Hey @everyone IMPORTANT update in Voiceflow that should help with a ton of your questions!
We know that token visibility has been a big question, so we've added in the ability to see how many tokens you're using at each step.
(NOTE: if you want to get updated for future updates - please add the @Server Updates role to your profile by going into 'channels & roles' and selecting it. We'll be only be tagging that in future updates so we dont spam people.)
A few core notes
- Total Tokens = Query + Answer tokens. Query tokens are what goes in to the AI, Answer tokens are what comes out. This is how OpenAI and other models calculate it.
- If you're using memory for the AI steps, you will consume more query tokens since more information is being passed in.
- The knowledge base will use anywhere from 500-2000 tokens per call. This is because its passing in relevant chunks of information from your documents to the AI to create a response.
- Different models have different costs. GPT-4 uses 25x more tokens than GPT-3.5 because its 25x more expensive. So make sure you use them intentionally.
More details here: https://learn.voiceflow.com/hc/en-us/articles/11696284619533
If you'd like a tutorial on how to optimize your token go vote on the poll in #
┃vf-new-videos and comment what you'd like to see!
We know that token visibility has been a big question, so we've added in the ability to see how many tokens you're using at each step.
(NOTE: if you want to get updated for future updates - please add the @Server Updates role to your profile by going into 'channels & roles' and selecting it. We'll be only be tagging that in future updates so we dont spam people.)
A few core notes
- Total Tokens = Query + Answer tokens. Query tokens are what goes in to the AI, Answer tokens are what comes out. This is how OpenAI and other models calculate it.
- If you're using memory for the AI steps, you will consume more query tokens since more information is being passed in.
- The knowledge base will use anywhere from 500-2000 tokens per call. This is because its passing in relevant chunks of information from your documents to the AI to create a response.
- Different models have different costs. GPT-4 uses 25x more tokens than GPT-3.5 because its 25x more expensive. So make sure you use them intentionally.
More details here: https://learn.voiceflow.com/hc/en-us/articles/11696284619533
If you'd like a tutorial on how to optimize your token go vote on the poll in #
┃vf-new-videos and comment what you'd like to see!Voiceflow
Purchasing Tokens
Additional tokens can only be purchased by pro or enterprise plans.
If you are on a paid plans follow the instructions below to purchase more tokens.
Pro & Business
Open the ...
Additional tokens can only be purchased by pro or enterprise plans.
If you are on a paid plans follow the instructions below to purchase more tokens.
Pro & Business
Open the ...
