MIT-licensed, local Git-backed long-term-memory plugin for DeepSeek Harness: durable memory in a private Git repo, read-only memory search/context tools, recovery-point clear with two-step confirmation, backup export/import, and a DSH settings-page UI.
DSH integration
Native runtime
Author-claimed
Safety audit
Unaudited
Last verified
2026-08-27
License
MIT
01What can it help you accomplish?
Retain durable knowledge across DSH conversations
Long-term memory stored in a local Git repository (`summary.md`, `handbook/`, `rollouts/`, `archive/`), with read-only `memory.search()` / `memory.context()` retrieval for the model
Developers and researchers running DeepSeek Harness who want the model to recall preferences and knowledge across sessions without a hosted memory service
Clear learned memory safely with a recoverable Git checkpoint
A two-step confirmed clear that first preserves a Git recovery point (existing pre-clear HEAD or a dedicated checkpoint commit), then records the cleared state
Users who need recoverable day-to-day memory resets from the DSH settings page
Sync idle local sessions into memory without supervision
An optional headless synchronizer (`./integrations/dsh/dsh-memory-sync`) that processes only idle local session logs in a private per-run workspace; the host validates and writes the live Git repository
Users who want unattended memory accumulation while the model edits only an isolated copy that the host validates
02How to install into DeepSeek Harness
Prerequisites
- Official `dsh` CLI — the README compatibility table lists the DSH runtime peer range as `@deepseek-ai/dsh@^0.1.0-rc.6` (rc.6 and rc.7)
- A local Git executable available on `PATH` (README compatibility table: Git | a local executable available on `PATH`)
- Node.js 22.x (README compatibility table)
- Python 3.11.x (README compatibility table)
- macOS — the README says macOS is the supported/tested integration target
Installation steps
- 01
Run `dsh plugin --profile web add github:seriousz158/dsh-memory` (GitHub source install; the root bundle `dsh-git-memory` activates both halves in one install)
$ dsh plugin --profile web add github:seriousz158/dsh-memory
- 02
Restart the selected DSH profile after installing
- 03
Source checkout (optional): `git clone https://github.com/seriousz158/dsh-memory.git`, `cd dsh-memory`, then `npm ci --ignore-scripts`
$ git clone https://github.com/seriousz158/dsh-memory.git
- 04
Source checkout (optional): `npm install --global @deepseek-ai/dsh@0.1.0-rc.7` and confirm with `dsh --version`
$ npm install --global @deepseek-ai/dsh@0.1.0-rc.7
- 05
Source checkout (optional): run `./integrations/dsh/install.sh` to install the two local packages into your DSH profile (defaults to `~/.dsh`; keep `DSH_HOME` / `DSH_MEMORY_ROOT` consistent for non-default paths)
- 06
Restart the DSH host and leave the **长期记忆** switch on in DSH Settings to enable recall for the next model call
Verify the integration
Not specified by the author
03DSH integration and capability boundaries
Native DSH plugin bundle: install with the official `dsh plugin` CLI and load the host + settings-page UI plugins through DSH's Cordis loader interfaces (not via MCP or a third-party agent adapter)
Local Git-backed memory
knowledge and preferences from DSH sessions that should persist→a local Git repository containing `summary.md`, `handbook/`, `rollouts/`, and `archive/`, with front matter, namespaced ids, provenance, and expiry projection
Safe synchronization and recovery
idle local session logs (optional)→dry-run/preview/apply flows, operation locking, health checks, bounded batches, retry backoff, duplicate-id diagnostics, and metadata-only journals
the model can edit only an isolated per-run workspace copy; the host validates and writes the live Git repositoryRead-only memory retrieval
a bounded, explicitly untrusted `summary.md` snapshot (12 KiB maximum)→`memory.search()` and `memory.context()` results with source citations and usage-aware deterministic ordering
read usage is stored as private metadata in `.sync/usage.json`; transcripts, prompts, credentials, and memory content are not written to journalsDSH settings-page integration
the DSH settings UI→memory toggle, repository status, recent-sync state, preview actions, rollback, and a deliberate two-step clear confirmation
the UI does not expose the filesystem root or execute Git directlyRecovery-point clear
a two-step confirmed clear request from the DSH settings UI→a preserved Git recovery point (existing pre-clear HEAD for clean repos, a checkpoint commit for dirty paths) with the clear commit created on that recovery point; `.git`, `README.md`, helper scripts, directory structure, and `.last-sync` stay intact
designed for recoverable day-to-day resets, not guaranteed privacy erasure
04Who is it for? When not to use it?
Good for
- Developers and researchers running DeepSeek Harness who want the model to recall preferences and knowledge across sessions without a hosted memory service
- Users who need recoverable day-to-day memory resets from the DSH settings page
- Users who want unattended memory accumulation while the model edits only an isolated copy that the host validates
Not for
- The plugin keeps the DSH `0.1.0-rc.6` peer-compatibility range (rc.6 and rc.7) and is tested against a pinned `0.1.0-rc.7` graph; DSH `rc.8` and later releases are unverified until they pass the repository's test suite.
- Clearing memory is designed for recoverable day-to-day resets, not guaranteed privacy erasure. For privacy-sensitive deletion, remove relevant local backups and follow your organization's retention policy; Git history alone is not a secure-erasure mechanism.
05Compatibility, maintenance and safety notes
- The plugin keeps the DSH `0.1.0-rc.6` peer-compatibility range (rc.6 and rc.7) and is tested against a pinned `0.1.0-rc.7` graph; DSH `rc.8` and later releases are unverified until they pass the repository's test suite.
- macOS is the supported/tested integration target. The compatibility table additionally lists Node.js 22.x, Python 3.11.x, and a local Git executable available on `PATH`.
- Clearing memory is designed for recoverable day-to-day resets, not guaranteed privacy erasure. For privacy-sensitive deletion, remove relevant local backups and follow your organization's retention policy; Git history alone is not a secure-erasure mechanism.
- The optional synchronizer defaults to `workspace-write`; wider privileges are never a repository default. It needs a user-installed `dsh` executable (or an explicitly selected `DSH_BIN`) rather than `npx --yes`, and its session filter redacts common credential shapes and home-directory prefixes before a transcript reaches the memory-extraction model.
- The project is distributed through GitHub source installs and GitHub Releases and is not published to npm. The two workspace implementation packages remain private; only the root `dsh-git-memory` bundle is publishable.
MIT · actively maintained (latest release v0.8.2, 2026-08-25)
06Frequently asked questions
How do I install dsh-memory for DeepSeek Harness?
Run `dsh plugin --profile web add github:seriousz158/dsh-memory`. The repository root is a public DSH bundle named `dsh-git-memory` containing the host plugin, the settings-page client bundle, and its `dsh.bundle` patch, so one install activates both halves. Restart the selected DSH profile after installing.
Does dsh-memory store my memory on a hosted service?
No. Memory stays in a local Git repository — the README states there is no hosted memory service or cloud vector database. The bundle also does not include any memory data, session logs, credentials, or the local `.dsh` directory.
Which DSH versions are compatible?
v0.8.2 keeps the DSH `0.1.0-rc.6` peer-compatibility range and has been tested and locally integrated with a consistently pinned `0.1.0-rc.7` graph. DSH `rc.8` and later releases are unverified until they pass the repository's test suite.
How does clearing memory work?
The settings UI requires two acknowledgements — click **删除记忆**, then enter exactly `删除记忆`. Before changing any target memory path it preserves a recovery point (existing pre-clear HEAD or a dedicated checkpoint commit). It is designed for recoverable day-to-day resets, not guaranteed privacy erasure.
What platforms and prerequisites are supported?
macOS is the supported/tested integration target. The compatibility table also lists Node.js 22.x, Python 3.11.x, and a local Git executable available on `PATH`.
07Related DSH workflows
weknora
by tencent
Open-source LLM knowledge platform: turn raw documents into a queryable RAG, an autonomous reasoning agent, and a self-maintaining Wiki.
mirage
by strukto-ai
The World's First Unified Virtual Filesystem For AI Agents
reme
by agentscope-ai
ReMe: Memory Management Kit for Agents - Remember Me, Refine Me.
memsearch
by zilliztech
A persistent, unified memory layer for all your AI agents (e.g. Claude Code, Codex, DSH), backed by Markdown and Milvus.
08Data and sources
The repository root is also a public DSH bundle named `dsh-git-memory`.
dsh plugin --profile web add github:seriousz158/dsh-memory
The package uses DSH's Cordis loader interfaces.
This page is generated from the project’s public documentation, repository metadata and a structured parse of DSH Plugins; last verified on 2026-08-27. Found an error? Submit a correction.
Best DeepSeek Harness Plugins
Twelve plugins worth installing first — picked from the whole catalog, across every category.