Hello Everyone,
I have been working on something which I think was very much needed. I have made sandboxes which you can trace. Underneath it is a custom linux sandboxes running containarized docker instances that provide kernel level isolation to your agents. And this way your agents are truly isolated and on top of that my tool tracks everything said/proposed by LLM and what was actually executed on sandboxes. Please do let me know your thoughts.
well actually .ai domain was very expensive, this was economical choice rather than me jumping on hype train. you can take a look at the website once, the npm package is open source but I am providing these sandboxes out of my own pocket for people to try out, so that is closed on gh. here is the npm link tho-https://github.com/Godcreatebugs/raytrace-cli. I initially started with open source but quickly realized the ebpf traces are locked inside the apple chamber. and you need 2 months to get permission from them. So, I thought rather than building a stack for windwos, linux, mac everyone can come on cloud sandboxes and I can design system better to catch the kernel traces.
I'm not even sure what you are offering after reading this, but cloud is an even harder sell because data / traces leave the network, it will be too expensive for a new project/company to get the compliances necessary
context, I assist others with making these kind of tool and vendor decisions
still, why is markdown and a ralph loop like thing insufficient for this, why is a dedicated tool needed? (hint, I do not think it is, and this will be a question you get asked a lot, so have a good answer that shows understanding)
Hello Everyone, I have been working on something which I think was very much needed. I have made sandboxes which you can trace. Underneath it is a custom linux sandboxes running containarized docker instances that provide kernel level isolation to your agents. And this way your agents are truly isolated and on top of that my tool tracks everything said/proposed by LLM and what was actually executed on sandboxes. Please do let me know your thoughts.
Not touching any project with .si domain, I'm unwilling to feed the ego beast
Do you have a github link? This level of the stack is only considered if it is open source
well actually .ai domain was very expensive, this was economical choice rather than me jumping on hype train. you can take a look at the website once, the npm package is open source but I am providing these sandboxes out of my own pocket for people to try out, so that is closed on gh. here is the npm link tho-https://github.com/Godcreatebugs/raytrace-cli. I initially started with open source but quickly realized the ebpf traces are locked inside the apple chamber. and you need 2 months to get permission from them. So, I thought rather than building a stack for windwos, linux, mac everyone can come on cloud sandboxes and I can design system better to catch the kernel traces.
there are tons of TLDs that are cheap
I'm not even sure what you are offering after reading this, but cloud is an even harder sell because data / traces leave the network, it will be too expensive for a new project/company to get the compliances necessary
context, I assist others with making these kind of tool and vendor decisions
still, why is markdown and a ralph loop like thing insufficient for this, why is a dedicated tool needed? (hint, I do not think it is, and this will be a question you get asked a lot, so have a good answer that shows understanding)