gateway/modules/features/chatbot/aiCenterAdapter.py

# Copyright (c) 2025 Patrick Motsch
# All rights reserved.
"""
Adapter to use AI Center as a LangChain-compatible chat model.
Maps LangChain message format to AI Center requests and responses.
"""

import logging
from typing import Any, AsyncIterator, Iterator, List, Optional

from langchain_core.language_models.chat_models import BaseChatModel
from langchain_core.messages import (
    AIMessage,
    BaseMessage,
    HumanMessage,
    SystemMessage,
)
from langchain_core.outputs import ChatGeneration, ChatResult
from langchain_core.callbacks import AsyncCallbackHandlerForLLMRun, CallbackManagerForLLMRun

from modules.datamodels.datamodelAi import AiCallRequest, AiCallOptions, OperationTypeEnum, ProcessingModeEnum

logger = logging.getLogger(__name__)


class AICenterChatModel(BaseChatModel):
    """
    Adapter to use AI center as LangChain chat model.
    Converts LangChain messages to AI center format and back.
    """
    
    def __init__(
        self,
        services,
        system_prompt: str = "",
        temperature: float = 0.2,
        **kwargs
    ):
        """
        Initialize AI Center chat model adapter.
        
        Args:
            services: Services instance with AI access
            system_prompt: System prompt to use
            temperature: Temperature for AI calls
        """
        super().__init__(**kwargs)
        self.services = services
        self.system_prompt = system_prompt
        self.temperature = temperature
    
    @property
    def _llm_type(self) -> str:
        """Return identifier of LLM type."""
        return "ai_center"
    
    def _generate(
        self,
        messages: List[BaseMessage],
        stop: Optional[List[str]] = None,
        run_manager: Optional[CallbackManagerForLLMRun] = None,
        **kwargs: Any,
    ) -> ChatResult:
        """
        Synchronous generation - not supported, use async version.
        """
        raise NotImplementedError("Use async version: _agenerate")
    
    async def _agenerate(
        self,
        messages: List[BaseMessage],
        stop: Optional[List[str]] = None,
        run_manager: Optional[AsyncCallbackHandlerForLLMRun] = None,
        **kwargs: Any,
    ) -> ChatResult:
        """
        Generate chat response using AI center.
        
        Args:
            messages: List of LangChain messages
            stop: Optional list of stop sequences
            run_manager: Optional callback manager
            **kwargs: Additional arguments
            
        Returns:
            ChatResult with generated message
        """
        # Convert LangChain messages to AI center prompt format
        prompt_parts = []
        
        # Add system prompt if present
        if self.system_prompt:
            prompt_parts.append(self.system_prompt)
        
        # Convert messages to text format
        for msg in messages:
            if isinstance(msg, SystemMessage):
                # System messages are already in system_prompt or can be added here
                if not self.system_prompt:
                    prompt_parts.append(f"System: {msg.content}")
            elif isinstance(msg, HumanMessage):
                prompt_parts.append(f"User: {msg.content}")
            elif isinstance(msg, AIMessage):
                prompt_parts.append(f"Assistant: {msg.content}")
            else:
                # Generic message
                prompt_parts.append(str(msg.content))
        
        # Combine into single prompt
        full_prompt = "\n\n".join(prompt_parts)
        
        # Create AI center request
        ai_request = AiCallRequest(
            prompt=full_prompt,
            options=AiCallOptions(
                resultFormat="txt",
                operationType=OperationTypeEnum.DATA_ANALYSE,
                processingMode=ProcessingModeEnum.DETAILED,
                temperature=self.temperature
            )
        )
        
        # Call AI center
        try:
            await self.services.ai.ensureAiObjectsInitialized()
            ai_response = await self.services.ai.callAi(ai_request)
            
            # Extract content
            content = ai_response.content if hasattr(ai_response, 'content') else str(ai_response)
            
            # Create AIMessage from response
            ai_message = AIMessage(content=content)
            
            # Create ChatGeneration
            generation = ChatGeneration(message=ai_message)
            
            # Return ChatResult
            return ChatResult(generations=[generation])
            
        except Exception as e:
            logger.error(f"Error calling AI center: {e}", exc_info=True)
            # Return error message
            error_message = AIMessage(content=f"Error: {str(e)}")
            generation = ChatGeneration(message=error_message)
            return ChatResult(generations=[generation])
    
    async def astream(
        self,
        messages: List[BaseMessage],
        stop: Optional[List[str]] = None,
        run_manager: Optional[AsyncCallbackHandlerForLLMRun] = None,
        **kwargs: Any,
    ) -> AsyncIterator[BaseMessage]:
        """
        Stream chat response (not fully supported by AI center, returns single chunk).
        
        Args:
            messages: List of LangChain messages
            stop: Optional list of stop sequences
            run_manager: Optional callback manager
            **kwargs: Additional arguments
            
        Yields:
            BaseMessage chunks
        """
        # For now, just return the full response as a single chunk
        # TODO: Implement proper streaming if AI center supports it
        result = await self._agenerate(messages, stop, run_manager, **kwargs)
        if result.generations:
            yield result.generations[0].message