Introduction

Under the plugin-based extension approach of DeepSeek Harness (DSH), community plugins can supplement the context capabilities of AI sessions. Spirtxiaoqi7/mindspace-dsh-session-memory is a DSH community plugin positioned as “Editable, session-isolated multi-person memory for DeepSeek Harness”. It addresses common issues in long sessions: the current user is just an object in the session, and the model needs to distinguish explicit AI requirements, character information, and general memories, rather than mixing everything into a lump.

Below is an introduction to what this plugin is, what it can do, how to install it, and the compatibility issues to note.

What is it

mindspace-dsh-session-memory is maintained by Spirtxiaoqi7, licensed under MIT, and the repository address is:

https://github.com/Spirtxiaoqi7/mindspace-dsh-session-memory

It targets the DeepSeek Harness 0.1.1 compatibility line and possesses the mindspaceSessionMemory Remote. The core positioning is to provide editable, session-isolated multi-person memory.

The key changes in version 0.5.0 are: explicit AI requirements are placed into the session-owned DSH persona section and replace Agent preset persona, rather than being appended after the original persona. When there are no explicit AI requirements, the persona is empty. Characters and general memories are still retained in a separate contextual section.

Core Features

The following lists the verified plugin capabilities.

Explicit AI Requirements Enter Persona Layer

Version 0.5.0 puts explicit AI requirements into the session-owned DSH persona section. These requirements replace Agent preset persona instead of being appended. If there are no explicit AI requirements, the persona is empty.

Separation of Characters and General Memories

Characters and general memories are kept in a separate contextual section and are not mixed with the persona layer. This allows for the separate management of “requirements for the AI” and “memories about characters.”

Model Read/Write Interfaces

The model reads and writes memory through the following two interfaces:

get_session_memory
update_session_memory

One read authorization is granted per classified mutation.

Chinese Rendering of Model Visible Fields

Model-facing fields are rendered in Chinese, including:

Character One
Individual Name
Individual Information
Character Preferences
Relationships to the current AI and background

V4 Sidecar Data Migration

V1, V2, and V3 data are losslessly migrated to the V4 sidecar format. Old text exceeding the new 300 character edit limit is retained until intentionally edited.

Memory Location

Memory is retained outside the canonical conversation JSONL at the path:

DSH_HOME/mindspace-session-memory/v1

Automatic Extraction Disabled by Default

Automatic extraction is disabled by default.

Session-scoped Context-compaction Controls Retained

Session-scoped context-compaction controls are still available and are isolated from the multi-person memory documents.

Installation and Enablement

The installation process uses PowerShell commands. First, clone the repository, then complete dependency installation, checking, and packaging locally, and finally install the locally built 0.5.0 tarball in the DeepSeek Harness directory.

The following commands complete the clone, install dependencies, run checks, and prepare the local package:

git clone https://github.com/Spirtxiaoqi7/mindspace-dsh-session-memory.git
Set-Location .\mindspace-dsh-session-memory
corepack pnpm install
corepack pnpm run check
$memoryTgz = (Get-ChildItem .\dist\mindspace-dsh-session-memory-0.5.0.tgz).FullName

Next, navigate to the DeepSeek Harness directory, add the plugin to the web profile, check the configuration, and start the web interface:

Set-Location C:\path\to\deepseek-harness
corepack pnpm dsh plugin --profile web add $memoryTgz
corepack pnpm dsh --profile web --dump-config
corepack pnpm dsh web

Before installation, check the source code and the MIT license. This plugin runs with the permissions of the current dsh process, so it is necessary to confirm the source and scope of permissions before installing.

Typical Usage

The following lists reproducible basic operations.

Install Local Tarball

Execute in the DeepSeek Harness directory:

corepack pnpm dsh plugin --profile web add $memoryTgz

This step installs the locally built plugin package to the web profile.

Check Configuration After Installation

Run:

corepack pnpm dsh --profile web --dump-config

This step is used to view the configuration status of the current web profile.

Start Web Interface

Run:

corepack pnpm dsh web

This step starts the DeepSeek Harness web interface.

Model Read and Write Memory

The model accesses memory through the following interfaces:

get_session_memory
update_session_memory

One read authorization per classified mutation.

View Memory Storage Location

Memory is located at:

DSH_HOME/mindspace-session-memory/v1

This path is independent of the canonical conversation JSONL.

Applicable Scenarios and Notes

This plugin is suitable for use in the DeepSeek Harness 0.1.1 compatibility line and scenarios that require maintaining editable, multi-person isolated memory within a session. It is suitable for DSH users who wish to manage explicit AI requirements, character information, and general memories separately.

Please note the following points:

  1. Do not install it together with the legacy embedded Mindspace Memory implementation.

  2. Automatic extraction is disabled by default.

  3. The plugin has the mindspaceSessionMemory Remote and targets the DeepSeek Harness 0.1.1 compatibility line.

  4. The repository materials reference README.zh-CN.md, but this article has not verified its content.

Conclusion

The value of mindspace-dsh-session-memory lies in splitting DSH session memory into three layers: persona, characters, and general memories, while providing V4 sidecar format migration, local installation process, and default-disabled automatic extraction controls. It is suitable for use as an editable multi-person memory plugin under the web profile in the DeepSeek Harness 0.1.1 compatibility line.

Repository address:

https://github.com/Spirtxiaoqi7/mindspace-dsh-session-memory