dsh-memoryhub DSH plugin

dsh-memoryhub wires MemoryHub's checkpoint-based memory into DeepSeek Harness. It auto-loads past memory on session start, provides six mh CLI tools (mh_load, mh_save, etc.), and adds a "Memory" tab in the web UI that embeds mh ui's checkpoint map. The plugin also registers a workflow skill teaching the model when to load and save memory.

skills-workflows developer-tools

How to install the dsh-memoryhub DSH plugin

dsh plugin --profile web add github:solknight48/dsh-memoryhub

Copying does not run this command. Review the repository and version before installing the dsh-memoryhub DSH plugin.

dsh-memoryhub DSH plugin data source

dsh-memoryhub DSH plugin snapshot date: Aug 16, 2026

discovered

What the dsh-memoryhub DSH plugin can do

Auto-load memory on session start: runs `mh load` in the workspace and injects checkpoint context.
Six mh tools: mh_load, mh_save, mh_status, mh_list, mh_search, mh_checkpoint – wrap the CLI for common operations.
Session-to-memory bridge: converts dsh session events to pi-format JSONL and saves via mh save --transcript.
Web UI Memory tab: embeds a live mh ui checkpoint map next to chat and trajectory, with per-session workspace discovery.
Workflow skill: teaches the model when to load, how to save (including compact summaries), and hub rules.
Configurable auto-load, tool registration, skill, UI tab, and timeouts.

Where the dsh-memoryhub DSH plugin fits

Persist project context across multiple dsh sessions without manual re-prompting.
Save a session's key decisions and summaries as versioned checkpoints for later recall.
Search through past memory checkpoints using mh_search directly from the chat.
View a visual timeline of memory checkpoints and token budgets in the web UI.
Integrate dsh with MemoryHub's existing workflow (e.g., using mh compact saves for clean summaries).

Who the dsh-memoryhub DSH plugin is for

DeepSeek Harness users who want persistent memory across sessions.
Developers already using MemoryHub (mh) who want to unify it with dsh.
Users of dsh web UI who prefer a graphical memory timeline.

dsh-memoryhub DSH plugin limitations

Requires MemoryHub (mh) installed and on PATH (uv tool install git+...).
Only supports mh's own checkpoint format; steering messages and file attachments are not bridged in v1.
Auto-load is opt-in per token budget; default is no budget (loads all sessions).
The test suite has not been exercised against a live model session (needs API key).
The mh ui --budget option requires mh from 2026-08-14 or later (for uiBudget: none default).

dsh-memoryhub DSH plugin: from the repository README

Quoted from the solknight48/dsh-memoryhub README, the upstream source of the dsh-memoryhub DSH plugin. Copyright remains with the original authors.

English | [简体中文](README.zh-CN.md) [MemoryHub](https://github.com/solknight48/memoryhub) (`mh`) integration for [DeepSeek Harness](https://github.com/deepseek-ai/deepseek-harness) (`dsh`). MemoryHub keeps project memory as **purified sessions in git-versioned checkpoints** under `.memoryhub/`. This plugin wires that into dsh: - **Auto-load on session start** — runs `mh load` in the session workspace and injects the checkpoint memory as durable plugin context. No prompt, no tool call; the model simply starts with its memory back. - **`mh_save` bridges dsh sessions to mh** — dsh session files are not among the transcript formats mh discovers (Claude Code / pi / Codex), so the plugin renders the live session's durable event log as a pi-format JSONL transcript (in the temp dir) and saves through mh's existing `--transcript` path. Purified saves and agent-written compact saves both work, under a stable per-session identity: re-saving **replaces**, never duplicates. - **Six tools** — `mh_load`, `mh_save`, `mh_status`, `mh_list`, `mh_search`, `mh_checkpoint` wrap the CLI; the model never has to shell out for the common operations. Rarer operations (`mh link`, `mh bac

Read the full READMERepository license: MIT

dsh-memoryhub DSH plugin questions

Do I need to install MemoryHub separately?

Yes. The plugin shells out to the `mh` CLI, so you must have MemoryHub installed and on PATH. Use `uv tool install git+https://github.com/solknight48/memoryhub` (requires git ≥ 2.32 and Python ≥ 3.12). The plugin will not work without it.

How do I configure the auto-load token budget?

Set `loadBudget` in the plugin config (e.g., `loadBudget: 6000`) to limit the number of tokens auto-loaded. If omitted, the plugin uses mh's default budget (which loads all sessions). You can also pass a `budget` argument to the `mh_load` tool individually.

Will saving a session duplicate memory checkpoints?

No. The plugin identifies each dsh session by a stable ID. When you save again, it replaces the previous checkpoint for that session (via mh's one-representation-per-session rule). The bridge is rebuilt each time, so you always get a fresh snapshot.

Can I use the Memory tab without a web server?

The Memory tab is only available when the dsh web UI is running (the plugin registers a `/plugins/memoryhub/mh-ui` route). Headless compositions never see this feature. If you run dsh without a web server, the Memory tab is simply not shown.

What happens if the workspace has no .memoryhub directory?

When `noHubHint: false` (the default), the plugin stays completely silent: no injection, no log spam. The Memory tab renders an empty state with a retry button. Auto-load does nothing. You can set `noHubHint: true` to inject a one-line hint.