📚 Documentation: Your /llms.txt isn't doing anything.
💬 Let us know how we can improve our documentation.
https://docs.copilotkit.ai/llms.txt
This link shows all links, but its linking to the actual pages. Not markdown versions of the pages. So it basically isn't any differnt from /sitemap.xml. AI's cannot read pages that is why llms.txt was created. But it is only effective when the team behind the docs creates markdown versions of the pages.
[NOT AVAILABLE] https://docs.copilotkit.ai/guides/backend-actions/remote-backend-endpoint
Hi @frederikhendrix, I've used this llms.txt file with a model that can search the internet by the URL's provided, & worked as expected.
Our AG-UI protocol has the llms.txt file in Markdown and we'd be open to a contribution if this is something that interests you.
Hi @frederikhendrix, I believe what you're looking for is https://docs.copilotkit.ai/llms-full.txt, that has our entire technical documentation. llms.txt is more for navigating systems.
Hey @frederikhendrix, we are planning on adding the llms-full.txt link that Tyler shared to the docs.
Thank you @NathanTarbert and @tylerslaton but what I am mainly missing is this:
Copilotkit has a very good looking and nicely functioning Chat UI specifically made for Agents and tool calling and way more things. Then you say you can couple to different Agents, but in a production setting I would want to build those agents and make them available at my API endpoint, this way I can start storing chats, making calls to other MCP's or API's from within my own API and not from within the Frontend where Copilotkit is.
But ofcourse when it is implemented in the API and not in Copilotkit I would still want to see everything happening.
This brings me to the AG-UI protocol. I haven't been able to find documentation where it is possible to have an AG-UI protocoled server just like at: https://github.com/ag-ui-protocol/ag-ui/blob/main/typescript-sdk/integrations/server-starter-all-features/server/python/example_server/init.py
Where I can simply do:
export default function RootLayout({
children,
}: Readonly<{
children: React.ReactNode;
}>) {
return (
<html lang="en">
<body
className={`${geistSans.variable} ${geistMono.variable} antialiased`}
>
<CopilotKit runtimeUrl="http://localhost:8000/agentic_chat" threadId="main">
{children}
<CopilotPopup />
</CopilotKit>
</body>
</html>
);
}
Now I have a server running the agent and the frontend simply getting the AG-UI events back from the server.
This would really make Copilotkit and AGUI together probably the most powerfull and flexible Agent implementation on the market now.
I'm really looking forward to your answer on if this is currently possible and if the answer is yes, where can I find the documentation on implementing this? (So no specific agentic framework is chosen and it is just simply Copilotkit -> AGUI endpoint)
Kind Regards,
Frederik Hendrix
P.S. When I find out how to have my own backend with AG-UI implemented and then a Copilotkit Chat UI functioning I will ofcourse contribute to both Copilotkit and AGUI with my findings. This would also be a valuable piece in the documentation.
Hello, I am interested in working on this issue. I can investigate how to generate and expose Markdown versions of documentation so that llms.txt is really useful for LLMs, as well as review how to integrate the AG-UI protocol part into the backend with clear examples. Could you assign it to me?
The current llms.txt seems to defy the core purpose (https://llmstxt.org/). Is anyone working on making this right or is it at least on the roadmap?
Hi @weipienlee, the llm.txt file is out of date and out of sync with our current code base. We are planning to fix this and keep it up to date.
cc @brandonmcconnell
tnx @frederikhendrix, please note that the sync is not the only problem, the links should point to other llm friendly (md) document in order give it practical utility for many.