dsh-memory-forget
Run the following command in DeepSeek Harness:
dsh plugin install XIAOke8698/dsh-memory-forget
Paste the following prompt into your AI chat to install this plugin:
Run dsh plugin install XIAOke8698/dsh-memory-forget in DeepSeek Harness to install this plugin; the full source is available at https://github.com/XIAOke8698/dsh-memory-forget
About this plugin
Most memory plugins teach agents to remember everything, yet stale facts and poisoned entries quietly consume the context window, degrading reasoning quality and even serving as persistent attack vectors. dsh-memory-forget turns forgetting from an accident into a design decision: every memory carries its own TTL and decay curve, and a sliding-renewal mechanism extends the life of actively used memories while physically deleting those that idle past their full shelf life.
Key capabilities include a hard injection budget (default 2000 tokens; over budget means fewer entries, never truncation), audit logs that retain only SHA-256 hashes, a recoverable restore mode, and an Ebbinghaus-inspired active → stale → forgotten lifecycle. The entire engine is zero-dependency, zero-embedding, and requires zero LLM calls on write.
Built for developers who need cross-session memory in multi-turn agents but want the context window to stay clean and controllable. The same core engine ships as a DSH plugin, a Claude Code skill, or a standalone Node module, so you can pick whichever fits your stack.
Use Cases
- Automatically expire stale facts in multi-turn conversations to keep context clean
- Auto-unplug sub-agent temporary memories when they terminate, preventing inheritance
- Enforce per-step token budgets for injection, preventing context window overflow
Best For
- Multi-turn agent developers who need cross-session memory with strict context control
- Production teams focused on memory security, auditability, and compliance
- Independent developers building DSH plugins or Claude Code skills
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.