Memory

给 DeepSeek Harness 补上时间轴记忆：定宽只追加的事实日志 + agent 自己维护的摘要金字塔，固定阅读预算下 recall 成本不随记忆量增长。

A structured work-state protocol for DeepSeek Harness agent handoffs

Native DeepSeek Harness (dsh) plugin: hybrid local and XMemo cloud memory for Cordis-based agents.

Evidence-backed, human-approved project memory for DeepSeek Harness

Automatic cross-session GoodMemory recall and writeback for DeepSeek Harness

DSH plugin: automatic dual-peer Honcho v3 memory, contextual recall, search, reasoning, conclusions, and health diagnostics

Auto-supplement QMD semantic search when the model greps/globs a knowledge-base directory: zero-dependency DSH plugin, async injection via next-step, no manual invocation.

Long-term memory injected at session start: preferences/facts/summaries/knowledge in global and per-project scopes, budgeted recall via `agent/pre-step`, durable JSON store.

Brings your existing Claude / ChatGPT history into DSH: instant zero-token import of the memories.json inside a Claude export, local Claude Code transcripts with no export needed, and conversations distilled through the model you already configured in DSH. Every memory carries verbatim evidence checked against the source by code — a paraphrase is dropped and counted, and evidence found only in the assistant turns is always routed to review rather than stored.

Project memory for coding agents on DeepSeek Harness: similar past edits surface as precedents, out-of-scope edits are blocked, and cognition persists across sessions (47 upstream-verified fixes, gold-free 20-task run).

Hana-style assistant manager for DSH: visual preset management (cards, yuan personalities, avatar cropping) plus a full Hana memory stack (rolling summaries, layered compilation, SQLite fact store, deep memory).

Cross-session event-cue memory for DeepSeek Harness: auto-extracts task/conversation cues and user language habits from every turn into a persistent global cue bank, then re-injects remembered context on topic-switch detection (keyword or vector matching).

Rewrites each request against a learned prompt profile and automatically updates habits, anti-hallucination rules, and learning records.

Project-scoped persistent key-value memories with set/get/list/delete/search model tools and JSON-file storage.

Focus board for DeepSeek Harness agents: durable, model-maintained notes in the session workspace that pin the objective, constraints, and decisions across compaction and sessions, with automatic context injection, archive on clear, and an optional web panel.

Self-maintaining handoff memory per working directory and git branch: records turns, folds them into concise Markdown, and injects the result into future sessions from ~/.agent/agent-handoff, byte-compatible with pi-handoff.

Local Markdown wiki as long-term memory for DeepSeek Harness — ported from llmwiki

Persistent long-term persona memory for DeepSeek Harness: MEMORY.md/USER.md storage, memory + memory_search tools, per-request context injection, and secret scanning. Ported in spirit from pi-hermes-memory.

A DeepSeek Harness plugin that turns the anywhere-claude-mem git sync into a DSH peer: pull cross-machine memory at session start and push new prompts/observations as they land in the claude-mem worker

A DeepSeek Harness plugin that integrates claude-mem: query, context injection, manual save, and session lifecycle over a local claude-mem worker's HTTP API

Cross-session memory for the model: full-text search all past sessions (SQLite FTS5 via ctx.sessionQuery) and bring the strongest matching excerpts back into the current context.

Cited memory over DSH's lossless session log: distilled facts carry `(sessionId, eventRange)` citations that expand back to the exact original log excerpt.

Manage memory files written by DSH memory plugins inside Settings (view/edit/delete), and sync them to one cloud drive with conflict-safe three-way sync; other drives take upload-only backups.

Cross-session memory and compaction archives for dsh, plus a persona kernel. After each substantive turn a background model call extracts facts, preferences, open tasks and commitments; entries carry ids, fade when they stop being mentioned, are capped per kind and can be corrected. When the host compacts a conversation the compressed span is written to the note vault as an archive card whose catalog is injected into the session, so details can be read back by id. Ships an append-only ledger, a settings page, a /whale command group and nine tools.