I've been working on a project since January of this year, dedicating about 10 hours of time per week outside of my regular job.
Key Features:
- Visual Interface: allows users to build processes and nest them to create powerful workflows. This is a significant shift from other similar projects like AutoGen, which don't offer this level of accessibility to non-technical users. I truly believe that for AI to take off, we as a community need to appeal more to every-day workers who don't have a technical background.
- Docker Container: each user has a dedicated docker container that the process engine can (eventually) automatically write and manage code for.
- Built for Sharing: Whereas many ai workflow systems limit themselves to running on a local machine, this system was built to allow collaboration between ALL users of the system. This means that processes that are developed are instantly available to the other users by default. ( There's a tutorial on the github to host it locally if you are not so open with your processes)
Challenges and Progress: 1) This is my first venture into a full stack project using Rust and Svelte. One of the key breakthroughs was adding protobuf to allow automatic serialization/deserialization between the frontend and backend. That coupled with the automatic type generation by Rust and Typescript significantly accelerated development. 2) I've never released an open source project. The process of coming up with a description and key features of the project has helped tremendously with keeping the focused on solving end-user goals and not just getting caught up with technicalities/refactoring bad code
Looking for Contributors: I'm on the lookout for beta testers to help refine the tool. Additionally, any Rust/Svelte developers (AND graphic designers!) interested in contributing to an AI-centric project are more than welcome. Building a community around this project would be ideal because I don't want to get stuck with my own limited vision.
Long-Term Goals: The ambition is quite big - to make this tool robust enough to create generalized processes that handle a wide range of computer-based tasks. To do this, the current feature roadmap is:
- Improve loop nodes
- Improve conditional nodes
- Add process scheduling / resuming
- Add persistent memory feature for processes to write important global context as needed
- Add versioning feature to protobuf schemas for easier data-structure evolution
How You Can Help: I'm keen to hear from anyone interested in testing or contributing to the project. Your feedback could be invaluable in steering its direction. Particularly, for implementing loops/conditional/code nodes. There are rudimentary implementations but these aren't as generalize-able as I would like. Just in the process of preparing for release, I discovered tons of bugs that wouldn't have been found otherwise.
Next Steps: For a better understanding of what's been developed so far, I'd recommend checking out this demo video . The video give you a glimpse into what the project is about and its capabilities.
Contact and More Information: If you're interested in being part of this journey, please feel free to reach out at [[email protected]](mailto:[email protected]) for more details, be it as a tester, developer, or just someone curious about the project.
Youtube Demo: https://www.youtube.com/watch?v=3pwcVoCyZ7A
Working Project: https://liminalnook.com/ -- you need to provide an openai api key, though, if you want a personal demo that doesn't need an api key then hit me up on slack and I'll show you how it works using my own api key.
Github Link: https://github.com/robert-at-pretension-io/Totally-Not-Skynet
Slack Channel: https://join.slack.com/t/slack-l4e7680/shared_invite/zt-2a2ndq2v4-OnEF7g05M6J7aGAgM3bPQQ -- I haven't created a channel before so please dm me here if you're having a hard time joining.
Please leave any feedback below or ask any questions about the project! Thanks for reading :)
[link] [comments]