Pete, I'm interested in the pattern of using the AI to prepare the query prompt for the knowledge base retrieval and validating the answer. Is this something something that is more generally useful or used or is this just to stress-test these models for the video? Could you share your VF? Cheers!
Pete uses this strategy to eliminate hallucinations, generating an answer and then checking that answer against a knowledge base. It catches errors better than just generating the answer once (plus it's a great stress tester). Here's Pete's template: creator.voiceflow.com/dashboard?import=665bc7cec0cb31d223ec2b0b
Pete, I'm interested in the pattern of using the AI to prepare the query prompt for the knowledge base retrieval and validating the answer. Is this something something that is more generally useful or used or is this just to stress-test these models for the video? Could you share your VF? Cheers!
Pete uses this strategy to eliminate hallucinations, generating an answer and then checking that answer against a knowledge base. It catches errors better than just generating the answer once (plus it's a great stress tester). Here's Pete's template: creator.voiceflow.com/dashboard?import=665bc7cec0cb31d223ec2b0b
Yep, this is something I do. Usually because I want to give the entire context of the conversation to the knowledge base retrieval.
If you could please share this voiceflow template then would really help :)
You can add it to your workspace here! creator.voiceflow.com/dashboard?import=665bc7cec0cb31d223ec2b0b