src/knowledge/engine.py

import sqlite3
import time
from typing import List, Dict, Any, Optional
from functools import lru_cache
from src.knowledge.repository import KnowledgeRepository

class KnowledgeEngine:
    def __init__(self, db_path: str = "data/knowledge/build/knowledge.db"):
        self.repo = KnowledgeRepository(db_path)
        self.repo.connect()

    @lru_cache(maxsize=1024)
    def get_entity_knowledge(self, entity_id: str) -> Optional[List[Dict[str, Any]]]:
        """
        Retrieves canonical visual traits for a given entity_id.
        Uses LRU Cache to maintain <1ms latency for repeated requests.
        """
        cursor = self.repo._conn.cursor()
        
        # We assume entity_id is already resolved to its canonical representation by the intent parser
        # Querying the visual traits
        cursor.execute("""
            SELECT t.trait_id, t.confidence, t.source_count, p.source_uri 
            FROM entity_visual_traits t
            LEFT JOIN trait_provenance p ON t.entity_id = p.entity_id AND t.trait_id = p.trait_id
            WHERE t.entity_id = ?
        """, (entity_id,))
        
        rows = cursor.fetchall()
        
        if not rows:
            return None
            
        traits = []
        for row in rows:
            traits.append({
                "trait_id": row["trait_id"],
                "confidence": row["confidence"],
                "source_count": row["source_count"],
                "provenance": row["source_uri"]
            })
            
        return traits

    def close(self):
        self.repo.close()