Cross-session memory vault: remember / recall / forget tools, per-turn prompt injection, and a settings-page entry browser.
Install
# from npm (prebuilt)
dsh plugin --profile web add dsh-memory-vault
# from GitHub (first run asks for allowBuilds approval — follow the hint, retry)
dsh plugin --profile web add github:flymysql/dsh-memory
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 — pnpm blocks those until you allow them, so an install can stop with ERR_PNPM_GIT_DEP_PREPARE_NOT_ALLOWED or ERR_PNPM_IGNORED_BUILDS; dsh prints the exact key to add under allowBuilds in your profile’s pnpm-workspace.yaml, and the install works on the next run. Allowing a build is a trust decision: only install sources you trust, and pin a commit (github:owner/repo#sha).
README
Maintained by @flymysql · Blog · Issues
Cross-session memory vault for DeepSeek Harness (DSH).
Every DSH session starts from zero. dsh-memory gives your agents a persistent second brain: facts, preferences, decisions, and project notes survive across sessions and are recalled automatically.
- 3 model tools —
memory_remember/memory_recall/memory_forget - Per-turn prompt injection — the most recent entries are injected into every system-prompt assembly, so the agent starts each turn already aware of stored context
- Durable storage — records live in a storage domain (
dsh_memoryunit, json backend by default; unit names are snake_case perUNIT_NAME_RE) - Browser management page — a "记忆库 / Memory" page in Settings to browse, add, and delete entries
Install
dsh plugin --profile web add dsh-memory-vault
The dsh.bundle manifest wires the dsh-memory row into the profile automatically. To install by hand instead:
npm install dsh-memory-vault
then add a row to your profile cordis.yml (or cordis.patch.yml):
- id: dsh-memory
name: dsh-memory-vault
Usage
Tell your agent to remember things, or do it yourself:
"记住:这个项目的部署目标是 Windows,测试命令是
pnpm test。"
The agent calls memory_remember with optional tags:
memory_remember(content="User prefers Windows deployment; test command is `pnpm test`", tags=["project:demo", "user"])
When context from an earlier session matters, the agent calls memory_recall(query="Windows deploy") — entries are scored by tag matches (×2) and content matches, newest first. Wrong or obsolete facts are removed with memory_forget(id="m-...") or memory_forget(tag="project:demo").
Configuration
| Key | Type | Default | Meaning |
|---|---|---|---|
injectLimit |
int | 8 | Max entries injected per system-prompt assembly |
recallLimit |
int | 10 | Max entries returned by memory_recall |
- id: dsh-memory
name: dsh-memory-vault
config:
injectLimit: 12
recallLimit: 20
Browser page
Settings → 记忆库 (Memory). Lists all entries newest-first with tags and timestamps; quick-add and delete buttons.
The page talks to the host half through the harness webServer service — same-origin JSON endpoints (GET/POST /dsh-memory/entries). The dynamic-plugin prototype used the sandbox-only harness.handle/host.call package RPC; the published package uses the public route so it works in any deployment.
Development
The plugin is one cordis package:
lib/index.js— host half: domain, tools, prompt injection, RPC handlerslib/client.js— web half: settings page
License
MIT
Links
More in this category
volcengine/OpenViking#examples/dsh-memory-plugin★ 38815
OpenViking memory and context bundle for DeepSeek Harness: pre-step auto-recall and profile injection, session capture, `viking://` URI guarding, and recall/write memory tools backed by an OpenViking server.
vectorize-io/hindsight#coding-agents★ 37670
Hindsight, agent memory that learns: long-term project memory with auto recall and retain, knowledge pages, deep reflection, and per-repo memory banks.
agentscope-ai/ReMe#dsh★ 3526
Connects DeepSeek Harness to ReMe's local-first, self-evolving personal knowledge base: automatically captures completed main-agent conversations as user-owned Markdown memory, searches conversations and source material through reme_search with BM25, optional embeddings, and wikilink expansion, and schedules daily memory consolidation.
zilliztech/memsearch#MemSearch★ 2670
Shared Markdown memory for DSH and other coding agents, with automatic capture, pre-step context injection, searchable recall, and memory-to-skill self-evolution through a review panel.
vshulcz/deja-vu#extensions/dsh★ 1074
Reads the session files thirty-three other coding agents on this machine already wrote — Claude Code, Codex, Cursor, VS Code Copilot Chat, opencode, OpenClaw, Hermes, Kimi, Cline, Zed and more — including sessions from before it was installed: six tools (deja_recall, deja_session, deja_blame, deja_fix, deja_how, deja_remember), a /deja command, and optional automatic recall added to the runtime context. Local BM25 index, no LLM, no embeddings, no network (dsh plugin --profile web add dsh-deja).
adoresever/graph-memory★ 630
Traceable, searchable cross-session memory for DeepSeek Harness — conversation knowledge as typed graph nodes (TASK/SKILL/EVENT) and typed edges.
Community comments
Comments are public GitHub Discussions. Loading them connects to GitHub and Giscus; a GitHub account is required to post.