AI Agent Hub
Back to plugins
🧠

dsh-session-reference

Memory Updated 2026.08.16

Run the following command in DeepSeek Harness:

dsh plugin install PUPPET17/dsh-session-reference

Paste the following prompt into your AI chat to install this plugin:

Run dsh plugin install PUPPET17/dsh-session-reference in your terminal to install; the source is available at https://github.com/PUPPET17/dsh-session-reference

About this plugin

Juggling multiple sessions in DeepSeek Harvest means constantly context-switching back to older conversations to pull in a snippet you need. dsh-session-reference plugs that gap as a standalone bundle: type @ in the current surface and search across your Home session list to reference content without copy-paste.

The plugin wires up a complete reference pipeline: an @ menu search over the active session list, same-working-directory-first candidate ranking, a responsive title-capsule inserted via ReferenceInsert with its own codec, and canonical dsh-session: mention serialization that stays consistent across copy, persistence, and submission. At send time it reads the current surface object, applies a bounded projection, and durably stores untrusted recall context so downstream steps can audit the reference within size limits.

Best suited to developers who manage several project sessions in DSH and need to cross-reference content frequently. It does not alter core Harness behavior; it mounts on the generic extension points the repository exposes and activates only when you install it. Defaults cap references at three per message and 65,536 UTF-8 bytes per referenced source, both adjustable via profile configuration.

Use Cases

  • When referencing content across multiple project sessions frequently
  • When bringing key context from prior sessions into the current conversation
  • When quickly locating relevant code across sessions in the same working directory

Best For

  • Developers managing multiple parallel sessions in DSH
  • Technical writers who frequently reference context across sessions
  • Teams collaborating on multi-project work with DeepSeek Harness