dsh-layered-memory
Layered long-term distillation memory plugin (L0-L3) with automatic context injection and local/remote embedding retrieval for DeepSeek Harness.
Install
$ dsh plugin --profile web add dsh-layered-memoryPlugin Overview & Capabilities
AI-assisted organization based on the public repository snapshot. The content must be grounded in source evidence and does not replace compatibility or security verification.
Key Capabilities
- Four-layer memory pipeline capturing L0 conversations, L1 atomic facts, L2 scene blocks, and L3 user persona
- Pre-step automatic context injection into agent prompts with duplicate suppression and decay weighting
- Built-in agent memory tools: memory_search, conversation_search, and memory_read_scene
- Tri-state semantic search supporting BM25 full-text search, local ONNX CPU inference, and remote OpenAI-compatible embeddings
- Interactive UI settings panel with token cost dashboard, layer-specific LLM routing, and session-level memory mode controls
Useful For
- Retaining user preferences, recurring project context, and workflows across separate chat sessions
- Separating personal/casual memory from professional coding and development tasks
- Reducing repeated prompt exploration and token costs in long-running agent workflows
- Enabling local, privacy-first semantic memory retrieval without external API dependencies
Who It Fits
- DeepSeek Harness users needing persistent cross-session memory and context continuity
- Developers building multi-turn workflows requiring structured background retrieval and tool invocation
- Teams seeking detailed token accounting and model fallback routing for memory distillation
Documented Limitations
- Requires Node.js >= 22.16.0
- Vector embeddings require downloading local ONNX models (~100-560MB) or configuring an external OpenAI-compatible embedding API
- Distillation token costs depend on conversation volume and configured distillation LLM models
DSH Compatibility
Version-specific runtime evidence collected by DSH Plugin. A missing result means we have not tested that combination yet.
Security Signals
Objective signals discovered from package metadata and source inspection. These are not a guarantee that a plugin is safe.
package.json declares 4 runtime, 14 development, 8 peer, and 0 optional dependencies.
package.json declares DSH bundle metadata.
GitHub reports the repository license as MIT.
A root package.json was captured and can be inspected by the registry.
Public GitHub source metadata is available for this registry snapshot.
Source & Registry Notes
Public provenance, Registry classification, and the latest source check for this entry, kept separate from runtime verification.
- Source repository
- JunNanLYS/dsh-layered-memory
- Registry source
- GitHub · dsh-plugin topic
- Registry classification
- Plugin
- Source checked
- 08b6109 · 2026-08-27
This project is independently indexed from public source information. DSH Plugin is not affiliated with DeepSeek or the plugin author. Always check the author repository before installation.
Repository Activity
- GitHub stars
- 10GitHub stars
- Forks
- 1
- Open issues
- 2
- Last commit
- 2026-08-27
- Last release
- 2026-08-27
Related DSH Plugins
Ranked by overlapping capabilities, use cases, plugin type, categories, and DSH profile.
Composable three-tier memory control plane for DeepSeek Harness with runtime context, searchable documents, and pluggable long-term memory.
View plugin →Comprehensive long-term memory, self-evolution, skills, task management, and multi-agent dispatch bundle for DeepSeek Harness.
View plugin →Local-first, self-evolving memory management kit and knowledge base integration for DeepSeek Harness.
View plugin →Import and resume chat histories from 18+ AI coding agents into DeepSeek Harness with reverse export and sync.
View plugin →