| """ |
| Base class for LLM provider implementations. |
| Supports both sync and async methods. |
| """ |
|
|
| from abc import ABC, abstractmethod |
| from typing import List, Dict, Any, Optional, Iterator, AsyncIterator |
| import logging |
| import asyncio |
|
|
| logger = logging.getLogger(__name__) |
|
|
|
|
| class BaseLLM(ABC): |
| """ |
| Abstract base class for LLM providers. |
| |
| All LLM providers must implement: |
| - generate(): Generate a complete response |
| - stream(): Stream response token by token (optional) |
| """ |
|
|
| @abstractmethod |
| def generate( |
| self, |
| user_prompt: str, |
| system_prompt: Optional[str] = None, |
| temperature: float = 0.7, |
| max_tokens: Optional[int] = None, |
| **kwargs |
| ) -> str: |
| """ |
| Generate a response from the LLM. |
| |
| Args: |
| user_prompt: The user's prompt, including any context. |
| system_prompt: System prompt to guide LLM behavior |
| temperature: Sampling temperature (0.0 - 1.0) |
| max_tokens: Maximum tokens to generate |
| **kwargs: Additional provider-specific parameters |
| |
| Returns: |
| Generated response as string |
| """ |
| pass |
|
|
| def stream( |
| self, |
| user_prompt: str, |
| system_prompt: Optional[str] = None, |
| temperature: float = 0.7, |
| max_tokens: Optional[int] = None, |
| **kwargs |
| ) -> Iterator[str]: |
| """ |
| Stream response token by token (optional). |
| |
| Args: |
| Same as generate() |
| |
| Yields: |
| Response tokens as they are generated |
| """ |
| |
| response = self.generate( |
| user_prompt=user_prompt, |
| system_prompt=system_prompt, |
| temperature=temperature, |
| max_tokens=max_tokens, |
| **kwargs |
| ) |
| yield response |
|
|
| def _get_default_system_prompt(self) -> str: |
| """ |
| Get default system prompt for RAG. |
| |
| Returns: |
| Default system prompt |
| """ |
| return """You are a helpful AI assistant. Answer the user's question based on the provided context. |
| If the context doesn't contain relevant information, say so. Be concise and accurate. |
| |
| Guidelines: |
| - Use information from the context to answer |
| - If unsure, acknowledge uncertainty |
| - Cite sources when possible |
| - Be concise but comprehensive""" |
|
|
| def _get_default_vietnamese_system_prompt(self) -> str: |
| """ |
| Get default Vietnamese system prompt. |
| |
| Returns: |
| Vietnamese system prompt |
| """ |
| return """Bạn là một trợ lý AI hữu ích. Trả lời câu hỏi của người dùng dựa trên ngữ cảnh được cung cấp. |
| Nếu ngữ cảnh không chứa thông tin liên quan, hãy nói rõ. Hãy ngắn gọn và chính xác. |
| |
| Hướng dẫn: |
| - Sử dụng thông tin từ ngữ cảnh để trả lời |
| - Nếu không chắc chắn, hãy thừa nhận điều đó |
| - Trích dẫn nguồn khi có thể |
| - Ngắn gọn nhưng đầy đủ""" |
|
|
| @abstractmethod |
| def get_model_info(self) -> Dict[str, Any]: |
| """ |
| Get information about the LLM model. |
| |
| Returns: |
| Dictionary with model information |
| """ |
| pass |
|
|
| |
|
|
| async def agenerate( |
| self, |
| user_prompt: str, |
| system_prompt: Optional[str] = None, |
| temperature: float = 0.7, |
| max_tokens: Optional[int] = None, |
| **kwargs |
| ) -> str: |
| """ |
| Async version of generate. |
| Default implementation wraps sync generate in thread executor. |
| |
| Override this method for native async implementation. |
| |
| Args: |
| Same as generate() |
| |
| Returns: |
| Generated response as string |
| """ |
| return await asyncio.to_thread( |
| self.generate, |
| user_prompt, |
| system_prompt, |
| temperature, |
| max_tokens, |
| **kwargs |
| ) |
|
|
| async def astream( |
| self, |
| user_prompt: str, |
| system_prompt: Optional[str] = None, |
| temperature: float = 0.7, |
| max_tokens: Optional[int] = None, |
| **kwargs |
| ) -> AsyncIterator[str]: |
| """ |
| Async stream response token by token. |
| Default implementation wraps sync stream. |
| |
| Override this method for native async streaming. |
| |
| Args: |
| Same as generate() |
| |
| Yields: |
| Response tokens as they are generated |
| """ |
| |
| for token in self.stream( |
| user_prompt, system_prompt, temperature, max_tokens, **kwargs |
| ): |
| yield token |
