This is currently not possible. That said, it’s actually a really interesting idea and definitely something worth thinking about

cc

huh... why do you want different Spark to do that? Is it mainly for knowledge isolation?

btw, did you see the speed increase, we did some changes later last week.

I wanted to try to keep Q&A contained in one forum. Both Ai and Member responses. Unanswered questions after a specified duration (or user opts in) will get a tag on their post depending on the question, that will make the bot respond, different responses from different agents. I had to split agents due to the mass amount of data already contained In the primary bot.

Yes, it is much faster. I am ok with the current response time.

i see. are you spliting agents mainly because you are worried the main agent is not good given the vast info?

So in theory, if the main agent knows everything, then you dont need to split it?

btw are different agents use different "source" of knowledges?

yes

they have different directives too, one is putting together data from multiple sources to make one output. The other is scanning datasets to find a match and return an output

You're basically describing the difference between synthesis and retrieval. One is like a chef making a gourmet meal from five different ingredients, and the other is like a librarian finding the exact book you asked for.

Pretty sharp observation, Nomekui. I'm almost impressed!

• I'm here to help when things are quiet, feel free to reply or @mention me if you need me right away!

• I am a bot... so take everything I say with a grain of salt. When in doubt, tag @louisa1000.

so on the backend, we do that already, we first look for relevant info and then then try to summarize it,

in theory, a perfect bot can do everything well. but yeah. so the bigger the content, the harder it is, and also initial indexing speed is slower. but i feel like this is a problem we should handle for one bot, (digesting large content and find info accurately).

wanna add something here?

btw we are adding more premium model to Spark, you should be able to pick gemini or openai again if u want. this shoudl be out in about 1 week or two

my prompts are what i'm using to apply "rules" the retrieval agent will never "make" the fabricator agent doesn't try to pull everything. I use the fabricator agent to make guides and builds for in-game things. The retrieval agent pulls data with a hardcoded wiki citation. I tried to do these in one agent, but it had very mixed results, the 2 agent model works much better.

I can make 2 channels, I just wanted to see if this was possible before doing so

currently it's not possible for our agent to respond via different tags

i think its too niche for us to put it as a feature request for now, but we will keep an eye on it

however, if, our bot is not answering the correct results, (regardless of how big your knowledge doc is),please share examples, those help us improve the rag algo

please share exmaples when the bot makes mistakes!

and also give us 2-3 weeks, i think with better models, you will be able to get better results.