I built Open Clank because I was using Openclaw but the idea that it's usable for the layperson is like a joke. The market for a good local AI assistant is still wide open; no offense guys but I don't think you can reasonably provide the level of personalization required to truly create unique feeling assistants; allowing people to save their most personal information locally turns your service into a verifiable "living journal" and transcends the boundaries that would normally be imposed by OpenAI having to take on the responsibilities of running the harness. Originally it was a fork of Openclaw with a bunch of bells-and-whistles (really intense RAG/semantic/vector memory/search stack, custom matrix client, etc);, but I got sick to death of that software (openclaw is.. clunky to say the least) and ported to Odysseus (because it has a nice chatgpt-like interface, but also has a sort of Hypnospace outlaw thing that gives it actual character (sorry lol)). I am almost all the way done porting a full fledged "Obsidian-like" notes suite into Odysseus, that I'm going to pair with a LMMS I'm building myself out of the components "skill-tree" (NSA, yes that one), and "Learnhouse" (popular open source LMS). This journey has really lazer focused me on what I think is important in the AI space (accessibility and discernible use-cases). Once my learning management system and everything else is stable my "final form" for Open Clank is to preload them with a whole bunch of domains of knowledge so that it's not up to the user to form useful memories, it'll come loaded with open questions "what's the users name?, what's their interests? etc"; but also a ton of "known-knows".

Currently it's in alpha state but if it wasn't for GPT Sol it wouldn't even be there! xD

Built With

Share this project:

Updates