""" Translation Agent Translates non-English comments to English using LLM """ from typing import Dict, Any import json from langchain_openai import ChatOpenAI from langchain.schema import HumanMessage, SystemMessage from agents.base_agent import BaseAgent import logging logger = logging.getLogger(__name__) class TranslationAgent(BaseAgent): """ Agent that translates text from source language to English. Uses LLM for high-quality, context-aware translation. """ def __init__(self, config: Dict[str, Any], api_key: str): """ Initialize the Translation Agent. Args: config: Configuration dictionary api_key: OpenAI API key """ super().__init__("TranslationAgent", config) self.api_key = api_key self.llm = ChatOpenAI( model=self.model, temperature=self.temperature, api_key=self.api_key ) def validate_input(self, input_data: Dict[str, Any]) -> bool: """ Validate that input contains required fields. Args: input_data: Input dictionary Returns: True if valid, False otherwise """ required_fields = ["comment_text", "is_english"] return all(field in input_data for field in required_fields) def translate_text(self, text: str, source_language: str) -> Dict[str, Any]: """ Translate text from source language to English using LLM. Args: text: Text to translate source_language: Source language name Returns: Dictionary with translation results """ system_prompt = """You are a professional translator specializing in social media content related to music and education. Translate the given text from the source language to English. The text is a comment on a musical content. Preserve the tone, intent, and any emojis or special characters. For informal social media language, maintain the casual tone in translation. Return your response in JSON format with the following fields: - translated_text: The English translation - translation_confidence: Your confidence level (high, medium, low) - notes: Any important notes about the translation (optional) """ user_prompt = f"""Translate this {source_language} comment to English: "{text}" Return JSON only.""" try: messages = [ SystemMessage(content=system_prompt), HumanMessage(content=user_prompt) ] response = self.llm.invoke(messages) result = self._parse_llm_json_response(response.content) return { "success": True, "translated_text": result.get("translated_text", text), "translation_confidence": result.get("translation_confidence", "medium"), "translation_notes": result.get("notes", "") } except json.JSONDecodeError as e: self.log_processing(f"JSON decode error: {str(e)}", "warning") # Try to extract text from response return { "success": False, "translated_text": text, "translation_confidence": "low", "translation_notes": "JSON parsing failed", "error": str(e) } except Exception as e: self.log_processing(f"Translation failed: {str(e)}", "error") return { "success": False, "translated_text": text, "translation_confidence": "low", "translation_notes": "Translation error", "error": str(e) } def process(self, input_data: Dict[str, Any]) -> Dict[str, Any]: """ Process comment and translate if needed. Args: input_data: Dictionary containing comment data with language info Returns: Dictionary with translation results """ try: # Validate input if not self.validate_input(input_data): return { "success": False, "error": "Invalid input: missing required fields", "translated_text": input_data.get("comment_text", ""), "translation_performed": False } comment_text = input_data["comment_text"] is_english = input_data["is_english"] source_language = input_data.get("language", "Unknown") # If already English, no translation needed if is_english: result = { "success": True, "translated_text": comment_text, "translation_performed": False, "translation_confidence": "N/A", "translation_notes": "Original text is English" } self.log_processing("Text is already English, skipping translation", "debug") else: # Perform translation self.log_processing( f"Translating from {source_language} to English", "debug" ) translation_result = self.translate_text(comment_text, source_language) result = { "success": translation_result.get("success", True), "translated_text": translation_result.get("translated_text", comment_text), "translation_performed": True, "translation_confidence": translation_result.get("translation_confidence", "medium"), "translation_notes": translation_result.get("translation_notes", "") } if "error" in translation_result: result["translation_error"] = translation_result["error"] # Preserve all original data for key, value in input_data.items(): if key not in result: result[key] = value return result except Exception as e: return self.handle_error(e, "translation") def _parse_llm_json_response(self, response_content: str) -> Dict[str, Any]: """ Parse LLM response that may contain JSON wrapped in markdown code blocks. Args: response_content: Raw response content from LLM Returns: Parsed JSON dictionary Raises: json.JSONDecodeError: If JSON cannot be parsed """ content = response_content.strip() # Check if response is wrapped in markdown code block if content.startswith("```json"): # Remove ```json prefix and ``` suffix content = content[7:] # Remove ```json if content.endswith("```"): content = content[:-3] # Remove trailing ``` content = content.strip() elif content.startswith("```"): # Remove generic ``` code block content = content[3:] if content.endswith("```"): content = content[:-3] content = content.strip() # Parse the cleaned JSON return json.loads(content)