It's AMAZING when you get a video that actually explains the topic in a clear and concise way. I found your video from a search on YT - "What is an AI Token?" Excellent job explaining the topic. Keep up the great work
One question from this video. If I follow your tutorial video to insert openAI's assistant api and use Voiceflow as UI mainly, in short I connect openAI's assistant with API. Is my price calculated by Voiceflow's or openAI's rate? I couldn't figure it out from the video. please answer Thanks for the video it's excellent.
If you’re using the Voiceflow AI features, then you’d be using Voiceflow AI tokens. If you’re connecting directly to OpenAI via the API step then you are using OpenAIs tokens directly.
Can you tell me exactly how you did this step by step with the direct API? Are there detailed instructions somewhere? I would also like to get the tokens directly from OpenAI. Thank you very much, that would help me a lot.
@@Voiceflow btw, is it possible to save the token consumption to the likes of google sheets or airtable? I'm planning to have a log that'll help me make pricing decisions.
At 8:47, you say in your gpt-4-turbo model you use 24,096 tokens, which only cost a couple of cents? Wouldn't that cost 24 cents? about 1 cent per 1000 tokens
@Voiceflow Thanks for the response! I want the AI to be able to access the knowledge base (e.g. files and URLs that I uploaded) because when I use the AI model it says: "It doesn't have access to the knowledge base"
Depending on your query, you'll find that the cost in Voiceflow is higher since we cover related costs like embeddings and hosting the Vector DB. If you're on the teams plan we do have volume discounts available that get it closer to cost when you do higher volumes. But if you prefer to use AI services directly you can also do so via. the API step in Voiceflow
@@marcozirrera They could continue making money by having the best chatgpt integration and charging more for the subscription... but, it has been a risky move.
yeah but the response speed is much faster than using the API, I think it's because their model using their own memory system which makes it really faster than using API.@@marcozirrera
Is there a way of making voice flow VOICE bots work on firefox and safari from a custom web page? WOrks well in IE/Chrome but can't record voice on FF/Safari
Its probably just ..psychological perception... but I feel like I'm burning tokens much faster when using Voiceflow Response AI step compared to my own openAI API key. ...dontknow
All the token information can be seen in the debug mode. The tracker in the app updates around every 30 seconds. Also be sure to keep an eye on what model you're using since they have different multipliers. One token in VF is not the same as one token in openAI since different models have different costs.
Is it fair to say then, that purchasing additional tokens through Voiceflow is kind of exactly the same as paying for OpenAI API tokens (in their Playground thing)...given the same model, and all else being the same? @@Voiceflow
I've been experiment with voiceflow, can VF introduce a function that link up different response AI block to different knowledgebase? let say i have text, doc, URL knowledgebase, can i just hook up differrent block with specific knowledgebase?