DEDUPED Hot AI repos, turned into use cases Get the Monday 5

How to run memory search fully offline with local embeddings with Mneme

Switch the embedding provider to local so semantic recall needs no API key.

Short answer: Install the plugin: dsh plugin --profile web add @modusensus/dsh-mneme; then start DSH with dsh web; open 设置 → 记忆库设置 and change embedProvider from openai to local. All steps.

Who it is forA user who wants recall to work without sending data to an external embedding service or needing an API key.

The problemSemantic search normally depends on an external embedding provider, which conflicts with keeping memory on the machine.

The outcomeRetrieval uses a local embedding model and reranking, and keyword and BM25 search keep working when no embedding is set.

What you need first

  • DeepSeek Harness (DSH) installed, with the dsh command available
  • Node.js and npm for the MCP server and the standalone daemon
  • Access to the DSH settings panel to generate an external access token

Steps

  1. Install the plugin: dsh plugin --profile web add @modusensus/dsh-mneme
  2. Start DSH with dsh web
  3. Open 设置 → 记忆库设置 and change embedProvider from openai to local
  4. Open the status panel and check that the vector index is building before relying on semantic recall (inferred, check the README)

Prompt for your agent

In the dsh-mneme repo, set embedProvider to local in the memory library settings, save one test memory, then run a semantic search for a paraphrase of it. Expected output: the test memory returned in the results. Verify that no API key field is required for the search to succeed.

Caveat. Setting embedProvider to local keeps retrieval offline, but consolidation and summaries still use your configured host model unless you clear the dream and summarize provider settings as the README describes.

Use this in your AI agent

Ask your agent

Using the Deduped MCP, get the setup steps for slow-stack/mneme and "Run memory search fully offline with local embeddings"

Add Deduped to Claude Code once

claude mcp add --transport http deduped https://dedupedai.com/mcp

Cursor, ChatGPT, others

Share: X LinkedIn Hacker News Email Copy link

Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer