isa-model 0.3.4__py3-none-any.whl → 0.3.6__py3-none-any.whl

isa_model/inference/ai_factory.py

@@ -3,27 +3,27 @@
 
 """
 Simplified AI Factory for creating inference services
-Uses the new service architecture with proper base classes and centralized API key management
+Uses the new unified service architecture with centralized managers
 """
 
-from typing import Dict, Type, Any, Optional, Tuple, List, TYPE_CHECKING, cast
+from typing import Dict, Any, Optional, TYPE_CHECKING
 import logging
-from isa_model.inference.providers.base_provider import BaseProvider
 from isa_model.inference.services.base_service import BaseService
-from isa_model.inference.base import ModelType
-from isa_model.inference.services.vision.base_vision_service import BaseVisionService
-from isa_model.inference.services.vision.base_image_gen_service import BaseImageGenService
+from isa_model.core.models.model_manager import ModelManager
+from isa_model.core.config import ConfigManager
 
 if TYPE_CHECKING:
     from isa_model.inference.services.audio.base_stt_service import BaseSTTService
     from isa_model.inference.services.audio.base_tts_service import BaseTTSService
+    from isa_model.inference.services.vision.base_vision_service import BaseVisionService
+    from isa_model.inference.services.img.base_image_gen_service import BaseImageGenService
 
 logger = logging.getLogger(__name__)
 
 class AIFactory:
     """
-    Simplified Factory for creating AI services with proper inheritance hierarchy
-    API key management is handled by individual providers
+    Modernized AI Factory using centralized ModelManager and ConfigManager
+    Provides unified interface with only 6 core methods: get_llm, get_vision, get_img, get_stt, get_tts, get_embed
     """
     
     _instance = None
@@ -37,239 +37,134 @@ class AIFactory:
     def __init__(self):
         """Initialize the AI Factory."""
         if not self._is_initialized:
-            self._providers: Dict[str, Type[BaseProvider]] = {}
-            self._services: Dict[Tuple[str, ModelType], Type[BaseService]] = {}
+            # Use centralized managers
+            self.model_manager = ModelManager()
+            self.config_manager = ConfigManager()
             self._cached_services: Dict[str, BaseService] = {}
-            self._initialize_services()
-            AIFactory._is_initialized = True
-    
-    def _initialize_services(self):
-        """Initialize available providers and services"""
-        try:
-            # Register Ollama services
-            self._register_ollama_services()
-            
-            # Register OpenAI services
-            self._register_openai_services()
-            
-            # Register Replicate services
-            self._register_replicate_services()
-            
-            logger.info("AI Factory initialized with centralized provider API key management")
-            
-        except Exception as e:
-            logger.error(f"Error initializing services: {e}")
-            logger.warning("Some services may not be available")
-    
-    def _register_ollama_services(self):
-        """Register Ollama provider and services"""
-        try:
-            from isa_model.inference.providers.ollama_provider import OllamaProvider
-            from isa_model.inference.services.llm.ollama_llm_service import OllamaLLMService
-            from isa_model.inference.services.embedding.ollama_embed_service import OllamaEmbedService
-            from isa_model.inference.services.vision.ollama_vision_service import OllamaVisionService
-            
-            self.register_provider('ollama', OllamaProvider)
-            self.register_service('ollama', ModelType.LLM, OllamaLLMService)
-            self.register_service('ollama', ModelType.EMBEDDING, OllamaEmbedService)
-            self.register_service('ollama', ModelType.VISION, OllamaVisionService)
-            
-            logger.info("Ollama services registered successfully")
-            
-        except ImportError as e:
-            logger.warning(f"Ollama services not available: {e}")
-    
-    def _register_openai_services(self):
-        """Register OpenAI provider and services"""
-        try:
-            from isa_model.inference.providers.openai_provider import OpenAIProvider
-            from isa_model.inference.services.llm.openai_llm_service import OpenAILLMService
-            from isa_model.inference.services.audio.openai_tts_service import OpenAITTSService
-            from isa_model.inference.services.audio.openai_stt_service import OpenAISTTService
-            from isa_model.inference.services.embedding.openai_embed_service import OpenAIEmbedService
-            from isa_model.inference.services.vision.openai_vision_service import OpenAIVisionService
-            
-            self.register_provider('openai', OpenAIProvider)
-            self.register_service('openai', ModelType.LLM, OpenAILLMService)
-            self.register_service('openai', ModelType.AUDIO, OpenAITTSService)
-            self.register_service('openai', ModelType.EMBEDDING, OpenAIEmbedService)
-            self.register_service('openai', ModelType.VISION, OpenAIVisionService)
-            
-            logger.info("OpenAI services registered successfully")
-            
-        except ImportError as e:
-            logger.warning(f"OpenAI services not available: {e}")
-    
-    def _register_replicate_services(self):
-        """Register Replicate provider and services"""
-        try:
-            from isa_model.inference.providers.replicate_provider import ReplicateProvider
-            from isa_model.inference.services.vision.replicate_image_gen_service import ReplicateImageGenService
-            from isa_model.inference.services.audio.replicate_tts_service import ReplicateTTSService
-            
-            self.register_provider('replicate', ReplicateProvider)
-            self.register_service('replicate', ModelType.VISION, ReplicateImageGenService)
-            self.register_service('replicate', ModelType.AUDIO, ReplicateTTSService)
-            
-            logger.info("Replicate services registered successfully")
-            
-        except ImportError as e:
-            logger.warning(f"Replicate services not available: {e}")
-    
-    def register_provider(self, name: str, provider_class: Type[BaseProvider]) -> None:
-        """Register an AI provider"""
-        self._providers[name] = provider_class
-    
-    def register_service(self, provider_name: str, model_type: ModelType, 
-                        service_class: Type[BaseService]) -> None:
-        """Register a service type with its provider"""
-        self._services[(provider_name, model_type)] = service_class
-    
-    def create_service(self, provider_name: str, model_type: ModelType, 
-                      model_name: str, config: Optional[Dict[str, Any]] = None) -> BaseService:
-        """Create a service instance with provider-managed configuration"""
-        try:
-            cache_key = f"{provider_name}_{model_type}_{model_name}"
-            
-            if cache_key in self._cached_services:
-                return self._cached_services[cache_key]
-            
-            # Get provider and service classes
-            provider_class = self._providers.get(provider_name)
-            service_class = self._services.get((provider_name, model_type))
-            
-            if not provider_class:
-                raise ValueError(f"No provider registered for '{provider_name}'")
-            
-            if not service_class:
-                raise ValueError(
-                    f"No service registered for provider '{provider_name}' and model type '{model_type}'"
-                )
-            
-            # Create provider with user config (provider handles .env loading)
-            provider = provider_class(config=config)
-            service = service_class(provider=provider, model_name=model_name)
-            
-            self._cached_services[cache_key] = service
-            return service
             
-        except Exception as e:
-            logger.error(f"Error creating service: {e}")
-            raise
+            logger.info("AI Factory initialized with centralized ModelManager and ConfigManager")
+            AIFactory._is_initialized = True
     
-    # Convenient methods for common services with updated defaults
-    def get_llm_service(self, model_name: Optional[str] = None, provider: Optional[str] = None,
-                       config: Optional[Dict[str, Any]] = None) -> BaseService:
+    # Core service methods using centralized architecture
+    def get_llm(self, model_name: Optional[str] = None, provider: Optional[str] = None,
+                config: Optional[Dict[str, Any]] = None) -> BaseService:
         """
         Get a LLM service instance with automatic defaults
         
         Args:
-            model_name: Name of the model to use (defaults: OpenAI="gpt-4.1-nano", Ollama="llama3.2:3b")
+            model_name: Name of the model to use (defaults: OpenAI="gpt-4.1-mini", Ollama="llama3.2:3b", YYDS="claude-sonnet-4-20250514")
             provider: Provider name (defaults to 'openai' for production, 'ollama' for dev)
-            config: Optional configuration dictionary (auto-loads from .env if not provided)
-                   Can include: streaming=True/False, temperature, max_tokens, etc.
+            config: Optional configuration dictionary
             
         Returns:
             LLM service instance
         """
         # Set defaults based on provider
         if provider == "openai":
-            final_model_name = model_name or "gpt-4.1-nano"
+            final_model_name = model_name or "gpt-4.1-mini"
             final_provider = provider
         elif provider == "ollama":
             final_model_name = model_name or "llama3.2:3b-instruct-fp16"
             final_provider = provider
+        elif provider == "yyds":
+            final_model_name = model_name or "claude-sonnet-4-20250514"
+            final_provider = provider
         else:
             # Default provider selection - OpenAI with cheapest model
             final_provider = provider or "openai"
             if final_provider == "openai":
-                final_model_name = model_name or "gpt-4.1-nano"
-            else:
+                final_model_name = model_name or "gpt-4.1-mini"
+            elif final_provider == "ollama":
                 final_model_name = model_name or "llama3.2:3b-instruct-fp16"
+            else:
+                final_model_name = model_name or "gpt-4.1-mini"
         
-        return self.create_service(final_provider, ModelType.LLM, final_model_name, config)
-    
-    def get_embedding_service(self, model_name: Optional[str] = None, provider: Optional[str] = None,
-                             config: Optional[Dict[str, Any]] = None) -> BaseService:
-        """
-        Get an embedding service instance with automatic defaults
-        
-        Args:
-            model_name: Name of the model to use (defaults: OpenAI="text-embedding-3-small", Ollama="bge-m3")
-            provider: Provider name (defaults to 'openai' for production, 'ollama' for dev)
-            config: Optional configuration dictionary (auto-loads from .env if not provided)
-            
-        Returns:
-            Embedding service instance
-        """
-        # Set defaults based on provider
-        if provider == "openai":
-            final_model_name = model_name or "text-embedding-3-small"
-            final_provider = provider
-        elif provider == "ollama":
-            final_model_name = model_name or "bge-m3"
-            final_provider = provider
-        else:
-            # Default provider selection
-            final_provider = provider or "openai"
+        # Create service using new centralized approach
+        try:
             if final_provider == "openai":
-                final_model_name = model_name or "text-embedding-3-small"
+                from isa_model.inference.services.llm.openai_llm_service import OpenAILLMService
+                return OpenAILLMService(provider_name=final_provider, model_name=final_model_name, 
+                                      model_manager=self.model_manager, config_manager=self.config_manager)
+            elif final_provider == "ollama":
+                from isa_model.inference.services.llm.ollama_llm_service import OllamaLLMService
+                return OllamaLLMService(provider_name=final_provider, model_name=final_model_name,
+                                      model_manager=self.model_manager, config_manager=self.config_manager)
+            elif final_provider == "yyds":
+                from isa_model.inference.services.llm.yyds_llm_service import YydsLLMService
+                return YydsLLMService(provider_name=final_provider, model_name=final_model_name,
+                                    model_manager=self.model_manager, config_manager=self.config_manager)
             else:
-                final_model_name = model_name or "bge-m3"
-        
-        return self.create_service(final_provider, ModelType.EMBEDDING, final_model_name, config)
+                raise ValueError(f"Unsupported LLM provider: {final_provider}")
+        except Exception as e:
+            logger.error(f"Failed to create LLM service: {e}")
+            raise
     
-    def get_vision_service(self, model_name: Optional[str] = None, provider: Optional[str] = None,
-                          config: Optional[Dict[str, Any]] = None) -> BaseVisionService:
+    def get_vision(
+        self,
+        model_name: Optional[str] = None,
+        provider: Optional[str] = None,
+        config: Optional[Dict[str, Any]] = None
+    ) -> 'BaseVisionService':
         """
-        Get a vision service instance with automatic defaults
+        Get vision service with automatic defaults
         
         Args:
-            model_name: Name of the model to use (defaults: OpenAI="gpt-4.1-mini", Ollama="gemma3:4b")
-            provider: Provider name (defaults to 'openai' for production, 'ollama' for dev)
-            config: Optional configuration dictionary (auto-loads from .env if not provided)
+            model_name: Model name. Special names:
+                       - "isa_vision_table": Table extraction service
+                       - "isa_vision_ui": UI detection service  
+                       - "isa_vision_doc": Document analysis service
+                       - Default: "gpt-4.1-mini"
+            provider: Provider name (auto-detected for ISA services)
+            config: Optional configuration override
             
         Returns:
             Vision service instance
         """
-        # Set defaults based on provider
+        # Handle special ISA vision services
+        if model_name in ["isa_vision_table", "isa_vision_ui", "isa_vision_doc"]:
+            try:
+                from isa_model.inference.services.vision.auto_deploy_vision_service import AutoDeployVisionService
+                logger.info(f"Creating auto-deploy service wrapper for {model_name}")
+                return AutoDeployVisionService(model_name, config)
+            except Exception as e:
+                logger.error(f"Failed to create ISA vision service: {e}")
+                raise
+        
+        # Set defaults for regular services
         if provider == "openai":
             final_model_name = model_name or "gpt-4.1-mini"
             final_provider = provider
         elif provider == "ollama":
             final_model_name = model_name or "llama3.2-vision:latest"
             final_provider = provider
+        elif provider == "replicate":
+            final_model_name = model_name or "meta/llama-2-70b-chat"
+            final_provider = provider
         else:
             # Default provider selection
             final_provider = provider or "openai"
             if final_provider == "openai":
                 final_model_name = model_name or "gpt-4.1-mini"
-            else:
+            elif final_provider == "ollama":
                 final_model_name = model_name or "llama3.2-vision:latest"
+            else:
+                final_model_name = model_name or "gpt-4.1-mini"
         
-        return cast(BaseVisionService, self.create_service(final_provider, ModelType.VISION, final_model_name, config))
-    
-    def get_image_generation_service(self, model_name: Optional[str] = None, provider: Optional[str] = None,
-                                   config: Optional[Dict[str, Any]] = None) -> 'BaseImageGenService':
-        """
-        Get an image generation service instance with automatic defaults
-        
-        Args:
-            model_name: Name of the model to use (defaults: "black-forest-labs/flux-schnell" for production)
-            provider: Provider name (defaults to 'replicate')
-            config: Optional configuration dictionary (auto-loads from .env if not provided)
-            
-        Returns:
-            Image generation service instance
-        """
-        # Set defaults based on provider
-        final_provider = provider or "replicate"
-        if final_provider == "replicate":
-            final_model_name = model_name or "black-forest-labs/flux-schnell"
-        else:
-            final_model_name = model_name or "black-forest-labs/flux-schnell"
-        
-        return cast('BaseImageGenService', self.create_service(final_provider, ModelType.VISION, final_model_name, config))
+        # Create service using new centralized approach
+        try:
+            if final_provider == "openai":
+                from isa_model.inference.services.vision.openai_vision_service import OpenAIVisionService
+                return OpenAIVisionService(provider_name=final_provider, model_name=final_model_name,
+                                         model_manager=self.model_manager, config_manager=self.config_manager)
+            elif final_provider == "replicate":
+                from isa_model.inference.services.vision.replicate_vision_service import ReplicateVisionService
+                return ReplicateVisionService(provider_name=final_provider, model_name=final_model_name,
+                                            model_manager=self.model_manager, config_manager=self.config_manager)
+            else:
+                raise ValueError(f"Unsupported vision provider: {final_provider}")
+        except Exception as e:
+            logger.error(f"Failed to create vision service: {e}")
+            raise
     
     def get_img(self, type: str = "t2i", model_name: Optional[str] = None, provider: Optional[str] = None,
                 config: Optional[Dict[str, Any]] = None) -> 'BaseImageGenService':
@@ -310,111 +205,138 @@ class AIFactory:
         else:
             raise ValueError(f"Unknown image generation type: {type}. Use 't2i' or 'i2i'")
         
-        return cast('BaseImageGenService', self.create_service(final_provider, ModelType.VISION, final_model_name, config))
-    
-    def get_audio_service(self, model_name: Optional[str] = None, provider: Optional[str] = None,
-                         config: Optional[Dict[str, Any]] = None) -> BaseService:
+        # Create service using new centralized architecture
+        try:
+            if final_provider == "replicate":
+                from isa_model.inference.services.img.replicate_image_gen_service import ReplicateImageGenService
+                return ReplicateImageGenService(provider_name=final_provider, model_name=final_model_name,
+                                              model_manager=self.model_manager, config_manager=self.config_manager)
+            else:
+                raise ValueError(f"Unsupported image generation provider: {final_provider}")
+        except Exception as e:
+            logger.error(f"Failed to create image generation service: {e}")
+            raise
+
+    def get_stt(self, model_name: Optional[str] = None, provider: Optional[str] = None,
+                config: Optional[Dict[str, Any]] = None) -> 'BaseSTTService':
         """
-        Get an audio service instance (TTS) with automatic defaults
+        Get Speech-to-Text service with automatic defaults
         
         Args:
-            model_name: Name of the model to use (defaults: OpenAI="tts-1")
+            model_name: Name of the model to use (defaults: "whisper-1")
             provider: Provider name (defaults to 'openai')
-            config: Optional configuration dictionary (auto-loads from .env if not provided)
+            config: Optional configuration dictionary
             
         Returns:
-            Audio service instance
+            STT service instance
         """
-        # Set defaults based on provider
+        # Set defaults
         final_provider = provider or "openai"
-        if final_provider == "openai":
-            final_model_name = model_name or "tts-1"
-        else:
-            final_model_name = model_name or "tts-1"
+        final_model_name = model_name or "whisper-1"
         
-        return self.create_service(final_provider, ModelType.AUDIO, final_model_name, config)
-    
-    def get_tts_service(self, model_name: Optional[str] = None, provider: Optional[str] = None,
-                       config: Optional[Dict[str, Any]] = None) -> 'BaseTTSService':
+        # Create service using new centralized approach
+        try:
+            if final_provider == "openai":
+                from isa_model.inference.services.audio.openai_stt_service import OpenAISTTService
+                return OpenAISTTService(provider_name=final_provider, model_name=final_model_name,
+                                      model_manager=self.model_manager, config_manager=self.config_manager)
+            else:
+                raise ValueError(f"Unsupported STT provider: {final_provider}")
+        except Exception as e:
+            logger.error(f"Failed to create STT service: {e}")
+            raise
+
+    def get_tts(self, model_name: Optional[str] = None, provider: Optional[str] = None,
+                config: Optional[Dict[str, Any]] = None) -> 'BaseTTSService':
         """
-        Get a Text-to-Speech service instance with automatic defaults
+        Get Text-to-Speech service with automatic defaults
         
         Args:
             model_name: Name of the model to use (defaults: Replicate="kokoro-82m", OpenAI="tts-1")
             provider: Provider name (defaults to 'replicate' for production, 'openai' for dev)
-            config: Optional configuration dictionary (auto-loads from .env if not provided)
+            config: Optional configuration dictionary
             
         Returns:
             TTS service instance
         """
         # Set defaults based on provider
         if provider == "replicate":
-            model_name = model_name or "kokoro-82m"
+            final_model_name = model_name or "kokoro-82m"
+            final_provider = provider
         elif provider == "openai":
-            model_name = model_name or "tts-1"
+            final_model_name = model_name or "tts-1"
+            final_provider = provider
         else:
             # Default provider selection
-            provider = provider or "replicate"
-            if provider == "replicate":
-                model_name = model_name or "kokoro-82m"
+            final_provider = provider or "replicate"
+            if final_provider == "replicate":
+                final_model_name = model_name or "kokoro-82m"
             else:
-                model_name = model_name or "tts-1"
+                final_model_name = model_name or "tts-1"
         
-        # Ensure model_name is never None
-        if model_name is None:
-            model_name = "tts-1"
-        
-        if provider == "replicate":
-            from isa_model.inference.services.audio.replicate_tts_service import ReplicateTTSService
-            from isa_model.inference.providers.replicate_provider import ReplicateProvider
-            
-            # Use full model name for Replicate
-            if model_name == "kokoro-82m":
-                model_name = "jaaari/kokoro-82m:f559560eb822dc509045f3921a1921234918b91739db4bf3daab2169b71c7a13"
-            
-            provider_instance = ReplicateProvider(config=config)
-            return ReplicateTTSService(provider=provider_instance, model_name=model_name)
-        else:
-            return cast('BaseTTSService', self.get_audio_service(model_name, provider, config))
+        # Create service using new centralized approach
+        try:
+            if final_provider == "replicate":
+                from isa_model.inference.services.audio.replicate_tts_service import ReplicateTTSService
+                # Use full model name for Replicate
+                if final_model_name == "kokoro-82m":
+                    final_model_name = "jaaari/kokoro-82m:f559560eb822dc509045f3921a1921234918b91739db4bf3daab2169b71c7a13"
+                return ReplicateTTSService(provider_name=final_provider, model_name=final_model_name,
+                                         model_manager=self.model_manager, config_manager=self.config_manager)
+            elif final_provider == "openai":
+                from isa_model.inference.services.audio.openai_tts_service import OpenAITTSService
+                return OpenAITTSService(provider_name=final_provider, model_name=final_model_name,
+                                      model_manager=self.model_manager, config_manager=self.config_manager)
+            else:
+                raise ValueError(f"Unsupported TTS provider: {final_provider}")
+        except Exception as e:
+            logger.error(f"Failed to create TTS service: {e}")
+            raise
     
-    def get_stt_service(self, model_name: Optional[str] = None, provider: Optional[str] = None,
-                       config: Optional[Dict[str, Any]] = None) -> 'BaseSTTService':
+    def get_embed(self, model_name: Optional[str] = None, provider: Optional[str] = None,
+                     config: Optional[Dict[str, Any]] = None) -> BaseService:
         """
-        Get a Speech-to-Text service instance with automatic defaults
+        Get embedding service with automatic defaults
         
         Args:
-            model_name: Name of the model to use (defaults: "whisper-1")
-            provider: Provider name (defaults to 'openai')
-            config: Optional configuration dictionary (auto-loads from .env if not provided)
+            model_name: Name of the model to use (defaults: OpenAI="text-embedding-3-small", Ollama="bge-m3")
+            provider: Provider name (defaults to 'openai' for production)
+            config: Optional configuration dictionary
             
         Returns:
-            STT service instance
+            Embedding service instance
         """
         # Set defaults based on provider
-        provider = provider or "openai"
         if provider == "openai":
-            model_name = model_name or "whisper-1"
-        
-        # Ensure model_name is never None
-        if model_name is None:
-            model_name = "whisper-1"
-        
-        from isa_model.inference.services.audio.openai_stt_service import OpenAISTTService
-        from isa_model.inference.providers.openai_provider import OpenAIProvider
+            final_model_name = model_name or "text-embedding-3-small"
+            final_provider = provider
+        elif provider == "ollama":
+            final_model_name = model_name or "bge-m3"
+            final_provider = provider
+        else:
+            # Default provider selection
+            final_provider = provider or "openai"
+            if final_provider == "openai":
+                final_model_name = model_name or "text-embedding-3-small"
+            else:
+                final_model_name = model_name or "bge-m3"
         
-        # Create provider and service directly with config
-        provider_instance = OpenAIProvider(config=config)
-        return OpenAISTTService(provider=provider_instance, model_name=model_name)
-    
-    def get_available_services(self) -> Dict[str, List[str]]:
-        """Get information about available services"""
-        services = {}
-        for (provider, model_type), service_class in self._services.items():
-            if provider not in services:
-                services[provider] = []
-            services[provider].append(f"{model_type.value}: {service_class.__name__}")
-        return services
-    
+        # Create service using new centralized approach
+        try:
+            if final_provider == "openai":
+                from isa_model.inference.services.embedding.openai_embed_service import OpenAIEmbedService
+                return OpenAIEmbedService(provider_name=final_provider, model_name=final_model_name,
+                                        model_manager=self.model_manager, config_manager=self.config_manager)
+            elif final_provider == "ollama":
+                from isa_model.inference.services.embedding.ollama_embed_service import OllamaEmbedService
+                return OllamaEmbedService(provider_name=final_provider, model_name=final_model_name,
+                                        model_manager=self.model_manager, config_manager=self.config_manager)
+            else:
+                raise ValueError(f"Unsupported embedding provider: {final_provider}")
+        except Exception as e:
+            logger.error(f"Failed to create embedding service: {e}")
+            raise
+
     def clear_cache(self):
         """Clear the service cache"""
         self._cached_services.clear()
@@ -425,126 +347,4 @@ class AIFactory:
         """Get the singleton instance"""
         if cls._instance is None:
             cls._instance = cls()
-        return cls._instance
-    
-    # Alias method for cleaner API
-    def get_llm(self, model_name: Optional[str] = None, provider: Optional[str] = None,
-                config: Optional[Dict[str, Any]] = None) -> BaseService:
-        """
-        Alias for get_llm_service with cleaner naming
-        
-        Usage:
-            llm = AIFactory().get_llm()  # Uses gpt-4.1-nano by default
-            llm = AIFactory().get_llm(model_name="llama3.2", provider="ollama")
-            llm = AIFactory().get_llm(model_name="gpt-4.1-mini", provider="openai", config={"streaming": True})
-        """
-        return self.get_llm_service(model_name, provider, config)
-    
-    def get_embed(self, model_name: Optional[str] = None, provider: Optional[str] = None,
-                     config: Optional[Dict[str, Any]] = None) -> 'BaseEmbedService':
-        """
-        Get embedding service with automatic defaults
-        
-        Args:
-            model_name: Name of the model to use (defaults: OpenAI="text-embedding-3-small", Ollama="bge-m3")
-            provider: Provider name (defaults to 'openai' for production)
-            config: Optional configuration dictionary (auto-loads from .env if not provided)
-            
-        Returns:
-            Embedding service instance
-            
-        Usage:
-            # Default (OpenAI text-embedding-3-small)
-            embed = AIFactory().get_embed()
-            
-            # Custom model
-            embed = AIFactory().get_embed(model_name="text-embedding-3-large", provider="openai")
-            
-            # Development (Ollama)
-            embed = AIFactory().get_embed(provider="ollama")
-        """
-        return self.get_embedding_service(model_name, provider, config)
-
-    def get_stt(self, model_name: Optional[str] = None, provider: Optional[str] = None,
-                config: Optional[Dict[str, Any]] = None) -> 'BaseSTTService':
-        """
-        Get Speech-to-Text service with automatic defaults
-        
-        Args:
-            model_name: Name of the model to use (defaults: "whisper-1")
-            provider: Provider name (defaults to 'openai')
-            config: Optional configuration dictionary (auto-loads from .env if not provided)
-            
-        Returns:
-            STT service instance
-            
-        Usage:
-            # Default (OpenAI whisper-1)
-            stt = AIFactory().get_stt()
-            
-            # Custom configuration
-            stt = AIFactory().get_stt(model_name="whisper-1", provider="openai")
-        """
-        return self.get_stt_service(model_name, provider, config)
-
-    def get_tts(self, model_name: Optional[str] = None, provider: Optional[str] = None,
-                config: Optional[Dict[str, Any]] = None) -> 'BaseTTSService':
-        """
-        Get Text-to-Speech service with automatic defaults
-        
-        Args:
-            model_name: Name of the model to use (defaults: Replicate="kokoro-82m", OpenAI="tts-1")
-            provider: Provider name (defaults to 'replicate' for production, 'openai' for dev)
-            config: Optional configuration dictionary (auto-loads from .env if not provided)
-            
-        Returns:
-            TTS service instance
-            
-        Usage:
-            # Default (Replicate kokoro-82m)
-            tts = AIFactory().get_tts()
-            
-            # Development (OpenAI tts-1)
-            tts = AIFactory().get_tts(provider="openai")
-            
-            # Custom model
-            tts = AIFactory().get_tts(model_name="tts-1-hd", provider="openai")
-        """
-        return self.get_tts_service(model_name, provider, config)
-    
-    def get_vision_model(self, model_name: str, provider: str,
-                        config: Optional[Dict[str, Any]] = None) -> BaseService:
-        """Alias for get_vision_service and get_image_generation_service"""
-        if provider == "replicate":
-            return self.get_image_generation_service(model_name, provider, config)
-        else:
-            return self.get_vision_service(model_name, provider, config)
-    
-    def get_vision(
-        self,
-        model_name: Optional[str] = None,
-        provider: Optional[str] = None,
-        config: Optional[Dict[str, Any]] = None
-    ) -> 'BaseVisionService':
-        """
-        Get vision service with automatic defaults
-        
-        Args:
-            model_name: Model name (default: gpt-4.1-nano)
-            provider: Provider name (default: openai)
-            config: Optional configuration override
-            
-        Returns:
-            Vision service instance
-        """
-        # Set defaults
-        if provider is None:
-            provider = "openai"
-        if model_name is None:
-            model_name = "gpt-4.1-nano"
-        
-        return self.get_vision_service(
-            model_name=model_name,
-            provider=provider,
-            config=config
-        ) 
+        return cls._instance