Scorp1o117/dsh-tdai-memory
Agent memory for DeepSeek Harness | DeepSeek Harness memory plug-in
dsh-tdai-memory is a port of TencentDB Agent Memory (Tencent Cloud's open-source four-layer memory system, originally an OpenClaw plugin) into DeepSeek Harness: L0 conversation capture writes every turn (turn end, request boundary) to raw conversation storage (JSONL + SQLite + FTS + vectors); L1 structured memory runs a background pipeline that uses an LLM to extract facts/preferences/events (persona/episodic/instruction) from conversations, stored in records/ + SQLite + FTS + vectors; L2 scenes and L3 persona generate scene blocks and user profiles (pipeline-scheduled); and automatic recall injection adds relevant memory on every prompt assembly. It is part of the DeepSeek Harness Enhancement Suite (Vision · Soul/Persona · Long-term Memory · Plugin Marketplace).
Install
dsh plugin --profile web add dsh-tdai-memorynpm package dsh-tdai-memory 0.2.14 (registry-verified 2026-08-24). Standard profile bundle (dsh.bundle.patch): dsh plugin --profile web add dsh-tdai-memory installs and mounts it in one step — no manual cordis.patch.yml edits needed. Then restart dsh web. LLM/embedding API keys can be set in the Web UI settings page (记忆 / Memory) or directly in settings.yaml under tdai-memory:. Note: version 0.2.13+ requires DSH 0.1.0-rc.7 or newer (tested against rc.7, rc.8, and 0.1.1-rc.1); DSH 0.1.0-rc.6 users must pin dsh-tdai-memory@0.2.11. node-llama-cpp is an optional peer used only by the fully local embedding backend (not installed by default). Part of the DeepSeek Harness Enhancement Suite.
Compatibility
DeepSeek Harness 0.1.0-rc.7+ (0.2.13+ line; tested against rc.7, rc.8, 0.1.1-rc.1); rc.6 users pin 0.2.11. Standard profile bundle (dsh.bundle.patch) — installs and mounts in one step. DSH exposes the registered tdai-memory settings namespace directly; the plugin does not modify files in the host installation. Settings changes apply after a restart (TdaiCore is built at startup). Optional node-llama-cpp peer for the fully local embedding backend.
Details
- Repo: Scorp1o117/dsh-tdai-memory
- Category: Agent Capabilities
- Stars: 6
- Version: npm package dsh-tdai-memory 0.2.14 (registry-verified 2026-08-24)
- Last push: 2026-09-10
- First seen: 2026-08-13
Recent updates
The current README documents: the four-layer memory architecture (L0 conversation capture, L1 structured memory, L2 scenes, L3 persona), automatic recall injection, install (dsh plugin add, one-step bundle), settings namespace (tdai-memory), the rc.7+/0.2.13 version requirement with the rc.6 0.2.11 pin, and the optional node-llama-cpp local embedding backend.
FAQ
- Which DSH versions are supported?
- Version 0.2.13+ requires DSH 0.1.0-rc.7 or newer (tested against rc.7, rc.8, and 0.1.1-rc.1). DSH 0.1.0-rc.6 users must pin dsh-tdai-memory@0.2.11.
- Do I need to edit cordis.patch.yml manually?
- No — it is a standard profile bundle (dsh.bundle.patch): dsh plugin --profile web add dsh-tdai-memory installs and mounts it in one step.
- Can embeddings run fully locally?
- Yes — node-llama-cpp is an optional peer used only by the fully local embedding backend; it is intentionally not installed by default.
Alternatives
FuRongJun-1999/dsh-memory · hychiang-dsh-memory · dsh-engramory