dsh-linghun-memory
Linghun Memory
Linghun memory architecture: a complete package comprising the Closer Soul Card, fully ontologized memory, and the goal.js goal system, independent of the two original dsh-linghun lines. It provides structured event, fact, and rule entries; goal-signal merging with Bridge trigger terms; grouped nati
The plugin will be installed here. Keep web if you are unsure.
npx -y @deepseek-ai/dsh plugin --profile web add github:syyr1987/dsh-linghun-memory#b6ffbe315f188af0631b6aed02f333f2b186354bDescription
Linghun memory architecture—a complete package comprising the Closer Soul Card, fully ontologized memory, and the goal.js goal system, independent of the two original dsh-linghun lines: the write side provides three structured entry types (event memory ontology / fact knowledge ontology / rule rule ontology), plus goal-signal merging and Bridge trigger terms for recording beliefs, values, and bottom lines; the read side uses warm mode to natively render the three ontologies grouped by kind, with a goal activation score (0.25·domainHit + 0.35·bridgeHit + 0.40·sim) to add long-range semantic associations for goal/value; the v0.1.7 retrieval-graph design renders the write-side mechanisms into a four-axis memory map (context / goal / profile axes + evidence layer) and injects it into the read side, allowing the cognitive agent to trace back through the map and drill into evidence via memory_read/journal; chronological ledger traces, scenario archiving, and downweighting/forgetting of low-confidence items; a complete read/write consolidation loop through memory_append / memory_read / memory_consolidate + linghun_goal_update / linghun_goal_get / linghun_goal_state. LoCoMo-Plus 16/16 (official Memory Awareness Judge). Not a persona card, but judgment—add a closer to DeepSeek Harness.
