Feature request: add native Ollama provider support for memos-local-openclaw

[a1461750564]

Summary

Please add native ollama provider support to memos-local-openclaw for both:

• embedding
• summarizer

This would make true fully-local MemOS deployments much easier.

Why this matters

memos-local-openclaw is positioned as a local memory solution, so users naturally expect to run both:

• local embedding model
• local summarizer model

In practice, Ollama is one of the most common local model runtimes for this use case.

Right now, the local plugin supports:

• embedding providers like local, openai_compatible, etc.
• summarizer providers like openai, openai_compatible, anthropic, gemini, etc.

But there is no native ollama provider.

That means users who want a real fully-local setup cannot configure Ollama directly in the official plugin.

Expected support

Embedding

Support Ollama embedding via:

• POST /api/embed

Example:

{
  "model": "bge-m3:latest",
  "input": "some text"
}

Summarizer

Support Ollama text generation via:

• POST /api/generate

Example:

{
  "model": "llama3.1:latest",
  "prompt": "summarize this text...",
  "stream": false,
  "options": {
    "temperature": 0
  }
}

Suggested config shape

Embedding

{
  "embedding": {
    "provider": "ollama",
    "endpoint": "http://127.0.0.1:11434",
    "model": "bge-m3:latest"
  }
}

Summarizer

{
  "summarizer": {
    "provider": "ollama",
    "endpoint": "http://127.0.0.1:11434",
    "model": "llama3.1:latest",
    "temperature": 0
  }
}

Viewer / Settings UI

It would also be helpful if the viewer settings page supported:

• provider = ollama
• endpoint
• model
• temperature (for summarizer)

Validation

We locally patched the plugin and verified that this works in practice with:

• embedding: bge-m3:latest
• summarizer: llama3.1:latest

The patched flow successfully completed:

• summarize
• embed
• write to MemOS SQLite
• retrieve via search / recall

So this is not just a theoretical request — it already works as a practical implementation direction.

Benefits

• true fully-local MemOS deployment
• better privacy
• less dependency on external APIs
• better alignment with the “local memory” positioning of the plugin

If maintainers are interested, I can also help provide a patch / PR direction for the Ollama provider integration.