This looks like it could be a sensible way to use LLMs in programming, although I'm not convinced AI-generated documentation can give meaningful explanations rather than paraphrase. However, since the generated wiki is editable, it seems it can be used to give a kick start to internal documentation and let the actual devs step in when it's required. I'm skittish about genAI in the workplace (or anywhere really) but this could be valuable.
However, and this might have been naïve of me, but I expected some sort of local model. And I see that you have to bring in your own vendor API keys, which implies that you let AI companies mine your codebase. Isn't that a no-go for most companies? So far I've only worked in places that banned ChatGPT over IP concerns like these. Is it already common for businesses to feed their codebases to third party LLMs?
But what if you edited a section about code piece A and then you change that code?
Does it have permission to overwrite?
Seems like a fairly hard problem to solve.
Exactly. The wiki is editable so AI gives a first draft, and developers refine it. Some companies are fine with vendor APIs, but we’ll add local model support for full privacy.
Thanks!
On our homepage (https://davia.ai/showcase) we already showcase examples like AI SDK, E2B, GPT OSS, and Transformers. We’ll also add more examples directly on our GitHub soon.
This looks like it could be a sensible way to use LLMs in programming, although I'm not convinced AI-generated documentation can give meaningful explanations rather than paraphrase. However, since the generated wiki is editable, it seems it can be used to give a kick start to internal documentation and let the actual devs step in when it's required. I'm skittish about genAI in the workplace (or anywhere really) but this could be valuable.
However, and this might have been naïve of me, but I expected some sort of local model. And I see that you have to bring in your own vendor API keys, which implies that you let AI companies mine your codebase. Isn't that a no-go for most companies? So far I've only worked in places that banned ChatGPT over IP concerns like these. Is it already common for businesses to feed their codebases to third party LLMs?
But what if you edited a section about code piece A and then you change that code? Does it have permission to overwrite? Seems like a fairly hard problem to solve.
Exactly. The wiki is editable so AI gives a first draft, and developers refine it. Some companies are fine with vendor APIs, but we’ll add local model support for full privacy.
Looks interesting and would love to try it out on my side project, but it'd be great if you could add your own package (https://github.com/davialabs/davia/tree/main/packages/agent) as a showcase.
Thanks! We’ve actually added our own package as a showcase: https://davia.ai/share/d9c79723-11a6-4a5b-a14b-04ddaa9f123f/....
It's just one short page? Or am I missing something?
There are actually 5 sub-pages. Also, the link I sent is a ‘share’ version, so it’s not editable once published.
Looks interesting but it needs some demos using real world complex projects.
Point it at famous open source projects like React, or three.js, and use those as examples on the homepage.
Thanks! On our homepage (https://davia.ai/showcase) we already showcase examples like AI SDK, E2B, GPT OSS, and Transformers. We’ll also add more examples directly on our GitHub soon.