dsh-memory 🧠

Cross-session memory plugin for DeepSeek Harness (dsh).

YAML-backed remember / recall / forget / pin tools with embedding search and automatic system-prompt injection — the agent "naturally carries" memory without having to remember to call a tool.

English | 中文

✨ Features

Feature	Description
remember / recall / forget / pin	Four tools for write, semantic search, soft-delete, and force-inject pinning
Embedding search	HybridSearch (keyword 2-gram + local embedding via ollama, RRF fusion) — catches paraphrases keyword search misses
Automatic injection	systemPrompt.context() at order -50: high-importance + recent memories injected every session, visible in the WebUI as a "上下文注入" row
Dynamic topic injection	agent/pre-step listener retrieves memories matching the user's current message (with a retrieval gate that skips "好/行/ok" chatter)
Provenance labeling	Injected memories are prefixed "Retrieved memories from the memory store (not conversation history)" so the model never mistakes them for dialogue
Data safety	Serialized write queue (no lost updates), corrupt-file quarantine + backup, atomic writes, mtime cache invalidation

📦 Install

# From source (inside a dsh checkout)
pnpm install
pnpm run build

# Register in your profile patch (e.g. ~/.dsh/profiles/web/cordis.patch.yml)
- insert:
    - id: dsh-memory
      name: '@towzai/dsh-memory'
      config:
        file: /path/to/memory.yaml
        injectLimit: 8

Requires a local ollama instance with an embedding model (default qwen3-embedding:0.6b; override via DSH_MEMORY_EMBED_MODEL).

🛠️ Tools

Tool	Purpose
remember	Save a memory (content / category / tags / importance / forceInject / source)
recall	Hybrid search (keyword + embedding) top-N
forget	Soft-delete (marks retired, keeps history)
pin	Toggle forceInject — pinned memories always appear at session start

🗂️ Storage

Single YAML file (default memory.yaml). Each entry:

- id: MEM-20260814-001
  content: "..."
  category: preference | project | lesson | fact
  tags: [tag1, tag2]
  importance: high | normal | low
  created: 2026-08-14
  updated: 2026-08-14
  source: user | agent | conversation
  retired: false
  forceInject: false
  vector: [...]   # 1024-dim embedding, auto-computed on write

🔧 Architecture

src/
├── index.ts    # plugin entry: tools + injection + pre-step listener
├── storage.ts  # Storage interface + YamlStorage (queue, mtime cache, quarantine)
├── search.ts   # KeywordSearch / EmbeddingSearch / HybridSearch (RRF) / OllamaEmbedder
├── inject.ts   # selectForInjection / renderSection / buildMemorySectionSync
├── dynamic.ts  # DynamicInjector (per-session dedup + retrieval gate)
└── types.ts    # data model (+ reserved fields: vector/scope/weight for future)

Design doc: docs/2026-08-13-dsh-memory-design.md.

⚠️ Notes

• Embedding model change invalidates stored vectors (dimension mismatch is detected and logged)
• The plugin is developed against dsh v0.1.0-rc.5; peer dependency ranges may need bumping

🔗 References

This plugin is part of the DeepSeek Harness (dsh) plugin ecosystem.

• Official repo: deepseek-ai/deepseek-harness
• Plugin docs: docs/user/develop — plugin lifecycle, config, and publish guide
• Ecosystem topic: dsh-plugin

📄 License

MIT