Tencent open-sources Agent Memory: local long-term memory, 61% fewer tokens, 48%→76% accuracy
solyarisoftware · x · 2026-10-13
Tencent open-sourced TencentDB Agent Memory, giving AI agents long-term memory that runs 100% locally on plain SQLite with zero external API dependencies (5.1k GitHub stars). Key numbers: 61% fewer tokens, PersonaMem accuracy up from 48% to 76%, and a 51.52% relative pass-rate improvement when integrated with OpenClaw. Instead of compressing history into an opaque vector pile, it builds a semantic pyramid (L0 Conversation → L1 Atom → L2 Scenario → L3 Persona), encodes short-term state as Mermaid graphs in context, offloads verbose tool logs to disk, and lets the agent drill back via nodeid — lossless, fully readable markdown at every layer.
More from coding & agent
- Celesto CI undercuts GitHub Actions 9x: 1000 jobs for $6.67 vs $60 — aniketmaurya · 2026-10-13
- Developer says the Sketch MCP tooling is 'actually useful' in practice — nathanborror · 2026-10-13
- Pulse: open-source mission control for Claude Code with live agent map and local voice alerts — ThirdEyeDry · 2026-10-13
- Every trades personal agents for one shared company agent in Slack — every · 2026-10-13
- Dev shares prepended prompt that stops coding agents from bloating code — gabriberton · 2026-10-13
- One prompt orchestrates Haiku 5.5 builders and Astra reviewers into an autonomous PR loop on t3code nightly — talkaboutdesign · 2026-10-13