Hi James, What would be interesting is the concept of message history for semantic routing. For example if someone asks "What is the market outlook of EV's" and then follow up with "Can you elaborate". Since NeMo Guardrails can handle the message history, I'm interested to see if the semantic routing library of aurelio labs you're involved in, would be thinking about similar support.
Hi James, great thanks for your videos! I´ve been tinkering a bit with NeMo, and i have noticed, that it is pretty expensive. It generates a lot of prompt tokens and requests. I´m wondering, what additional cost impact it would have actually employing it. Not sure if this is a good solution for actual guard railing, but only "flow operations".
hi, i had the same thoughts, have you been able to find out more about the resources/computational costs? It would help me out a lot since I am currently writing a paper about it.
Thanks for your videos! Really nice!! I'm implementing this in Python but it seems that the synthax $name=... is generating some issues. do you have any idea about that?
Hey James! Really nice videos :) I noticed you did not give a name to the final flow example of the video. Something like: Define flow User greeting I thought the syntax required a name? So: Define flow user name greeting if exists Does it not work this way?
Thanks James, quick question, is it possible to set a fallback question in the colang file such as IF not of the above flows then trigger a RAG, appreciate your work!
Hi James, I was working on Guardrails on a demo project and already have a Chat Bot using custom data. Now it gives me all the answers but missing guardrails features. I used Langchain and OpenAI embeddings. I am not sure how to integrate langchain on that bot. User_input (contains user question) and ai_output (contains bot generated response). Could you help me here? Thank you
I'm going to cover this as soon as possible - but if you use guardrails actions you can call your chatbot via those, that is how I typically integrate the two
Hi James, What would be interesting is the concept of message history for semantic routing. For example if someone asks "What is the market outlook of EV's" and then follow up with "Can you elaborate". Since NeMo Guardrails can handle the message history, I'm interested to see if the semantic routing library of aurelio labs you're involved in, would be thinking about similar support.
Hi James, great thanks for your videos! I´ve been tinkering a bit with NeMo, and i have noticed, that it is pretty expensive. It generates a lot of prompt tokens and requests. I´m wondering, what additional cost impact it would have actually employing it. Not sure if this is a good solution for actual guard railing, but only "flow operations".
hi, i had the same thoughts, have you been able to find out more about the resources/computational costs? It would help me out a lot since I am currently writing a paper about it.
Thanks for the video. Do you have a video on langchain using guardrails
Yes, coming soon :)
Thanks for your videos! Really nice!! I'm implementing this in Python but it seems that the synthax $name=... is generating some issues. do you have any idea about that?
Hey James! Really nice videos :)
I noticed you did not give a name to the final flow example of the video.
Something like:
Define flow
User greeting
I thought the syntax required a name? So:
Define flow user name greeting if exists
Does it not work this way?
no the canonical form "greeting" will also be encoded into the semantic vector space and searched against, we can do either
Thanks James, quick question, is it possible to set a fallback question in the colang file such as IF not of the above flows then trigger a RAG, appreciate your work!
yes you can use:
define flow
user ...
How does this compare with Palm API guardrails?
Hi James, how would you integrate that with custom tools ?
Hi James,
I was working on Guardrails on a demo project and already have a Chat Bot using custom data. Now it gives me all the answers but missing guardrails features. I used Langchain and OpenAI embeddings. I am not sure how to integrate langchain on that bot. User_input (contains user question) and ai_output (contains bot generated response). Could you help me here? Thank you
I'm going to cover this as soon as possible - but if you use guardrails actions you can call your chatbot via those, that is how I typically integrate the two
I still do not understand how this is working under the hood. A bit convoluted.