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

isa_model/inference/services/llm/ollama_llm_service.py

@@ -1,99 +1,288 @@
 import logging
-from typing import Dict, Any, List, Union, AsyncGenerator, Optional
-from isa_model.inference.services.base_service import BaseLLMService
+import httpx
+import json
+from typing import Dict, Any, List, Union, AsyncGenerator, Optional, Callable
+from isa_model.inference.services.llm.base_llm_service import BaseLLMService
 from isa_model.inference.providers.base_provider import BaseProvider
 
 logger = logging.getLogger(__name__)
 
 class OllamaLLMService(BaseLLMService):
-    """Ollama LLM service using backend client"""
+    """Ollama LLM service with unified invoke interface"""
     
-    def __init__(self, provider: 'BaseProvider', model_name: str = "llama3.1"):
+    def __init__(self, provider: 'BaseProvider', model_name: str = "llama3.2:3b-instruct-fp16"):
         super().__init__(provider, model_name)
+        
+        # Create HTTP client for Ollama API
+        base_url = self.config.get("base_url", "http://localhost:11434")
+        timeout = self.config.get("timeout", 60)
+        
+        self.client = httpx.AsyncClient(
+            base_url=base_url,
+            timeout=timeout
+        )
             
         self.last_token_usage = {"prompt_tokens": 0, "completion_tokens": 0, "total_tokens": 0}
-        logger.info(f"Initialized OllamaLLMService with model {model_name}")
+        self.total_token_usage = {"prompt_tokens": 0, "completion_tokens": 0, "total_tokens": 0, "requests_count": 0}
+        
+        # Tool binding attributes
+        self._bound_tools: List[Any] = []
+        self._tool_binding_kwargs: Dict[str, Any] = {}
+        self._tool_functions: Dict[str, Callable] = {}
+        
+        logger.info(f"Initialized OllamaLLMService with model {model_name} at {base_url}")
     
-    async def ainvoke(self, prompt: Union[str, List[Dict[str, str]], Any]):
-        """Universal invocation method"""
-        if isinstance(prompt, str):
-            return await self.acompletion(prompt)
-        elif isinstance(prompt, list):
-            return await self.achat(prompt)
-        else:
-            raise ValueError("Prompt must be string or list of messages")
+    def _ensure_client(self):
+        """Ensure the HTTP client is available and not closed"""
+        if not hasattr(self, 'client') or not self.client or self.client.is_closed:
+            base_url = self.config.get("base_url", "http://localhost:11434")
+            timeout = self.config.get("timeout", 60)
+            self.client = httpx.AsyncClient(base_url=base_url, timeout=timeout)
+    
+    def _create_bound_copy(self) -> 'OllamaLLMService':
+        """Create a copy of this service for tool binding"""
+        bound_service = OllamaLLMService(self.provider, self.model_name)
+        bound_service._bound_tools = self._bound_tools.copy()
+        bound_service._tool_binding_kwargs = self._tool_binding_kwargs.copy()
+        bound_service._tool_functions = self._tool_functions.copy()
+        return bound_service
     
-    async def achat(self, messages: List[Dict[str, str]]):
-        """Chat completion method"""
+    def bind_tools(self, tools: List[Union[Dict[str, Any], Callable]], **kwargs) -> 'OllamaLLMService':
+        """Bind tools to this LLM service for function calling"""
+        bound_service = self._create_bound_copy()
+        # 使用基类的适配器管理器方法
+        bound_service._bound_tools = tools
+        bound_service._tool_binding_kwargs = kwargs
+        
+        return bound_service
+    
+    async def ainvoke(self, input_data: Union[str, List[Dict[str, str]], Any]) -> Union[str, Any]:
+        """
+        Universal async invocation method that handles different input types
+        
+        Args:
+            input_data: Can be:
+                - str: Simple text prompt
+                - list: Message history like [{"role": "user", "content": "hello"}]
+                - Any: LangChain message objects or other formats
+            
+        Returns:
+            Model response (string for simple cases, object for complex cases)
+        """
         try:
+            # Ensure client is available
+            self._ensure_client()
+            
+            # Convert input to messages format
+            messages = self._prepare_messages(input_data)
+            
+            # Prepare request parameters
             payload = {
                 "model": self.model_name,
                 "messages": messages,
-                "stream": False
+                "stream": self.streaming,
+                "options": {
+                    "temperature": self.config.get("temperature", 0.7),
+                    "top_p": self.config.get("top_p", 0.9),
+                    "num_predict": self.config.get("max_tokens", 2048)
+                }
             }
-            response = await self.backend.post("/api/chat", payload)
             
-            # Update token usage if available
-            if "eval_count" in response:
-                self.last_token_usage = {
-                    "prompt_tokens": response.get("prompt_eval_count", 0),
-                    "completion_tokens": response.get("eval_count", 0),
-                    "total_tokens": response.get("prompt_eval_count", 0) + response.get("eval_count", 0)
-                }
+            # Add tools if bound using adapter manager
+            tool_schemas = await self._prepare_tools_for_request()
+            if tool_schemas:
+                payload["tools"] = tool_schemas
             
-            return response["message"]["content"]
+            # Handle streaming
+            if self.streaming:
+                return self._stream_response(payload)
             
-        except Exception as e:
-            logger.error(f"Error in chat completion: {e}")
-            raise
-    
-    async def acompletion(self, prompt: str):
-        """Text completion method"""
-        try:
-            payload = {
-                "model": self.model_name,
-                "prompt": prompt,
-                "stream": False
-            }
-            response = await self.backend.post("/api/generate", payload)
+            # Regular request
+            response = await self.client.post("/api/chat", json=payload)
+            response.raise_for_status()
+            result = response.json()
             
             # Update token usage if available
-            if "eval_count" in response:
-                self.last_token_usage = {
-                    "prompt_tokens": response.get("prompt_eval_count", 0),
-                    "completion_tokens": response.get("eval_count", 0),
-                    "total_tokens": response.get("prompt_eval_count", 0) + response.get("eval_count", 0)
-                }
+            if "eval_count" in result:
+                self._update_token_usage(result)
+            
+            # Handle tool calls if present
+            message = result["message"]
+            if "tool_calls" in message and message["tool_calls"]:
+                return await self._handle_tool_calls(message, messages)
             
-            return response["response"]
+            # Return appropriate format based on input type
+            return self._format_response(message["content"], input_data)
             
+        except httpx.RequestError as e:
+            logger.error(f"HTTP request error in ainvoke: {e}")
+            raise
         except Exception as e:
-            logger.error(f"Error in text completion: {e}")
+            logger.error(f"Error in ainvoke: {e}")
             raise
     
-    async def agenerate(self, messages: List[Dict[str, str]], n: int = 1) -> List[str]:
-        """Generate multiple completions"""
-        results = []
-        for _ in range(n):
-            result = await self.achat(messages)
-            results.append(result)
-        return results
+    def _prepare_messages(self, input_data: Union[str, List[Dict[str, str]], Any]) -> List[Dict[str, str]]:
+        """Convert various input formats to Ollama messages format (same as OpenAI)"""
+        if isinstance(input_data, str):
+            # Simple string prompt
+            return [{"role": "user", "content": input_data}]
+        
+        elif isinstance(input_data, list):
+            if not input_data:
+                raise ValueError("Empty message list provided")
+            
+            # Check if it's LangChain messages or standard messages
+            first_msg = input_data[0]
+            if hasattr(first_msg, 'content') and hasattr(first_msg, 'type'):
+                # LangChain message objects - use base class method
+                return self._convert_langchain_to_openai(input_data)
+            elif isinstance(first_msg, dict):
+                # Standard message dictionaries
+                return input_data
+            else:
+                # List of strings or other formats
+                messages = []
+                for i, msg in enumerate(input_data):
+                    if isinstance(msg, str):
+                        role = "user" if i % 2 == 0 else "assistant"
+                        messages.append({"role": role, "content": msg})
+                    elif isinstance(msg, dict):
+                        messages.append(msg)
+                    else:
+                        messages.append({"role": "user", "content": str(msg)})
+                return messages
+        
+        else:
+            # Handle single LangChain message objects or other objects
+            if hasattr(input_data, 'content') and hasattr(input_data, 'type'):
+                return self._convert_langchain_to_openai([input_data])
+            else:
+                return [{"role": "user", "content": str(input_data)}]
     
-    async def astream_chat(self, messages: List[Dict[str, str]]) -> AsyncGenerator[str, None]:
-        """Stream chat responses"""
-        # Note: This would require modifying the backend to support streaming
-        # For now, return the full response
-        response = await self.achat(messages)
-        yield response
+    def _format_response(self, content: str, original_input: Any) -> Union[str, Any]:
+        """Format response based on original input type"""
+        # For LangGraph compatibility, return AIMessage object if needed
+        if hasattr(original_input, 'type') or (isinstance(original_input, list) and 
+                                               original_input and hasattr(original_input[0], 'type')):
+            try:
+                from langchain_core.messages import AIMessage
+                return AIMessage(content=content)
+            except ImportError:
+                # Fallback to simple object
+                class SimpleAIMessage:
+                    def __init__(self, content):
+                        self.content = content
+                        self.type = "ai"
+                return SimpleAIMessage(content)
+        
+        # Default to string
+        return content
     
-    def get_token_usage(self):
+    async def _stream_response(self, payload: Dict[str, Any]) -> AsyncGenerator[str, None]:
+        """Handle streaming responses"""
+        async def stream_generator():
+            try:
+                async with self.client.stream("POST", "/api/chat", json=payload) as response:
+                    response.raise_for_status()
+                    async for line in response.aiter_lines():
+                        if line.strip():
+                            try:
+                                chunk = json.loads(line)
+                                if "message" in chunk and "content" in chunk["message"]:
+                                    content = chunk["message"]["content"]
+                                    if content:
+                                        yield content
+                            except json.JSONDecodeError:
+                                continue
+            except Exception as e:
+                logger.error(f"Error in streaming: {e}")
+                raise
+        
+        return stream_generator()
+    
+    async def _handle_tool_calls(self, assistant_message: Dict[str, Any], original_messages: List[Dict[str, str]]) -> str:
+        """Handle tool calls from the assistant using adapter manager"""
+        tool_calls = assistant_message.get("tool_calls", [])
+        
+        # Add assistant message with tool calls to conversation
+        messages = original_messages + [assistant_message]
+        
+        # Execute each tool call using adapter manager
+        for tool_call in tool_calls:
+            function_name = tool_call["function"]["name"]
+            
+            try:
+                # Parse arguments if they're a string
+                arguments = tool_call["function"]["arguments"]
+                if isinstance(arguments, str):
+                    arguments = json.loads(arguments)
+                
+                # Use adapter manager to execute tool
+                result = await self._execute_tool_call(function_name, arguments)
+                
+                # Add tool result to messages
+                messages.append({
+                    "role": "tool",
+                    "content": str(result),
+                    "tool_call_id": tool_call.get("id", function_name)
+                })
+                
+            except Exception as e:
+                logger.error(f"Error executing tool {function_name}: {e}")
+                messages.append({
+                    "role": "tool",
+                    "content": f"Error executing {function_name}: {str(e)}",
+                    "tool_call_id": tool_call.get("id", function_name)
+                })
+        
+        # Get final response from the model
+        return await self.ainvoke(messages)
+    
+    def _update_token_usage(self, result: Dict[str, Any]):
+        """Update token usage statistics"""
+        self.last_token_usage = {
+            "prompt_tokens": result.get("prompt_eval_count", 0),
+            "completion_tokens": result.get("eval_count", 0),
+            "total_tokens": result.get("prompt_eval_count", 0) + result.get("eval_count", 0)
+        }
+        
+        # Update total usage
+        self.total_token_usage["prompt_tokens"] += self.last_token_usage["prompt_tokens"]
+        self.total_token_usage["completion_tokens"] += self.last_token_usage["completion_tokens"]
+        self.total_token_usage["total_tokens"] += self.last_token_usage["total_tokens"]
+        self.total_token_usage["requests_count"] += 1
+    
+    def get_token_usage(self) -> Dict[str, Any]:
         """Get total token usage statistics"""
-        return self.last_token_usage
+        return self.total_token_usage
     
     def get_last_token_usage(self) -> Dict[str, int]:
         """Get token usage from last request"""
         return self.last_token_usage
+    
+    def get_model_info(self) -> Dict[str, Any]:
+        """Get information about the current model"""
+        return {
+            "name": self.model_name,
+            "max_tokens": self.config.get("max_tokens", 2048),
+            "supports_streaming": True,
+            "supports_functions": True,
+            "provider": "ollama"
+        }
+    
+    def _has_bound_tools(self) -> bool:
+        """Check if this service has bound tools"""
+        return bool(self._bound_tools)
+    
+    def _get_bound_tools(self) -> List[Any]:
+        """Get the bound tools schema"""
+        return self._bound_tools
         
     async def close(self):
-        """Close the backend client"""
-        await self.backend.close() 
+        """Close the HTTP client"""
+        if hasattr(self, 'client') and self.client:
+            try:
+                if not self.client.is_closed:
+                    await self.client.aclose()
+            except Exception as e:
+                logger.warning(f"Error closing Ollama client: {e}")