dsh-biomemory · Biomimetic Memory for DeepSeek Harness

中文文档 · English

A cross-session memory plugin for DeepSeek Harness (DSH), designed like a human brain: layered memory, graded approval, memory metabolism, fully transparent.

v0.5.2 (2026-08-20) — editable memories + conflict surfacing:

• Edit entries in place: memory action=update fp="..." text="..." (tool), /memory edit <fp> <new text> (command) and an Edit button on the Knowledge tab — metadata (pin/weight/layer) is preserved, the stale vector is cleared, and an UPDATE audit event is recorded; duplicates are rejected.
• Conflict surfacing: behavior memories that clash with user preferences are pinned to the top of memory action=list, the Knowledge tab (red badge + left border) and the frozen session snapshot (marked [冲突]) — so conflicts float up for you to judge and fix, instead of being silently auto-arbitrated away. Search results keep relevance order but still carry the conflict marker.
• Conflict arbitration in the Reflect tab: conflicts are listed at the top of the reflect view, each with inline Edit (large auto-growing editor) and Delete buttons, plus an Undo bar after deletion.
• Single-entry rollback: memory action=restore fp="..." (tool), /memory undo <fp> (command), POST /entries/restore (API) restore a deleted entry from the newest backup DB — metadata preserved, RESTORE audit event recorded.
• Reflect data source fixed to SQLite (was scanning the Markdown read-only backups — deleted entries resurrected on the next reflect).

v0.5.1 (2026-08-20) — pet bridge extracted:

• The DSH↔desktop-pet bridge (session/event state forwarding + approval panel) moved out to its own plugin dsh-whale-pet-bridge — biomemory now only handles memory (save notifications to the pet are kept).
• No functional changes to memory features.

v0.5 (2026-08-19) — SQLite + semantic retrieval:

• SQLite data layer (~/.dsh/biomemory/biomemory.db, Node 24 built-in node:sqlite, WAL mode, zero external deps) — L2/L3 structured entries + vector blobs + audit log
• Offline embedding model (bge-small-zh-v1.5, 512-dim, quantized ONNX ~24MB at ~/.dsh/models/) via transformers.js — pure JS, no native modules
• Three retrieval modes: exact (keyword) / semantic (vector) / hybrid (default, Reciprocal Rank Fusion per v0.5 design doc §3.4)
• Automatic Markdown migration: existing ~/.dsh/memory entries imported once on first boot (Markdown kept as read-only backup)
• Audit aggregation (P1-003): group by action / day / entry
• Dream checkpointing (P0-002): resume interrupted metabolism from the last checkpoint
• Graceful degradation: model unavailable → keyword retrieval only; memory features unaffected

Core features (unchanged from v0.4):

…