dsh-memory
Run the following command in DeepSeek Harness:
dsh plugin install navid-kianfar/dsh-memory
Paste the following prompt into your AI chat to install this plugin:
Run dsh plugin install navid-kianfar/dsh-memory in DeepSeek Harness to install the plugin, available at https://github.com/navid-kianfar/dsh-memory , then add it to your profile's dsh.profile.bundles to activate it.
About this plugin
Every new agent session starts with amnesia. The architecture decisions you made an hour ago, the coding conventions your team agreed on, the sprint goals on the board—all of it vanishes the moment the context window fills up. dsh-memory pins each project's knowledge into a single DuckDB file inside the project directory, so you stop re-explaining the same things. It provides hybrid search blending BM25F keyword scoring with optional vector similarity, and every memory carries a category, a retention window, and a full audit trail.
Rules are the hardest guarantee in this plugin. Mandatory and forbidden rules live in the system prompt, which is reassembled before every model request, so they survive aggressive context compaction by design. Edit a rule in the built-in Memory tab and the very next prompt reflects the change—no daemon, no restart, no second process owning your data. Sessions also open with the previous summary, current sprint goals, and recent decisions already injected, so continuity is automatic rather than manual.
It is built for developers working daily in DeepSeek Harness who want project knowledge to persist and stay enforceable. The memory file is a normal path you can commit to share a team's rule set or gitignore to keep it local. No separate server, no extra process to babysit—install it and it composes itself into the harness from its own patch manifest.
Use Cases
- Stop re-explaining project decisions and coding conventions in every new session
- Share mandatory and forbidden rules across the team so the agent always complies
- Carry sprint goals, architecture notes, and recent decisions across sessions without manual context pasting
Best For
- Developers using DeepSeek Harness who want persistent, searchable project knowledge
- Engineering teams that need the agent to always follow shared coding conventions
- Individuals or small teams who want memory files to be committable, shareable, or kept local
Related Plugins
Traceable, searchable cross-session memory for AI agents that turns conversation knowledge into a typed knowledge graph and recalls relevant subgraphs instead of replaying full history, natively integrated with DeepSeek Harness.
Gives DSH AI cross-session long-term memory, to-do and skill management, plus multi-session orchestration, external AI delegation, and an infinite canvas that grows with you.
dsh-mnemon is a three-tier, pluggable, Agent-driven memory system for DeepSeek Harness, combining Runtime memory, Project Documents, and replaceable Memory Spaces with nine long-term providers.
An opinionated, zero-infrastructure file-based memory protocol: human-readable markdown files plus a bounded always-loaded index, curation discipline, and no database or embeddings.