dsh-plugin-memory

Persistent Knowledge Graph & Long-Term Memory Plugin for DeepSeek Harness (dsh)

dsh-plugin-memory gives DeepSeek Harness agents persistent long-term memory, episodic recall, and user preference tracking across multiple conversation sessions and workspace lifecycles.

Features

• 🧠 Long-Term Fact & Preference Storage: Remember user coding preferences, architectural decisions, and project conventions.
• 🔍 Semantic & Keyword Recall: Ranked relevance matching retrieves the right context for the prompt.
• 💻 CLI Management (dsh-memory): List, search, add, and forget memories directly from your terminal.
• 🧩 Cordis Native: Automatically exposes memory_remember and memory_recall tools to the agent.

Installation

# Add from GitHub
dsh plugin --profile web add "github:menotbobbybrown/dsh-plugin-memory"

CLI Usage

# List all remembered items
npx dsh-memory list

# Remember a new preference or fact
npx dsh-memory add "User prefers TypeScript and strict ESLint rules" --type preference --tags coding,style

# Search and recall
npx dsh-memory search "coding style"

# Delete a memory
npx dsh-memory forget <id>

Programmatic Usage (Cordis API)

import { Context } from 'cordis';
import * as MemoryPlugin from 'dsh-plugin-memory';

const ctx = new Context();
ctx.plugin(MemoryPlugin);

// Save to memory
ctx.memory.remember('Target database is PostgreSQL 16');

// Recall from memory
const results = ctx.memory.recall('database version');

License

MIT © DeepSeek Harness Community