MIT-licensed, Python-based memory plugin for DeepSeek Harness (dsh) — automatic capture, pre-step memory injection, native skill-based recall, background maintenance, and a read-only memory browser, with Markdown as the source of truth.
DSH integration
Native runtime
Author-claimed
Safety audit
Unaudited
Last verified
2026-08-25
License
MIT
01What can it help you accomplish?
Give your DeepSeek Harness sessions persistent memory that travels across agents
Daily `.md` memory journals under `.memsearch/memory/`, plus semantic recall so past conversations become searchable context in DSH and other agents
DeepSeek Harness users who want automatic conversation capture and retrieval without manual setup
Recover past decisions and debugging context directly inside DSH
Relevant memories injected before the first model step and on-demand `memory-recall` skill answers drawn from your history
Developers tracing why an architecture or decision was chosen, or resuming a prior debugging thread
02How to install into DeepSeek Harness
Installation steps
- 01
Install the memsearch CLI with the ONNX extra: `uv tool install "memsearch[onnx]"`
- 02
Add the published DSH plugin to your profile: `dsh plugin --profile web add @zilliz/memsearch-dsh`
$ dsh plugin --profile web add @zilliz/memsearch-dsh
- 03
Restart that DSH profile, or start a new session
Verify the integration
- ls .memsearch/memory/
03DSH integration and capability boundaries
Install the published `@zilliz/memsearch-dsh` plugin into your DSH profile; completed turns are captured automatically and relevant memories are injected before the first model step.
Automatic conversation capture
Completed DeepSeek Harness conversation turns→Daily `.md` memory journals appended under `.memsearch/memory/`
Pre-step memory injection
Stored memory + current session→Relevant memories injected before the first model step, only when useful
Semantic recall (memory-recall skill)
Natural-language question in DSH→Answers drawn from past conversations via the registered `memory-recall` skill
Read-only memory browser dock
MemSearch web profile→A compact dock to review skill candidates and browse `.memsearch/` files read-only
04Who is it for? When not to use it?
Good for
- DeepSeek Harness users who want automatic conversation capture and retrieval without manual setup
- Developers tracing why an architecture or decision was chosen, or resuming a prior debugging thread
05Compatibility, maintenance and safety notes
- On first launch, the default ONNX bge-m3 embedding model (~558 MB) is downloaded from HuggingFace Hub, so the initial run needs network access and disk space.
- Advanced memory maintenance (PROJECT.md / USER.md notes) and skill distillation (procedural memory) are off by default; enable them by asking your agent or editing the config.
MIT · actively maintained (latest release v0.4.19, 2026-08-23)
06Frequently asked questions
How do I install memsearch for DeepSeek Harness?
Install the memsearch CLI with `uv tool install "memsearch[onnx]"`, then add the plugin to your profile with `dsh plugin --profile web add @zilliz/memsearch-dsh`, and restart that DSH profile. Completed turns are then captured automatically.
Does memsearch need an API key to run on DSH?
No. The default embedding provider is ONNX bge-m3, which runs locally on CPU with no API key and no cost. On first launch the ~558 MB model is downloaded from HuggingFace Hub. You only need a key if you switch to OpenAI, Voyage, or a hosted Milvus backend.
How is memory recalled inside DeepSeek Harness?
Relevant memories are injected before the first model step only when they are useful, and you can also ask naturally or tell DSH to use the registered `memory-recall` skill for on-demand answers from your history.
Can memory created in DSH be used by other agents?
Yes. A conversation in one agent becomes searchable context in all others with no extra setup — memory flows across Claude Code, Codex, OpenClaw and OpenCode, all working from the same `.memsearch/` Markdown files. Milvus is just a rebuildable shadow index.
Is my memory editable and portable?
Yes. Memories are plain `.md` files under `.memsearch/memory/` — human-readable, editable and version-controllable. The web profile also adds a read-only dock to review skill candidates and browse those files without editing them.
07Related DSH workflows
openviking
by volcengine
Self-evolving Context Database for AI Agents. Unify Agent Memory, Knowledge RAG and Skills.
distilly
by titanwings
Distill how experts and agents think into reusable Skills for any bot or agent. Turn workflows and personas into shareable skill packs across Claude Code, Codex, OpenClaw and more.
nocobase
by nocobase
NocoBase is an open-source AI + no-code platform for building business systems fast. Instead of generating everything from scratch, AI works on top of production-proven infrastructure and a WYSIWYG no-code interface, so you get both speed and reliability.
archify
by tt-a1i
Agent skill for beautiful, verifiable architecture, workflow, sequence, data-flow, and lifecycle diagrams—self-contained HTML with motion and crisp export.
08Data and sources
automatic capture, pre-step memory injection, native skill-based recall, background maintenance, and a read-only memory…
dsh plugin --profile web add @zilliz/memsearch-dsh
This page is generated from the project’s public documentation, repository metadata and a structured parse of DSH Plugins; last verified on 2026-08-25. Found an error? Submit a correction.
Best DeepSeek Harness Plugins
Twelve plugins worth installing first — picked from the whole catalog, across every category.