# Advies over een supportbot voor een technisch supportforum (Discourse AI vs Discourse Chatbot)

**URL:** https://meta.discourse.org/t/advice-on-a-support-bot-for-a-technical-support-forum-discourse-ai-vs-discourse-chatbot/286349
**Category:** General
**Tags:** ai, ai-bot
**Created:** [22 november 2023 om 17:18 UTC](https://meta.discourse.org/t/advice-on-a-support-bot-for-a-technical-support-forum-discourse-ai-vs-discourse-chatbot/286349 "2023-11-22T17:18:18Z")
**Posts on this page:** 1
**Showing post:** 18

<div class="post-metadata">

### Author: ![EricGT](https://avatars.discourse-cdn.com/v4/letter/e/f1d935/32.png) [@EricGT](https://meta.discourse.org/u/EricGT)
#### Post date: [26 november 2023 om 16:00 UTC](https://meta.discourse.org/t/advice-on-a-support-bot-for-a-technical-support-forum-discourse-ai-vs-discourse-chatbot/286349/18 "2023-11-26T16:00:28Z")

</div>

> [@37Rb](#):
>
> I wouldn’t want a user to be able to update the knowledge base anyway.

Good point. Glad you noted this, I should have stated such in my reply.

> [@37Rb](#):
>
> Only staff should be able to do that

Seems like it would segue into the [review queue](https://meta.discourse.org/t/discourse-moderation-guide/63116#the-review-queue-14) nicely. That is Discourses call.

> [@37Rb](#):
>
> it doesn’t seem like that’s possible with RAG.

No.

RAG is one of the many technologies needed to get what you seek. In the case of rating the AI reply that is different from the purpose of RAG. The human feedback for a generated completion is part of the RLHF technology.

> [@37Rb](#):
>
> That’s cool!

It is!

Most of this technology is so new and moving so fast one should be open to ideas from anyone and anywhere.

> [@37Rb](#):
>
> I wonder what it should provide as a reference if the source is in our hidden knowledge base though?

The same as if you were a human and noting it in a response to a real person. I would think the name of the manual, the date of publishing, the section and paragraph.

> [@37Rb](#):
>
> This question was poorly worded due to my naivety with how LLMs work. RAG should be sufficient for now.

Thanks for sharing that. I see many post from many users and will tell you that it is a pleasure to help you.

> [@37Rb](#):
>
> What I meant was that I don’t want to develop custom code for features that merefield or the discourse dev team have already built in their plugins.

That was funny but many will not even know why. The technology is moving so fast that I am sure many are creating code one month then realizing it has now been built-in to a technology and what they created is now redundant and may even be obsolete. If you take time to learn about [Langchain](https://www.langchain.com/), you may see users noting that they threw out just about every bit of code for their app and just started using Langchain.

Another source for learning such is the [DeepLearning.AI short courses](https://www.deeplearning.ai/short-courses/) (presently free).

> [@37Rb](#):
>
> Thanks again Eric!

You are quite welcome.

Please ask more questions, that is why I am helping you. The more questions people ask about this technology the more we can help each other reach our goals.

---

_[View the full topic](https://meta.discourse.org/t/advice-on-a-support-bot-for-a-technical-support-forum-discourse-ai-vs-discourse-chatbot/286349)._