DeepSeek Harness Plugins · Memory
Plugins that extend DeepSeek Harness (DSH) with vision, workflows, memory, web tools and more.
Mirrors every session and workspace title into a local dsh-session-vault folder, making each workspace a self-contained portable archive that restores automatically on any machine.
Persistent per-project memory for DeepSeek Harness with hybrid search, binding rules injected into every model request, and cross-session continuity.
Adds a /btw slash command to DSH, letting you queue a supplementary note during an agent run without interrupting the current turn; notes are delivered FIFO on the next turn.
Give your agent a dream cycle: replay past sessions, reflect and distill lessons, write a permanent dream journal, and merge high-frequency lessons into AGENTS.md as long-term memory.
Select any text in an assistant reply to add inline annotations that are bundled with your next message, so the model responds to each one individually.
Attach local files to DSH Web conversations, injecting text and image content directly into model context without manual pasting.
A trajectory-tree context system for DeepSeek Harness that turns completed sessions into forkable checkpoints, recalls relevant prior context into new sessions, and visualizes the cross-session forest.
SQLite-indexed Obsidian vault memory for the DeepSeek Harness agent, enabling full-text and structured search with zero runtime dependencies.
A Cordis plugin that adds reliable Engram lifecycle management to DeepSeek Harness, seeding memory context into model steps and saving learnings asynchronously.
Error-driven hierarchical memory and rule system with mechanical trigger-based reminders; all auto-generated entries require user approval before taking effect.