""" THE QS HUB - AI Memory Loop Handles user-specific and global AI classification rules. """ from utils.db import safe_query, safe_execute def record_mapping(user_id: int, original_text: str, mapped_category: str) -> bool: """ Records a mapping rule. Creates a 'personal' rule for the user, and a 'pending' rule for the Admin to review. """ original_text = original_text.strip().lower() mapped_category = mapped_category.strip() # Check if a global rule already exists global_df = safe_query( "SELECT id FROM qto_memory_rules WHERE status='global' AND original_text=%s", (original_text,) ) if not global_df.empty: return True # Already globally mapped # Check if this user already mapped it personal_df = safe_query( "SELECT id FROM qto_memory_rules WHERE user_id=%s AND status='personal' AND original_text=%s", (user_id, original_text) ) if personal_df.empty: # Create personal rule safe_execute( "INSERT INTO qto_memory_rules (user_id, original_text, mapped_category, status) VALUES (%s, %s, %s, 'personal')", (user_id, original_text, mapped_category) ) else: # Update personal rule safe_execute( "UPDATE qto_memory_rules SET mapped_category=%s WHERE user_id=%s AND status='personal' AND original_text=%s", (mapped_category, user_id, original_text) ) # Queue a pending rule for Admin (if one doesn't already exist for this exact mapping) pending_df = safe_query( "SELECT id FROM qto_memory_rules WHERE status='pending' AND original_text=%s AND mapped_category=%s", (original_text, mapped_category) ) if pending_df.empty: safe_execute( "INSERT INTO qto_memory_rules (user_id, original_text, mapped_category, status) VALUES (%s, %s, %s, 'pending')", (user_id, original_text, mapped_category) ) return True def get_active_rules(user_id: int = None) -> dict: """ Returns a dictionary of active rules. Global rules apply to everyone. Personal rules override global rules for this specific user. """ rules = {} # 1. Load global rules global_df = safe_query("SELECT original_text, mapped_category FROM qto_memory_rules WHERE status='global'") for _, row in global_df.iterrows(): rules[row['original_text']] = row['mapped_category'] # 2. Load personal rules (Overrides global if they conflict) if user_id: personal_df = safe_query( "SELECT original_text, mapped_category FROM qto_memory_rules WHERE user_id=%s AND status='personal'", (user_id,) ) for _, row in personal_df.iterrows(): rules[row['original_text']] = row['mapped_category'] return rules def format_rules_for_prompt(user_id: int) -> str: """ Formats the active rules into a string block to be injected into the AI Prompt. """ rules = get_active_rules(user_id) if not rules: return "" prompt_lines = ["\nCRITICAL USER MAPPING RULES:"] prompt_lines.append("You MUST apply these specific mappings if you encounter the exact original text (case-insensitive):") for original, mapped in rules.items(): prompt_lines.append(f"- If you see '{original}', classify it as '{mapped}'") return "\n".join(prompt_lines) + "\n"