dsh-memory: Local-First Long-Term Memory for DeepSeek Harness
hr98w/dsh-memory
dsh-memory brings local-first long-term memory to DeepSeek Harness, combining Claude Code-style progressive disclosure with Codex-style session consolidation, storing memories in Markdown and distilling reusable insights.
dsh-memory is a long-term memory plugin for DeepSeek Harness, inspired by Claude Code's Auto Memory and Codex's session consolidation. It stores user preferences, project constraints, and validated experiences in local Markdown files. With Global/Workspace isolation and progressive disclosure, it provides only necessary memory indexes to the agent, avoiding context bloat. It supports safe writes, session consolidation, and observability, making memory management simple and transparent.
How to Install
dsh plugin --profile web add @hr98w/dsh-memory- Category
- Data & Knowledge
- Platform
- DSH Plugin
- Author
- hr98w
- Distribution
- Plugin
dsh-memory Key Features
dsh-memory Repo Summary
What Does It Do?
dsh-memory is a DSH plugin for DeepSeek Harness that combines Claude Code's Auto Memory with Codex's Session consolidation to bring simple, transparent, and context-efficient long-term memory to the harness. It solves the problem of agents forgetting user preferences, project constraints, and validated experiences after a session ends, by storing long-term information in local Markdown files and providing only Global memory and the current Workspace's memory index on subsequent requests, avoiding stuffing all history into context. Core capabilities include local Markdown storage, Global/Workspace isolation, progressive disclosure, safe writes, session consolidation, and observability.
Core Features
- Local Markdown storage: authoritative memory lives in
$DSH_HOME/memory, readable and backup-friendly. - Global/Workspace isolation: Global holds cross-project preferences; project facts stay in the current Workspace.
- Progressive disclosure: Global memory and the current Workspace index are provided first; detailed Markdown is opened on demand only when relevant.
- Safe writes: Agent and Web UI share the same write mechanism, refusing to overwrite on concurrent modifications.
- Session consolidation: evidence is gathered from ended, stable sessions and an independent consolidator agent proposes memory updates.
- Observability: each consolidation saves a result record; optional debug logs capture stages and error chains without copying conversation content or credentials.
How to Use This Plugin?
After enabling it in DSH, first configure and activate at least one text model in the Models page, then choose the consolidation model under Settings → Memory → Settings. Next, save cross-project preferences in Global Memory and manage project facts and constraints in Workspace Memory. In Session Consolidation, pick an ended, stable session and click consolidate; after it finishes, review the results. To troubleshoot, enable Debug in the settings page and re-trigger.
This page is an independent rewrite of the plugin's official README — for authoritative documentation and the latest changes, refer to the source: hr98w/dsh-memory. The plugin is third-party code that runs on your machine once installed; inclusion does not imply endorsement — please review the source before installing.