OpenWhale

OpenClaw but ONLY with 8b or less Ollama models so that you have total data ownership and on-device AI assistant capabilities

Why?

I love OpenClaw (or Moltbot... or whatever), I really do. It is genuinely useful as long as you know how to use it safely and keep an eye on what it's doing (running, git commit etc....)

However, for it to be most useful, I want to guarantee that everything is processed on my device, that I own, in my house. And to do so, I do NOT want to have to go out and buy an M4 or some fat GPU like all the slop-posters on LinkedIn raving about moltbot must have done.

So here we are. It will likely never be as well tested/loved as Moltbot, but if it is useful to myself and others, that's awesome!

WARNING

Just like Moltbot says, I cannot control what your LLM models might do. I have not yet implemented safeguards like Moltbot has for command execution. And besides, wether you the reader are a real human or OpenAI codex trying to help someone out, you should never just run something that a stranger on the internet made......

WARNING v2

Also, just because the model runs on your local machine, that still doesn't mean an attacker couldn't convince it to run a bash command to install malware, or push your files somewhere, etc. As a rule of thumb: never ask OpenWhale to operate on a document or website you haven't checked over in plain-text form. (Since prompt injection could happen in the HTML comments of a website it's loading, for example. Or in white text on a PDF.)