MCP Memory Service

#!/usr/bin/env python3 """Re-score all DeBERTa memories with corrected model.""" import asyncio import sys from pathlib import Path sys.path.insert(0, str(Path(__file__).parent.parent.parent)) # Use SQLite directly to avoid Cloudflare network timeouts from mcp_memory_service.storage.sqlite_vec import SqliteVecMemoryStorage from mcp_memory_service.config import SQLITE_VEC_PATH from mcp_memory_service.quality.onnx_ranker import get_onnx_ranker_model async def rescore(): print("Loading DeBERTa...") deberta = get_onnx_ranker_model('nvidia-quality-classifier-deberta', 'auto') print("Connecting to storage (SQLite-vec only, no network)...") storage = SqliteVecMemoryStorage(SQLITE_VEC_PATH) await storage.initialize() print("Fetching memories...") all_memories = await storage.get_all_memories() to_rescore = [m for m in all_memories if m.metadata and m.metadata.get('quality_provider') == 'onnx_deberta'] print(f"Re-scoring {len(to_rescore)} memories...") for i, m in enumerate(to_rescore, 1): new_score = deberta.score_quality("", m.content) await storage.update_memory_metadata( content_hash=m.content_hash, updates={'quality_score': new_score} ) if i % 100 == 0: print(f" [{i:5d}/{len(to_rescore)}] Score: {new_score:.3f}") print(f"\n✓ Re-scored {len(to_rescore)} memories") print("Note: Changes saved to SQLite. Hybrid backend will sync to Cloudflare automatically.") if __name__ == "__main__": asyncio.run(rescore())