dsh-memory-graph

A zero-dependency, git-backed micro-lesson library for AI agents to asynchronously share and search verified debugging experience. Python stdlib only, with MCP server, CLI, and DeepSeek Harness adapter.

Theory repository for the Common Trust Protocol (Intelligentics v3.4). README states the repository holds the protocol's theoretical layer, while the DeepSeek Harness plugin implementation lives in the separate repository FuRongJun-1999/dsh-memory (npm: @furongjun1999/dsh-memory). It describes the Lingshu AEIS engineering layer: an MCP server for trustworthy agent memory and knowledge graphs, five-layer memory, condition-routed knowledge cards, an eight-step self-iteration loop, roleplay engine,

Third-party context compaction engine for DeepSeek Harness (dsh). It uses a dual-engine design: Stage-1 per-atom compression where the LLM proposes extract/summary/false and deterministic guards decide, and Stage-2 lazy reference-graph pruning with 0 LLM calls. An append-only log is the source of truth, and recall_summary/recall_detail provide paginated recall. npm default mounts only Stage-2; Stage-1 is enabled via a peratom config block.

dsh-hub is a DSH plugin hub that integrates global memory, graph-memory, plugin market, and self-update checks into one settings-page plugin.