Upload groq_client.py

llm/groq_client.py

@@ -0,0 +1,163 @@
+"""Groq LLM Client for AI Recommendations"""
+import sys
+from pathlib import Path
+from typing import Optional
+
+sys.path.insert(0, str(Path(__file__).parent.parent))
+from config import GROQ_API_KEY, LLM_MODEL
+
+try:
+    from groq import Groq
+except ImportError:
+    Groq = None
+
+
+SYSTEM_PROMPT = """You are AniVerse AI, an expert anime and manga recommendation assistant.
+
+## YOUR CORE MISSION
+Provide HIGHLY RELEVANT, PRECISE recommendations. Quality over quantity. Every suggestion must directly address what the user is looking for.
+
+## RECOMMENDATION RULES
+1. **Match the Query Exactly**: If user asks for "dark fantasy", recommend dark fantasy - not action comedy.
+2. **Use Context Wisely**: Reference the "Relevant Anime/Manga" data provided. These are semantically matched to the query.
+3. **Explain Your Picks**: For EACH recommendation, give 1-2 sentences on WHY it fits the request.
+4. **Limit Recommendations**: Suggest 2-4 titles max per response. Be selective.
+5. **Format Clearly**: Use bold for titles, include scores and genres inline.
+
+## PERSONALIZATION (When User Profile Available)
+- Reference their high-rated titles: "Since you gave Attack on Titan a 9..."
+- Avoid genres from low-rated shows
+- Connect new suggestions to their favorites
+
+## RESPONSE FORMAT
+When recommending, use this structure:
+**[Title]** (★ score/10) - [Brief reason why this matches their request]
+
+## ACTION HANDLING
+1. **Verifying Actions**: You verify if an action (like adding to list) succeeded by checking the "=== ACTIONS EXECUTED ===" section in the context.
+2. **Success**: If an action is listed there, confirm it enthusiastically (e.g., "Done! Added X to your list").
+3. **Failure/No Action**: If the user asked for an action but it is NOT in the "=== ACTIONS EXECUTED ===" block, DO NOT say you did it. Instead, assume the backend failed to understand the command.
+   - Ask the user to try again with valid syntax: "Add [Title] to [completed/watching/planned]"
+   - Or "Rate [Title] [Score]"
+4. **Ambiguity**: If the user says "Add to list" without specifying which list, ask them to specify (Completed, Watching, Planned).
+
+Context about relevant titles will be provided below."""
+
+
+class GroqClient:
+    """Groq LLM client for AI-powered recommendations"""
+    
+    def __init__(self):
+        if not Groq:
+            raise ImportError("groq package not installed. Run: pip install groq")
+        
+        if not GROQ_API_KEY:
+            raise ValueError("GROQ_API_KEY not set. Add it to your .env file")
+        
+        self.client = Groq(api_key=GROQ_API_KEY)
+        self.model = LLM_MODEL
+    
+    def chat(
+        self,
+        user_message: str,
+        context: str = "",
+        history: list[dict] = None,
+        max_tokens: int = 1024
+    ) -> str:
+        """Send a chat message and get a response"""
+        messages = [{"role": "system", "content": SYSTEM_PROMPT}]
+        
+        # Add context if provided
+        if context:
+            messages.append({
+                "role": "system",
+                "content": f"Here is relevant anime data from our database:\n\n{context}"
+            })
+        
+        # Add conversation history
+        if history:
+            messages.extend(history)
+        
+        # Add current user message
+        messages.append({"role": "user", "content": user_message})
+        
+        # Call Groq API
+        response = self.client.chat.completions.create(
+            model=self.model,
+            messages=messages,
+            max_tokens=max_tokens,
+            temperature=0.7,
+        )
+        
+        return response.choices[0].message.content
+    
+    def summarize_reviews(
+        self,
+        reviews: list[str],
+        anime_title: str
+    ) -> dict:
+        """Summarize multiple reviews into pros/cons"""
+        reviews_text = "\n---\n".join(reviews[:10])  # Limit to 10 reviews
+        
+        prompt = f"""Analyze these reviews for "{anime_title}" and provide:
+1. Overall sentiment (positive/negative/mixed)
+2. Top 3 pros (things reviewers loved)
+3. Top 3 cons (things reviewers criticized)
+4. A 2-3 sentence summary
+5. Aspect scores (1-10) for: story, animation, characters, music, enjoyment
+
+Reviews:
+{reviews_text}
+
+Respond in JSON format:
+{{
+    "sentiment": "positive|negative|mixed",
+    "pros": ["pro1", "pro2", "pro3"],
+    "cons": ["con1", "con2", "con3"],
+    "summary": "...",
+    "aspects": {{"story": 8, "animation": 9, ...}}
+}}"""
+        
+        response = self.chat(prompt, max_tokens=512)
+        
+        # Parse JSON response (with fallback)
+        import json
+        try:
+            return json.loads(response)
+        except json.JSONDecodeError:
+            return {
+                "sentiment": "mixed",
+                "pros": [],
+                "cons": [],
+                "summary": response,
+                "aspects": {}
+            }
+    
+    def generate_recommendation_reason(
+        self,
+        user_query: str,
+        anime_data: dict
+    ) -> str:
+        """Generate a personalized reason why an anime matches the user's request"""
+        prompt = f"""The user asked: "{user_query}"
+
+This anime was matched:
+- Title: {anime_data.get('title', 'Unknown')}
+- Genres: {anime_data.get('genres', 'Unknown')}
+- Score: {anime_data.get('score', 'N/A')}
+
+In 1-2 sentences, explain why this anime matches what the user is looking for. Be specific about the connection."""
+        
+        return self.chat(prompt, max_tokens=150)
+
+
+# Singleton
+_client: Optional[GroqClient] = None
+
+
+def get_llm_client() -> GroqClient:
+    """Get or create LLM client instance"""
+    global _client
+    if _client is None:
+        _client = GroqClient()
+    return _client