How can the AI chat reference external sources? (Websites, documentation)

Hi all,

I watched this video on how to configure the AI chat inside We Web, and it was really useful!

However, I’m building an AI support chat for my website whereby all my knowledge documentation is housed in Notion. Therefore I’m wondering what needs to be done (or if there’s a resource / video I can follow) that shows me how to do this?

Thanks so much!

1 Like

Hey Jamie! Quick question so we can point you to the right setup:

What backend are you using behind WeWeb right now (if any)?

  • Xano

  • or just WeWeb + OpenAI directly?

​​If you’re using Xano, you can use it to set up the key pieces for a RAG support chatbot (vector search + chatbot endpoint) and then have WeWeb route the chat messages to your Xano endpoint.

Here are the official Xano docs that cover the core parts:

If you don’t have a backend yet, a simple alternative is n8n as the middle layer to build the RAG workflow (pull Notion content, retrieval, answer generation) and then route the AI chat requests through n8n via a webhook/API call. Link to n8n’s RAG starter template.

Does that help?

2 Likes

The way I’ve done it is that you have an in-between function.

So in your case, you would get the user input to be a vector/keyword search in your database, select the results, then pass them to an llm with the conversation trail and get it to provide an answer.

There may be better ways in implementing this, based on the architecture you have.

2 Likes

Thanks for the feedback :). I decided to use a REST API request to n8n and manage my files from there.

I just have another question - does the AI chat component automatically come with the ‘three dots’ whilst the user is waiting back for their response, or does this need to be configured custom on the component in We Web?

I don’t think it comes with the 3 dots…might need to be corrected there.

But, if you fork the component, its fairly easy to add it.

Fork the component, then take the files and add them to Cursor/Claude and ask it to add the functionality. Then update the files in your forked component as a new version. I’ve done this with a range of custom requirements.

1 Like