Memory
dsh-wsl-vecmem
173787247/dsh-wsl-vecmem · ★ 1
DeepSeek Harness WSL plugin: tiny local vector memory via Ollama embeddings + JSON store.
安装
dsh plugin --profile web add 'github:173787247/dsh-wsl-vecmem#599c35ef12464c8d1ba30d3b6bf374dd65a46dfd'GitHub 源码安装:首次需按提示配置 allowBuilds 构建授权后重试
For install and environment setup, plus plugin development tutorials, see DSH community docs ↗
README
Table of contents
Languages: 中文(首页) · English (this file)
Tiny local vector memory for dsh: Ollama embeddings + JSON store under ~/.dsh/vecmem/.
Needs an embedding model (default nomic-embed-text). Complements Obsidian — not a vault replacement. Can share the same Ollama host as dsh-wsl-ollama.
Quick start
ollama pull nomic-embed-text
dsh plugin --profile web add github:173787247/dsh-wsl-vecmem
Tools
| Tool | Role |
|---|---|
vecmem_status |
Store path, embed model, entry count |
vecmem_add |
Embed + store one text item |
vecmem_search |
Semantic nearest-neighbor search |
Config
config:
enabled: true
storePath: "" # empty = ~/.dsh/vecmem/store.json
ollamaBase: "http://127.0.0.1:11434"
embedModel: "nomic-embed-text"
timeoutMs: 60000
maxItems: 5000
If Ollama runs on Windows, set ollamaBase to an address reachable from WSL (see the ollama plugin notes).
Compatibility
| Field | Value |
|---|---|
| Plugin | dsh-wsl-vecmem 0.2.0 |
| Minimum dsh | ≥ 0.1.2 (web UI one-shot ?token= on Windows relay :3081) |
| Latest verified | See dsh-wsl-kit Compatibility (currently 0.2.1-rc.5) — single source of truth for the suite |
| Kit set | optional (not in install.sh / KIT_SET=daily by default) |
License
MIT