CodeLLM Language/Documentation Requests

First of all, I just want to start off by saying THANK YOU for adding CodeLLM, it is a huge upgrade to Brave Search.

I ran across a language that wasn’t registering any searches with CodeLLM, so rather than searching for a different language that works (let’s say Python) and using a correct response to hit the feedback button, I figured a community thread would be the better option for anybody else coming across this.

So in my case the only language not registering any questions would be Splunk (SPL / SPL2), is it possible CodeLLM support can be added for this?


(Long winded appreciation/feedback below)

I’ve used Brave as my daily driver ever since it came out, and it’s been great to see the progression over these last few years. It was cool to see the stackoverflow code cards rolled out, then it was great to see the additional “code” responses that Brave AI started off doing with the search, but I’ve got to say… this latest upgrade to CodeLLM takes the cake.

I’ve played around with it for a few hours now, there was one search that I was pleasantly surprised to see CodeLLM prompted for and just wanted to throw some focus on it.

To make a long story short, it’s not uncommon for me to be working with and reading up on API documentation for tools I’m unfamiliar with. Normally, my experience using Brave would be similar to any dev using Google (example below)

So I was pleasantly surprised to find today that a simple prompt such as f5 api create ltm pool would trigger a CodeLLM response

(Context links)

And not only that, give a fantastic response. Sure the endpoint varies slightly (should be mgmt/tm/ltm/pool from link 1), but the ability to pull an example from a community page/knowledgebase giving me a detailed, near perfect response without having to even click a link is incredible. Not even ChatGPT could muster up something like that with a simple prompt like f5 api create ltm pool

Now of course this is a cherry picked example, CodeLLM doesn’t prompt for similar search f5 api delete ltm pool

But I can’t emphasize enough how much above and beyond other search engines Brave would be if CodeLLM (and Brave search) expanded their crawling of knowledgebases and community pages for different tools and technologies. In these examples here, there aren’t many results for community.f5.com or my.f5.com which contain the bulk of their examples and best info, so I assume that is why CodeLLM is unable to pull prompts consistently (and got lucky with the previous search that it pulled from a clouddocs.f5.com example which ranks pretty well in the results).

So my biggest feedback to CodeLLM would be this: In it’s current state, it handles general language questions/searches really well (anything you’d find up on Stackoverflow). The area it lacks (but shows major promise) would be incorporating responses from those lesser known, more specific communities/sites. Accomplishing this would give Brave the edge over other models like ChatGPT as they also struggle with the more specific docs/examples, however Brave’s data would be recent and up to date, not months or years out of date. Not sure of how best to start scrubbing for these types of sites other than to start a discussion/request thread like this one and get some feedback from those currently using Brave+CodeLLM.