LLM-driven remember/forget memory: after each turn a model decides what to keep (MemoryDirector), injects relevant memories before each step, dedups, and persists cross-session.
Install
# from GitHub (first run asks for allowBuilds approval — follow the hint, retry)
dsh plugin --profile web add github:ljsysfurryACE/dsh-memory-director
Any plugin you install runs third-party code with your own permissions — it can read your files, use your credentials, and reach the network, and tool approvals don’t sandbox it. GitHub-sourced plugins also run build scripts at install time. Only install sources you trust, and pin a commit (github:owner/repo#sha).
README
dsh-memory-director
MemoryDirector 插件 —— 给 DeepSeek Harness 加"长期记忆"能力。
官方 Harness 的 compaction 只做摘要压缩，没有任何"跨会话记住重要事实"的概念。 这个插件补上 AgentFrame 的 MemoryDirector：
- 每轮对话后：LLM 分析本轮，决定记住什么 / 忘记什么
- 每次请求前：检索相关记忆，注入模型上下文
- 持久化：记忆存 JSON 文件（
~/.dsh/memory.json）
为什么是差异化
| 能力 | 官方 Harness | dsh-memory-director |
|---|---|---|
| 上下文压缩 | ✓ compaction | ✓ (可配合) |
| 记住重要事实 | × 没有 | ✓ LLM 决策 |
| 跨会话记忆 | × | ✓ 持久化 |
| 记忆检索注入 | × | ✓ pre-step 注入 |
| 遗忘管理 | × | ✓ 重要度衰减 |
接入
# profile 的 cordis.patch.yml
- id: memory-director
name: '@agentframe/dsh-memory-director'
config:
provider: deepseek-official
model: deepseek-v4-flash
maxTokens: 256
dedupThreshold: 0.8
forgetThreshold: 0.1
storePath: ~/.dsh/memory.json
auto: true
机制
agent/turn-stopping (每轮结束)
→ 提取本轮文本 → LLM 决策 {remember, forget, importance}
→ remember → 存入记忆库 (去重)
→ forget → 删除过时记忆
agent/pre-step (每次请求前)
→ 取最后用户消息 → 检索相关记忆
→ 注入 system 消息 (memory-director block)
→ 模型可基于记忆回答
配置
| 字段 | 默认 | 说明 |
|---|---|---|
provider |
deepseek-official | 决策用 provider |
model |
deepseek-v4-flash | 决策用模型 (便宜快速) |
maxTokens |
256 | 决策输出上限 |
dedupThreshold |
0.8 | 记忆去重阈值 |
forgetThreshold |
0.1 | 遗忘阈值 (重要度衰减) |
storePath |
~/.dsh/memory.json | 持久化路径 |
auto |
true | 自动挂钩 agent 循环 |
License
GPL-3.0 © Cloud LTE Studio / AgentFrame
Links
More in this category
omdsh-dev/dsh-mnemon★ 21
Cross-agent, local-first persistent memory plugin for DeepSeek Harness (DSH), powered by Mnemon. It shares long-term memory across Mnemon-enabled agents and adds runtime memory, searchable project documents, semantic recall, knowledge graph, and a Sidebar UI.
LoserFox/distill★ 16
Automatic conversation distillation: background subagent reflection + skill create/update.
modusensus/dsh-mneme#dsh-mneme★ 9
Cross-session memory for DSH: SQLite + human-editable Markdown mirror, autoDream consolidation, six memory tools, and fully-offline semantic search (local embeddings, reranking, clustering).
nowledge-co/nowledge-mem-deepseek-harness★ 5
One memory layer for every AI tool and agent: Context Bundle injection, prompt-time recall, MCP tools, and turn-end DSH thread capture.
Aik358/dsh-auto-memory★ 4
Auto-memory for DSH: three-layer memory (user / project notes / daily logs) with automatic injection, per-turn auto-consolidation, AI greetings, smart search, a calendar view and a settings page, plus inheritance of other AI tools' memories.
PerryLink/dsh-memento★ 3
Bounded, layered, approval-gated, auditable cross-session memory: a typed `ctx.memory` seam with a zero-dependency SQLite provider, a `memory` tool, and frozen snapshot injection; every write passes the approval gate and stays reconstructable from the session log.