Original Discord Post by qqqo_92491 | 2024-01-08 20:24:57
<@1023671043287699568> How is the data that goes into the knowledge bank trained? Is it based on the GPT3 API(Is GPT3 correct?) and referencing the TXT file as it is? Or is the data trained by convai itself with machine learning techniques?
I’m curious about how the knowledge bank is utilized. And what does the character’s backstory do, does it have anything to do with the knowledge bank?
Hi <@1138879756113293334> , at the backend we use a number of different LLMs. The knowledge bank pipeline is setup in such a way that based on the user’s query, the relevant info ( if any present in the knowledge bank) is identified and the information is forwarded for the generation of the response.