正在加载视频...
视频加载失败
Introducing the Knowledge Agent Template Vercel dropped the Chat SDK so we built a full template on top of it that you can deploy in one click. Here's what's inside: • Add your sources: GitHub repos, YouTube, etc. directly from the UI • File-based agent using grep/find/cat • Powered... show more
54,794 次观看 • 8 个月前 •via X (Twitter)
25 条评论

Demo: GitHub repo:

@nuxt_js @vercel The Nuxt + Vercel stack just keeps getting better. Can't wait to try this template out

@vercel Very nice! Can you extend this further with more tools?! This kind of feels like Vercel’s answer to openclaw haha

@vercel Yes, it is completely extensible. (and I actually have some things in stock to release on that subject 👀). And although that wasn't the goal, it can clearly serve as a basis for it!

@vercel Incredible! Yeah that was the first thing I thought of haha! So this agent can execute code via vercel sandboxes?

@vercel Yes !

I'm currently building a similar knowledge-agent chat feature inside an existing Nuxt 4 app and want to understand what this template provides beyond the base AI SDK primitives. My current stack: Chat class, streamText, smoothStream)Streaming: ai + @ai-sdk/vue UI: @nuxt/ui v4 chat components (UChatMessages, UChatPrompt, UChatPromptSubmit) RAG: Genkit embeddings → Firestore vector search Model: Google Gemini via @ai-sdk/google Hosting: Firebase App Hosting

@vercel is there something similar in nextjs?

@vercel We have a lot of templates and starters for doing similar things, but not everything is put together in the same way, at least not to my knowledge🤔

@vercel this gives me an idea of what is possible. it is like lego block. later, while seeing something/some problem, if there is a feeling of "i wish", i can build them immediately using these kinds of lego blocks. bookmarked.

@vercel That’s interesting gonna have to give it a whirl

@vercel this is a rly interesting way on how u manage webhooks and stuff but looks like it makes it more complex on code but easier to deploy?

@vercel I spent a lot of time trying to make it as easy to deploy as possible (potentially at the expense of code complexity). That's why I also added documentation and skills to really explain how to extend, customize, etc. the project

@vercel one-click deploy on top of a well-designed SDK is genuinely hard to pull off cleanly. nice work

@vercel Deploying in one click is great, but the customization is key...

@vercel it is nice that we finally have a way to run grep on a youtube video without leaving the browser.

@nicoalbanese10 @vercel whoa... any reason for using nuxt over next?

@nicoalbanese10 @vercel I'm probably biased, but DX is better, uses Vue, whose syntax I much prefer, Nuxt is much more extensible (module/layer system), and there are many other reasons!

@vercel One-click deployment will accelerate serious experimentation.

@vercel doesn't this burn more tokens compared to RAG and vector search?

@vercel That sounds like an interesting project! It's awesome that you're building on top of Vercel's Chat SDK to make it more accessible for people.

@vercel trying to understand what this is. Would it be comparable to NotebookLM?

@vercel don't forget to use our premium components😅

@vercel grep/find/cat for file-based agent? interesting choice. how's that working out performance-wise at scale? been building something for this (atomchat) so im biased but we ended up needing more indexing.

@vercel Knowledge Agent Template sounds impressive until you realize it's just a fancy wrapper around stuff we were doing manually two years ago. Cool branding though.
