Update chat_handler.py

chat_handler.py

@@ -18,6 +18,7 @@ from validation_engine import validate
 from session import session_store, Session
 from llm_interface import LLMInterface, SparkLLM, GPT4oLLM
 from config_provider import ConfigProvider
+from llm_factory import LLMFactory
 
 # ─────────────────────────  CONFIG ───────────────────────── #
 # Global config reference
@@ -66,129 +67,42 @@ def _safe_intent_parse(raw: str) -> tuple[str, str]:
 
 # ─────────────────────────  LLM SETUP ───────────────────────── #
 def setup_llm_provider():
-    """Initialize LLM provider based on internal_prompt config"""
+    """Initialize LLM provider based on config"""
     global llm_provider
     
-    cfg = ConfigProvider.get()
-    internal_prompt = cfg.global_config.internal_prompt
-    
-    if not internal_prompt:
-        log("⚠️ No internal_prompt configured, using default Spark")
-        # Get Spark token
-        spark_token = _get_spark_token()
-        if not spark_token:
-            log("❌ SPARK_TOKEN not found")
-            raise ValueError("SPARK_TOKEN not configured")
-        
-        spark_endpoint = str(cfg.global_config.spark_endpoint).rstrip("/")
-        work_mode = cfg.global_config.work_mode
-        
-        log(f"🔌 Initializing SparkLLM: {spark_endpoint}")
-        log(f"🔧 Work mode: {work_mode}")
-        
-        llm_provider = SparkLLM(
-            spark_endpoint=spark_endpoint,
-            spark_token=spark_token,
-            work_mode=work_mode
-        )
-        
-        log("✅ SparkLLM initialized")
-        return
-    
-    # Check if it's a GPT-4o config
-    if internal_prompt.get("provider") == "gpt-4o":
-        api_key = internal_prompt.get("api_key")
-        if not api_key:
-            if cfg.global_config.is_cloud_mode():
-                api_key = os.environ.get("OPENAI_API_KEY")
-            else:
-                from dotenv import load_dotenv
-                load_dotenv()
-                api_key = os.getenv("OPENAI_API_KEY")
-        
-        if not api_key:
-            raise ValueError("OpenAI API key not configured")
-        
-        llm_provider = GPT4oLLM(
-            api_key=api_key,
-            model=internal_prompt.get("model", "gpt-4o"),
-            max_tokens=internal_prompt.get("max_tokens", 4096),
-            temperature=internal_prompt.get("temperature", 0.7)
-        )
-        log("✅ GPT-4o LLM initialized")
-    else:
-        # Default to Spark
-        spark_token = _get_spark_token()
-        if not spark_token:
-            raise ValueError("SPARK_TOKEN not configured")
-        
-        spark_endpoint = str(cfg.global_config.spark_endpoint).rstrip("/")
-        work_mode = cfg.global_config.work_mode
-        
-        llm_provider = SparkLLM(
-            spark_endpoint=spark_endpoint,
-            spark_token=spark_token,
-            work_mode=work_mode
-        )
-        log("✅ SparkLLM initialized (via internal_prompt)")
-
-def _get_spark_token() -> Optional[str]:
-    """Get Spark token based on work mode"""
-    cfg = ConfigProvider.get()
-    
-    if cfg.global_config.is_cloud_mode():
-        # Cloud mode - use HuggingFace secrets
-        token = os.environ.get("SPARK_TOKEN")
-        if token:
-            log("🔑 Using SPARK_TOKEN from environment")
-        return token
-    else:
-        # On-premise mode - use .env file
-        from dotenv import load_dotenv
-        load_dotenv()
-        return os.getenv("SPARK_TOKEN")
+    try:
+        llm_provider = LLMFactory.create_provider()
+        log("✅ LLM provider initialized successfully")
+    except Exception as e:
+        log(f"❌ Failed to initialize LLM provider: {e}")
+        raise
 
 # ─────────────────────────  SPARK/LLM CALL ───────────────────────── #
-async def spark_generate(s: Session, prompt: str, user_msg: str) -> str:
-    """Call LLM (Spark or configured provider) with proper error handling"""
-    global llm_provider
-    
-    if llm_provider is None:
-        setup_llm_provider()
-    
+async def llm_generate(s: Session, prompt: str, user_msg: str) -> str:
+    """Call LLM provider with proper error handling"""
     try:
-        # Get version config from session
-        version = s.get_version_config()
-        if not version:
-            # Fallback: get from project config
-            project = next((p for p in cfg.projects if p.name == s.project_name), None)
-            if not project:
-                raise ValueError(f"Project not found: {s.project_name}")
-            version = next((v for v in project.versions if v.published), None)
-            if not version:
-                raise ValueError("No published version found")
-        
-        log(f"🚀 Calling LLM for session {s.session_id[:8]}...")
-        log(f"📋 Prompt preview (first 200 chars): {prompt[:200]}...")
-        
-        # Call the configured LLM provider
-        raw = await llm_provider.generate(
-            project_name=s.project_name,
-            user_input=user_msg,
+        # Get conversation context
+        context = [
+            {"role": msg["role"], "content": msg["content"]} 
+            for msg in s.chat_history[-10:]  # Last 10 messages
+        ]
+        
+        # Generate response
+        raw_response = await llm_provider.generate(
             system_prompt=prompt,
-            context=s.chat_history[-10:],
-            version_config=version
+            user_input=user_msg,
+            context=context
         )
         
-        log(f"🪄 LLM raw response: {raw[:120]!r}")
-        return raw
-            
+        log(f"📥 LLM response length: {len(raw_response)}")
+        return raw_response
+        
     except httpx.TimeoutException:
-        log(f"⏱️ LLM timeout for session {s.session_id[:8]}")
-        raise
+        log("⏱️ LLM timeout - returning fallback")
+        return "İsteğiniz zaman aşımına uğradı. Lütfen tekrar deneyin."
     except Exception as e:
-        log(f"❌ LLM error: {e}")
-        raise
+        log(f"❌ LLM error: {str(e)}")
+        return "Bir hata oluştu. Lütfen daha sonra tekrar deneyin."
 
 # ─────────────────────────  ALLOWED INTENTS ───────────────────────── #
 ALLOWED_INTENTS = {"flight-booking", "flight-info", "booking-cancel"}