DeepSeek Harness Plugins · Memory
Plugins that extend DeepSeek Harness (DSH) with vision, workflows, memory, web tools and more.
Local-first A-share industry chain and concept knowledge graph MCP server with zero dependencies, no API key, structured queries for stock mapping, peers, and supply chains.
Journals every compaction summary as a timestamped file in the session workspace and provides a History subtab to browse them as Markdown.
Permanent memory plugin for DSH offering cross-session vector semantic search, LLM-driven reflection, knowledge page generation, and codebase survey; supports local Ollama or free Gemini API under the MIT License.
Local-first vector memory plugin with SQLite storage, automatic recall injection, conflict detection, and DeepSeek cloud extraction at session flush.
Automatically purge archived session records past a configurable retention period (default 7 days) to keep the Web GUI lightweight, with sweep schedule adjustable from the settings page.
A DeepSeek Harness plugin that provides persistent project-level session memory storage, retaining project context across conversations.
Remembers your last workspace directory and prefills it in the Add workspace dialog via two lightweight hooks, with a single on/off toggle in Settings and no official flow replaced.
Adds a restart-and-resume flow to DSH that records running top-level sessions, delegates the restart to dsh-restart, and safely submits one continuation turn after the new Host starts.
Persistent conversation memory for DeepSeek Harness with memory tags, AI auto-naming, multi-AI privacy isolation, experience archiving, and sidebar AI names.
Gives long sessions an infinite-context feel via multi-tier memory, progressive compression, and a mid-thinking guard.
A session governance plugin for DeepSeek Harness featuring multi-tag organization, archive lifecycle management, tag relation graph, safe deletion, and full conversation fold navigation.
Provides workspace-isolated long-term memory for DSH, enabling cross-session reuse of project facts and user preferences with search, correction, deletion, and auditable revisions, without requiring a vector database.