Tencent Ships Fully-Local 4-Tier Agent Memory With Hard Benchmarks — No External API

GitHub·high signal

TencentCloud/TencentDB-Agent-Memory (+581 stars this week) delivers long-term agent memory through a local semantic pyramid — L0 Conversation → L1 Atom → L2 Scenario → L3 Persona — plus symbolic short-term memory that condenses tool logs into Mermaid symbols, with zero external API dependencies. Tencent reports that wiring it into OpenClaw cut token usage up to 61.38%, improved pass rate 51.52% (relative), and raised PersonaMem accuracy from 48% to 76%. Builder takeaway: this is a rare local-first memory layer that publishes concrete numbers instead of flat vector-store hand-waving.