedback.
+
+        Args:
+            db: Database session
+
+        Returns:
+            Learning context dict
+        """
+        # Get positive examples from feedback
+        positive_examples = self._get_positive_examples(db, limit=5)
+
+        # Get negative examples
+        negative_examples = self._get_negative_examples(db, limit=3)
+
+        # Calculate stats
+        stats = self._get_feedback_stats(db)
+
+        # Use seed examples if insufficient real data
+        if len(positive_examples) < 3:
+            # Mix real and seed examples
+            seed_to_add = 3 - len(positive_examples)
+            positive_examples.extend(self.SEED_EXAMPLES[:seed_to_add])
+            stats['using_seed_examples'] = True
+        else:
+            stats['using_seed_examples'] = False
+
+        return {
+            'positive_examples': positive_examples,
+            'negative_examples': negative_examples,
+            'negative_patterns': self.NEGATIVE_PATTERNS,
+            'stats': stats,
+            'generated_at': datetime.now().isoformat()
+        }
+
+    def _get_positive_examples(self, db: Session, limit: int = 5) -> List[FeedbackExample]:
+        """
+        Get positive feedback examples for few-shot learning.
+
+        Prioritizes:
+        1. Most recent positive feedback
+        2. With comments (more context)
+        3. Diverse queries (different topics)
+
+        Args:
+            db: Database session
+            limit: Max examples to return
+
+        Returns:
+            List of FeedbackExample objects
+        """
+        examples = []
+
+        # Query positive feedback (rating=2 = thumbs up)
+        positive_messages = db.query(AIChatMessage).filter(
+            AIChatMessage.role == 'assistant',
+            AIChatMessage.feedback_rating == 2
+        ).order_by(desc(AIChatMessage.feedback_at)).limit(limit * 2).all()
+
+        for msg in positive_messages:
+            if len(examples) >= limit:
+                break
+
+            # Get the user query that preceded this response
+            user_query = db.query(AIChatMessage).filter(
+                AIChatMessage.conversation_id == msg.conversation_id,
+                AIChatMessage.id < msg.id,
+                AIChatMessage.role == 'user'
+            ).order_by(desc(AIChatMessage.id)).first()
+
+            if user_query:
+                # Extract company names mentioned (simple heuristic)
+                companies = self._extract_company_names(msg.content)
+
+                examples.append(FeedbackExample(
+                    query=user_query.content,
+                    response=msg.content,
+                    rating=2,
+                    feedback_comment=msg.feedback_comment,
+                    companies_mentioned=companies
+                ))
+
+        return examples
+
+    def _get_negative_examples(self, db: Session, limit: int = 3) -> List[FeedbackExample]:
+        """
+        Get negative feedback examples to learn what to avoid.
+
+        Args:
+            db: Database session
+            limit: Max examples to return
+
+        Returns:
+            List of FeedbackExample objects (with rating=1)
+        """
+        examples = []
+
+        # Query negative feedback (rating=1 = thumbs down)
+        negative_messages = db.query(AIChatMessage).filter(
+            AIChatMessage.role == 'assistant',
+            AIChatMessage.feedback_rating == 1
+        ).order_by(desc(AIChatMessage.feedback_at)).limit(limit).all()
+
+        for msg in negative_messages:
+            # Get the user query
+            user_query = db.query(AIChatMessage).filter(
+                AIChatMessage.conversation_id == msg.conversation_id,
+                AIChatMessage.id < msg.id,
+                AIChatMessage.role == 'user'
+            ).order_by(desc(AIChatMessage.id)).first()
+
+            if user_query:
+                examples.append(FeedbackExample(
+                    query=user_query.content,
+                    response=msg.content[:200] + "..." if len(msg.content) > 200 else msg.content,
+                    rating=1,
+                    feedback_comment=msg.feedback_comment
+                ))
+
+        return examples
+
+    def _get_feedback_stats(self, db: Session) -> Dict:
+        """
+        Get feedback statistics.
+
+        Args:
+            db: Database session
+
+        Returns:
+            Stats dict
+        """
+        total_responses = db.query(AIChatMessage).filter(
+            AIChatMessage.role == 'assistant'
+        ).count()
+
+        with_feedback = db.query(AIChatMessage).filter(
+            AIChatMessage.feedback_rating.isnot(None)
+        ).count()
+
+        positive_count = db.query(AIChatMessage).filter(
+            AIChatMessage.feedback_rating == 2
+        ).count()
+
+        negative_count = db.query(AIChatMessage).filter(
+            AIChatMessage.feedback_rating == 1
+        ).count()
+
+        return {
+            'total_responses': total_responses,
+            'with_feedback': with_feedback,
+            'positive_count': positive_count,
+            'negative_count': negative_count,
+            'feedback_rate': round(with_feedback / total_responses * 100, 1) if total_responses > 0 else 0,
+            'positive_rate': round(positive_count / with_feedback * 100, 1) if with_feedback > 0 else 0
+        }
+
+    def _extract_company_names(self, text: str) -> List[str]:
+        """
+        Extract company names from response text.
+
+        Simple heuristic: looks for **bold** text which typically marks company names.
+
+        Args:
+            text: Response text
+
+        Returns:
+            List of company names
+        """
+        import re
+        # Find text between ** markers (markdown bold)
+        pattern = r'\*\*([^*]+)\*\*'
+        matches = re.findall(pattern, text)
+        # Filter out non-company text (too short, common words)
+        return [m for m in matches if len(m) > 2 and not m.lower() in ['kontakt', 'tel', 'www', 'email']]
+
+    def format_for_prompt(self, context: Optional[Dict] = None) -> str:
+        """
+        Format learning context as text for inclusion in AI prompt.
+
+        Args:
+            context: Learning context (fetches if not provided)
+
+        Returns:
+            Formatted string for prompt injection
+        """
+        if context is None:
+            context = self.get_learning_context()
+
+        lines = []
+
+        # Add positive examples section
+        if context['positive_examples']:
+            lines.append("\n📚 PRZYKŁADY DOBRYCH ODPOWIEDZI (ucz się z nich):")
+            for i, ex in enumerate(context['positive_examples'][:3], 1):
+                lines.append(f"\nPrzykład {i}:")
+                lines.append(f"Pytanie: {ex.query}")
+                lines.append(f"Odpowiedź: {ex.response[:300]}{'...' if len(ex.response) > 300 else ''}")
+
+        # Add negative patterns to avoid
+        if context['negative_patterns']:
+            lines.append("\n\n⚠️ UNIKAJ takich odpowiedzi:")
+            for pattern in context['negative_patterns']:
+                lines.append(f"- {pattern}")
+
+        # Add guidance based on negative feedback
+        if context['negative_examples']:
+            lines.append("\n\n❌ Użytkownicy ocenili negatywnie odpowiedzi typu:")
+            for ex in context['negative_examples'][:2]:
+                if ex.feedback_comment:
+                    lines.append(f"- Pytanie: '{ex.query[:50]}...' - komentarz: {ex.feedback_comment}")
+
+        return '\n'.join(lines)
+
+    def invalidate_cache(self):
+        """Force cache refresh on next request"""
+        self._cache = None
+        self._cache_time = None
+
+    def record_feedback_used(self, message_id: int, examples_used: List[int]):
+        """
+        Record which feedback examples were used for a response.
+
+        This helps track the effectiveness of few-shot learning.
+
+        Args:
+            message_id: The AI response message ID
+            examples_used: List of message IDs that were used as examples
+        """
+        # TODO: Implement tracking in separate table for analytics
+        logger.info(f"Feedback examples {examples_used} used for message {message_id}")
+
+
+# Global service instance
+_feedback_service: Optional[FeedbackLearningService] = None
+
+
+def get_feedback_learning_service() -> FeedbackLearningService:
+    """Get or create global FeedbackLearningService instance"""
+    global _feedback_service
+    if _feedback_service is None:
+        _feedback_service = FeedbackLearningService()
+    return _feedback_service

nordabiz_chat.py

@@ -39,6 +39,13 @@ from database import (
     AIChatMessage
 )
 
+# Import feedback learning service for few-shot learning
+try:
+    from feedback_learning_service import get_feedback_learning_service
+    FEEDBACK_LEARNING_AVAILABLE = True
+except ImportError:
+    FEEDBACK_LEARNING_AVAILABLE = False
+
 
 class NordaBizChatEngine:
     """
@@ -452,6 +459,18 @@ class NordaBizChatEngine:
 - Odpowiadaj PO POLSKU
 """
 
+        # Add feedback-based learning context (few-shot examples)
+        if FEEDBACK_LEARNING_AVAILABLE:
+            try:
+                feedback_service = get_feedback_learning_service()
+                learning_context = feedback_service.format_for_prompt()
+                if learning_context:
+                    system_prompt += learning_context
+            except Exception as e:
+                # Don't fail if feedback learning has issues
+                import logging
+                logging.getLogger(__name__).warning(f"Feedback learning error: {e}")
+
         # Add ALL companies in compact JSON format
         if context.get('all_companies'):
             system_prompt += "\n\n🏢 PEŁNA BAZA FIRM (wybierz najlepsze):\n"

run_migration.py

@@ -64,7 +64,10 @@ GRANT ALL ON TABLE company_website_analysis TO nordabiz_app;
 
 def run_migration():
     print(f"Connecting to database...")
-    print(f"URL: {DATABASE_URL.replace('NordaBiz2025Secure', '****')}")
+    # Mask password in output
+    import re
+    masked_url = re.sub(r':([^:@]+)@', ':****@', DATABASE_URL)
+    print(f"URL: {masked_url}")
 
     try:
         conn = psycopg2.connect(DATABASE_URL)

templates/admin/chat_analytics.html

@@ -217,7 +217,142 @@
         {% endfor %}
     </ul>
     {% else %}
-    <p class="text-muted">Brak ocen - poproś użytkowników o feedback!</p>
+    <p class="text-muted">Brak ocen - popros uzytkownikow o feedback!</p>
     {% endif %}
 </div>
+
+<!-- AI Learning Status Section -->
+<div class="section">
+    <h2>Uczenie AI z feedbacku</h2>
+    <div id="learningStatus">
+        <p class="text-muted">Ladowanie statusu uczenia...</p>
+    </div>
+</div>
+
+<style>
+.learning-grid {
+    display: grid;
+    grid-template-columns: repeat(auto-fit, minmax(200px, 1fr));
+    gap: var(--spacing-lg);
+    margin-bottom: var(--spacing-lg);
+}
+.learning-card {
+    background: var(--background);
+    padding: var(--spacing-lg);
+    border-radius: var(--radius);
+    text-align: center;
+}
+.learning-card.active {
+    border-left: 4px solid var(--success);
+}
+.learning-card.seed {
+    border-left: 4px solid var(--warning);
+}
+.learning-value {
+    font-size: var(--font-size-2xl);
+    font-weight: 700;
+    color: var(--primary);
+}
+.learning-label {
+    font-size: var(--font-size-sm);
+    color: var(--text-secondary);
+}
+.example-card {
+    background: var(--background);
+    padding: var(--spacing-md);
+    border-radius: var(--radius);
+    margin-bottom: var(--spacing-md);
+    border-left: 3px solid var(--success);
+}
+.example-card.negative {
+    border-left-color: var(--error);
+}
+.example-query {
+    font-weight: 600;
+    color: var(--text-primary);
+    margin-bottom: var(--spacing-xs);
+}
+.example-response {
+    font-size: var(--font-size-sm);
+    color: var(--text-secondary);
+}
+</style>
+{% endblock %}
+
+{% block extra_js %}
+// Load AI Learning Status
+async function loadLearningStatus() {
+    try {
+        const response = await fetch('/api/admin/ai-learning-status');
+        const data = await response.json();
+
+        if (!data.success) {
+            document.getElementById('learningStatus').innerHTML =
+                '<p class="text-muted">Blad ladowania statusu</p>';
+            return;
+        }
+
+        if (!data.learning_active) {
+            document.getElementById('learningStatus').innerHTML =
+                '<p class="text-muted">Uczenie z feedbacku nieaktywne</p>';
+            return;
+        }
+
+        const stats = data.stats || {};
+        const usingSeed = data.using_seed_examples;
+
+        let html = `
+            <div class="learning-grid">
+                <div class="learning-card ${usingSeed ? 'seed' : 'active'}">
+                    <div class="learning-value">${usingSeed ? 'Seed' : 'Aktywne'}</div>
+                    <div class="learning-label">${usingSeed ? 'Uzywa przykladow startowych' : 'Uczy sie z feedbacku'}</div>
+                </div>
+                <div class="learning-card">
+                    <div class="learning-value">${data.positive_examples_count}</div>
+                    <div class="learning-label">Pozytywnych przykladow</div>
+                </div>
+                <div class="learning-card">
+                    <div class="learning-value">${stats.feedback_rate || 0}%</div>
+                    <div class="learning-label">Wskaznik feedbacku</div>
+                </div>
+                <div class="learning-card">
+                    <div class="learning-value">${stats.positive_rate || 0}%</div>
+                    <div class="learning-label">Pozytywnych ocen</div>
+                </div>
+            </div>
+        `;
+
+        // Show positive examples
+        if (data.positive_examples && data.positive_examples.length > 0) {
+            html += '<h3 style="margin: var(--spacing-lg) 0 var(--spacing-md);">Przyklady uzywane do nauki</h3>';
+            for (const ex of data.positive_examples.slice(0, 3)) {
+                html += `
+                    <div class="example-card">
+                        <div class="example-query">Q: ${ex.query}</div>
+                        <div class="example-response">${ex.response}</div>
+                    </div>
+                `;
+            }
+        }
+
+        // Show patterns to avoid
+        if (data.negative_patterns && data.negative_patterns.length > 0) {
+            html += '<h3 style="margin: var(--spacing-lg) 0 var(--spacing-md);">Wzorce do unikania</h3>';
+            html += '<ul style="color: var(--error); font-size: var(--font-size-sm);">';
+            for (const pattern of data.negative_patterns) {
+                html += `<li>${pattern}</li>`;
+            }
+            html += '</ul>';
+        }
+
+        document.getElementById('learningStatus').innerHTML = html;
+    } catch (error) {
+        console.error('Error loading learning status:', error);
+        document.getElementById('learningStatus').innerHTML =
+            '<p class="text-muted">Blad ladowania statusu</p>';
+    }
+}
+
+// Load on page load
+loadLearningStatus();
 {% endblock %}