isa-model 0.2.0__py3-none-any.whl → 0.3.1__py3-none-any.whl

isa_model/inference/ai_factory.py

@@ -1,17 +1,29 @@
-from typing import Dict, Type, Any, Optional, Tuple
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+
+"""
+Simplified AI Factory for creating inference services
+Uses the new service architecture with proper base classes and centralized API key management
+"""
+
+from typing import Dict, Type, Any, Optional, Tuple, List, TYPE_CHECKING, cast
 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
-import os
+from isa_model.inference.services.vision.base_vision_service import BaseVisionService
+from isa_model.inference.services.vision.base_image_gen_service import BaseImageGenService
+
+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
 
-# 设置基本的日志配置
-logging.basicConfig(level=logging.INFO)
 logger = logging.getLogger(__name__)
 
 class AIFactory:
     """
-    Factory for creating AI services based on the Single Model pattern.
+    Simplified Factory for creating AI services with proper inheritance hierarchy
+    API key management is handled by individual providers
     """
     
     _instance = None
@@ -24,72 +36,85 @@ class AIFactory:
     
     def __init__(self):
         """Initialize the AI Factory."""
-        self.triton_url = os.environ.get("TRITON_URL", "http://localhost:8000")
-        
-        # Cache for services (singleton pattern)
-        self._llm_services = {}
-        self._embedding_services = {}
-        self._speech_services = {}
-        
         if not self._is_initialized:
             self._providers: Dict[str, Type[BaseProvider]] = {}
             self._services: Dict[Tuple[str, ModelType], Type[BaseService]] = {}
             self._cached_services: Dict[str, BaseService] = {}
-            self._initialize_defaults()
+            self._initialize_services()
             AIFactory._is_initialized = True
     
-    def _initialize_defaults(self):
-        """Initialize default providers and services"""
+    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:
-            # Import providers and services
             from isa_model.inference.providers.ollama_provider import OllamaProvider
-            from isa_model.inference.services.embedding.ollama_embed_service import OllamaEmbedService
             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
             
-            # Register Ollama provider and services
             self.register_provider('ollama', OllamaProvider)
-            self.register_service('ollama', ModelType.EMBEDDING, OllamaEmbedService)
             self.register_service('ollama', ModelType.LLM, OllamaLLMService)
+            self.register_service('ollama', ModelType.EMBEDDING, OllamaEmbedService)
+            self.register_service('ollama', ModelType.VISION, OllamaVisionService)
             
-            # 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
-                
-                self.register_provider('openai', OpenAIProvider)
-                self.register_service('openai', ModelType.LLM, OpenAILLMService)
-                logger.info("OpenAI services registered successfully")
-            except ImportError as e:
-                logger.warning(f"OpenAI services not available: {e}")
-            
-            # 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 ReplicateVisionService
-                
-                self.register_provider('replicate', ReplicateProvider)
-                self.register_service('replicate', ModelType.VISION, ReplicateVisionService)
-                logger.info("Replicate provider and vision service registered successfully")
-            except ImportError as e:
-                logger.warning(f"Replicate services not available: {e}")
-            except Exception as e:
-                logger.warning(f"Error registering Replicate services: {e}")
-            
-            # Try to register Triton services
-            try:
-                from isa_model.inference.providers.triton_provider import TritonProvider
-                
-                self.register_provider('triton', TritonProvider)
-                logger.info("Triton provider registered successfully")
-                
-            except ImportError as e:
-                logger.warning(f"Triton provider not available: {e}")
-            
-            logger.info("Default AI providers and services initialized with backend architecture")
-        except Exception as e:
-            logger.error(f"Error initializing default providers and services: {e}")
-            # Don't raise - allow factory to work even if some services fail to load
-            logger.warning("Some services may not be available due to import errors")
+            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"""
@@ -102,31 +127,27 @@ class AIFactory:
     
     def create_service(self, provider_name: str, model_type: ModelType, 
                       model_name: str, config: Optional[Dict[str, Any]] = None) -> BaseService:
-        """Create a service instance"""
+        """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]
             
-            # 基础配置
-            base_config = {
-                "log_level": "INFO"
-            }
-            
-            # 合并配置
-            service_config = {**base_config, **(config or {})}
-            
-            # 创建 provider 和 service
-            provider_class = self._providers[provider_name]
+            # 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}"
+                    f"No service registered for provider '{provider_name}' and model type '{model_type}'"
                 )
             
-            provider = provider_class(config=service_config)
+            # 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
@@ -136,224 +157,394 @@ class AIFactory:
             logger.error(f"Error creating service: {e}")
             raise
     
-    # Convenient methods for common services
-    def get_llm(self, model_name: str = "llama3.1", provider: str = "ollama",
-                config: Optional[Dict[str, Any]] = None, api_key: Optional[str] = None) -> BaseService:
+    # 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:
         """
-        Get a LLM service instance
+        Get a LLM service instance with automatic defaults
         
         Args:
-            model_name: Name of the model to use
-            provider: Provider name ('ollama', 'openai', 'replicate', etc.)
-            config: Optional configuration dictionary
-            api_key: Optional API key for the provider (OpenAI, Replicate, etc.)
+            model_name: Name of the model to use (defaults: OpenAI="gpt-4.1-nano", Ollama="llama3.2:3b")
+            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.
             
         Returns:
             LLM service instance
-            
-        Example:
-            # Using with API key directly
-            llm = AIFactory.get_instance().get_llm(
-                model_name="gpt-4o-mini", 
-                provider="openai", 
-                api_key="your-api-key-here"
-            )
-            
-            # Using without API key (will use environment variable)
-            llm = AIFactory.get_instance().get_llm(
-                model_name="gpt-4o-mini", 
-                provider="openai"
-            )
         """
+        # Set defaults based on provider
+        if provider == "openai":
+            final_model_name = model_name or "gpt-4.1-nano"
+            final_provider = provider
+        elif provider == "ollama":
+            final_model_name = model_name or "llama3.2:3b-instruct-fp16"
+            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 "llama3.2:3b-instruct-fp16"
         
-        # Special case for DeepSeek service
-        if model_name.lower() in ["deepseek", "deepseek-r1", "qwen3-8b"]:
-            if "deepseek" in self._cached_services:
-                return self._cached_services["deepseek"]
+        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
         
-        # Special case for Llama3-8B direct service
-        if model_name.lower() in ["llama3", "llama3-8b", "meta-llama-3"]:
-            if "llama3" in self._cached_services:
-                return self._cached_services["llama3"]
+        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"
+            if final_provider == "openai":
+                final_model_name = model_name or "text-embedding-3-small"
+            else:
+                final_model_name = model_name or "bge-m3"
         
-        basic_config: Dict[str, Any] = {
-            "temperature": 0
-        }
+        return self.create_service(final_provider, ModelType.EMBEDDING, final_model_name, config)
+    
+    def get_vision_service(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
         
-        # Add API key to config if provided
-        if api_key:
-            if provider == "openai":
-                basic_config["api_key"] = api_key
-            elif provider == "replicate":
-                basic_config["api_token"] = api_key
+        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)
+            
+        Returns:
+            Vision service instance
+        """
+        # Set defaults based on provider
+        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
+        else:
+            # Default provider selection
+            final_provider = provider or "openai"
+            if final_provider == "openai":
+                final_model_name = model_name or "gpt-4.1-mini"
             else:
-                logger.warning(f"API key provided but provider '{provider}' may not support it")
+                final_model_name = model_name or "llama3.2-vision:latest"
+        
+        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"
         
-        if config:
-            basic_config.update(config)
-        return self.create_service(provider, ModelType.LLM, model_name, basic_config)
+        return cast('BaseImageGenService', self.create_service(final_provider, ModelType.VISION, final_model_name, config))
     
-    def get_vision_model(self, model_name: str = "gemma3-4b", provider: str = "triton",
-                       config: Optional[Dict[str, Any]] = None, api_key: Optional[str] = None) -> BaseService:
+    def get_img(self, type: str = "t2i", model_name: Optional[str] = None, provider: Optional[str] = None,
+                config: Optional[Dict[str, Any]] = None) -> 'BaseImageGenService':
         """
-        Get a vision model service instance
+        Get an image generation service with type-specific defaults
         
         Args:
-            model_name: Name of the model to use
-            provider: Provider name ('openai', 'replicate', 'triton', etc.)
+            type: Image generation type:
+                  - "t2i" (text-to-image): Uses flux-schnell ($3 per 1000 images)
+                  - "i2i" (image-to-image): Uses flux-kontext-pro ($0.04 per image)
+            model_name: Optional model name override
+            provider: Provider name (defaults to 'replicate')
             config: Optional configuration dictionary
-            api_key: Optional API key for the provider (OpenAI, Replicate, etc.)
             
         Returns:
-            Vision service instance
+            Image generation service instance
             
-        Example:
-            # Using with API key directly
-            vision = AIFactory.get_instance().get_vision_model(
-                model_name="gpt-4o", 
-                provider="openai", 
-                api_key="your-api-key-here"
-            )
-            
-            # Using Replicate for image generation
-            image_gen = AIFactory.get_instance().get_vision_model(
-                model_name="stability-ai/sdxl", 
-                provider="replicate", 
-                api_key="your-replicate-token"
-            )
+        Usage:
+            # Text-to-image (default)
+            img_service = AIFactory().get_img()
+            img_service = AIFactory().get_img(type="t2i")
+            
+            # Image-to-image
+            img_service = AIFactory().get_img(type="i2i")
+            
+            # Custom model
+            img_service = AIFactory().get_img(type="t2i", model_name="custom-model")
         """
+        # Set defaults based on type
+        final_provider = provider or "replicate"
+        
+        if type == "t2i":
+            # Text-to-image: flux-schnell
+            final_model_name = model_name or "black-forest-labs/flux-schnell"
+        elif type == "i2i":
+            # Image-to-image: flux-kontext-pro
+            final_model_name = model_name or "black-forest-labs/flux-kontext-pro"
+        else:
+            raise ValueError(f"Unknown image generation type: {type}. Use 't2i' or 'i2i'")
         
-        # Special case for Gemma3-4B direct service
-        if model_name.lower() in ["gemma3", "gemma3-4b", "gemma3-vision"]:
-            if "gemma3" in self._cached_services:
-                return self._cached_services["gemma3"]
+        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:
+        """
+        Get an audio service instance (TTS) with automatic defaults
         
-        # Special case for Replicate's image generation models
-        if provider == "replicate" and "/" in model_name:
-            replicate_config: Dict[str, Any] = {
-                "guidance_scale": 7.5,
-                "num_inference_steps": 30
-            }
-            
-            # Add API key if provided
-            if api_key:
-                replicate_config["api_token"] = api_key
-            
-            if config:
-                replicate_config.update(config)
-            return self.create_service(provider, ModelType.VISION, model_name, replicate_config)
+        Args:
+            model_name: Name of the model to use (defaults: OpenAI="tts-1")
+            provider: Provider name (defaults to 'openai')
+            config: Optional configuration dictionary (auto-loads from .env if not provided)
+            
+        Returns:
+            Audio service instance
+        """
+        # Set defaults based on provider
+        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"
         
-        basic_config: Dict[str, Any] = {
-            "temperature": 0.7,
-            "max_new_tokens": 512
-        }
+        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':
+        """
+        Get a Text-to-Speech service instance with automatic defaults
         
-        # Add API key to config if provided
-        if api_key:
-            if provider == "openai":
-                basic_config["api_key"] = api_key
-            elif provider == "replicate":
-                basic_config["api_token"] = api_key
+        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
+        """
+        # Set defaults based on provider
+        if provider == "replicate":
+            model_name = model_name or "kokoro-82m"
+        elif provider == "openai":
+            model_name = model_name or "tts-1"
+        else:
+            # Default provider selection
+            provider = provider or "replicate"
+            if provider == "replicate":
+                model_name = model_name or "kokoro-82m"
             else:
-                logger.warning(f"API key provided but provider '{provider}' may not support it")
+                model_name = model_name or "tts-1"
         
-        if config:
-            basic_config.update(config)
-        return self.create_service(provider, ModelType.VISION, model_name, basic_config)
+        # 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))
     
-    def get_embedding(self, model_name: str = "bge-m3", provider: str = "ollama",
-                     config: Optional[Dict[str, Any]] = None) -> BaseService:
-        """Get an embedding service instance"""
-        return self.create_service(provider, ModelType.EMBEDDING, model_name, config)
+    def get_stt_service(self, model_name: Optional[str] = None, provider: Optional[str] = None,
+                       config: Optional[Dict[str, Any]] = None) -> 'BaseSTTService':
+        """
+        Get a Speech-to-Text service instance 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
+        """
+        # 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
+        
+        # Create provider and service directly with config
+        provider_instance = OpenAIProvider(config=config)
+        return OpenAISTTService(provider=provider_instance, model_name=model_name)
     
-    def get_rerank(self, model_name: str = "bge-m3", provider: str = "ollama",
-                   config: Optional[Dict[str, Any]] = None) -> BaseService:
-        """Get a rerank service instance"""
-        return self.create_service(provider, ModelType.RERANK, model_name, config)
+    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
     
-    def get_embed_service(self, model_name: str = "bge-m3", provider: str = "ollama",
-                         config: Optional[Dict[str, Any]] = None) -> BaseService:
-        """Get an embedding service instance"""
-        return self.get_embedding(model_name, provider, config)
+    def clear_cache(self):
+        """Clear the service cache"""
+        self._cached_services.clear()
+        logger.info("Service cache cleared")
     
-    def get_speech_model(self, model_name: str = "whisper_tiny", provider: str = "triton",
-                       config: Optional[Dict[str, Any]] = None) -> BaseService:
-        """Get a speech-to-text model service instance"""
-        
-        # Special case for Whisper Tiny direct service
-        if model_name.lower() in ["whisper", "whisper_tiny", "whisper-tiny"]:
-            if "whisper" in self._cached_services:
-                return self._cached_services["whisper"]
+    @classmethod
+    def get_instance(cls) -> '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
         
-        basic_config = {
-            "language": "en",
-            "task": "transcribe"
-        }
-        if config:
-            basic_config.update(config)
-        return self.create_service(provider, ModelType.AUDIO, model_name, basic_config)
+        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)
     
-    async def get_embedding_service(self, model_name: str) -> Any:
+    def get_embed(self, model_name: Optional[str] = None, provider: Optional[str] = None,
+                     config: Optional[Dict[str, Any]] = None) -> 'BaseEmbedService':
         """
-        Get an embedding service for the specified model.
+        Get embedding service with automatic defaults
         
         Args:
-            model_name: Name of the model
+            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")
         """
-        if model_name in self._embedding_services:
-            return self._embedding_services[model_name]
-        
-        else:
-            raise ValueError(f"Unsupported embedding model: {model_name}")
-    
-    async def get_speech_service(self, model_name: str) -> Any:
+        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 a speech service for the specified model.
+        Get Speech-to-Text service with automatic defaults
         
         Args:
-            model_name: Name of the model
+            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:
-            Speech service instance
+            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':
         """
-        if model_name in self._speech_services:
-            return self._speech_services[model_name]
+        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_model_info(self, model_type: Optional[str] = None) -> Dict[str, Any]:
+    def get_vision(
+        self,
+        model_name: Optional[str] = None,
+        provider: Optional[str] = None,
+        config: Optional[Dict[str, Any]] = None
+    ) -> 'BaseVisionService':
         """
-        Get information about available models.
+        Get vision service with automatic defaults
         
         Args:
-            model_type: Optional filter for model type
+            model_name: Model name (default: gpt-4.1-nano)
+            provider: Provider name (default: openai)
+            config: Optional configuration override
             
         Returns:
-            Dict of model information
+            Vision service instance
         """
-        models = {
-            "llm": [
-                {"name": "deepseek", "description": "DeepSeek-R1-0528-Qwen3-8B language model"},
-                {"name": "llama", "description": "Llama3-8B language model"},
-                {"name": "gemma", "description": "Gemma3-4B language model"}
-            ],
-            "embedding": [
-                {"name": "bge_embed", "description": "BGE-M3 text embedding model"}
-            ],
-            "speech": [
-                {"name": "whisper", "description": "Whisper-tiny speech-to-text model"}
-            ]
-        }
+        # Set defaults
+        if provider is None:
+            provider = "openai"
+        if model_name is None:
+            model_name = "gpt-4.1-nano"
         
-        if model_type:
-            return {model_type: models.get(model_type, [])}
-        return models
-    
-    @classmethod
-    def get_instance(cls) -> 'AIFactory':
-        """Get the singleton instance"""
-        if cls._instance is None:
-            cls._instance = cls()
-        return cls._instance
+        return self.get_vision_service(
+            model_name=model_name,
+            provider=provider,
+            config=config
+        )