genetic-orange•17mo ago
slow response
my travel assistant is very simple, just add in openAI 4.0 turbo and google API. but why it response so slow. what is the problem? the google link given also becoming fully wrong when asking more on the chatbot.
2 Replies
The slow response time of your travel assistant could be due to several factors. One of them is the use of GPT-4.0 Turbo. Although it provides advanced reasoning and conciseness, it is known for slower results compared to other models like GPT-3.5-Turbo or Claude 1.
Another factor could be the number of tokens you're using. The more tokens you use, the longer the response time can be. The maximum number of tokens available per response is 512, and this includes your prompt and settings.
As for the incorrect Google link, it might be due to the way the assistant is handling the conversation logic. If the assistant is not properly trained or if the conversation logic is not well defined, it might not be able to provide the correct Google link.
To improve the performance of your assistant, you might want to consider using a faster model, reducing the number of tokens, and reviewing your conversation logic. Also, make sure to train your assistant whenever you make changes to the dialogue model or Model Manager.
For optimal performance, you should have at least 5-10 Utterances for each Intent. For entities, at least 10 values & variations for your Entity types, synonyms, and mentions in other utterances.
Lastly, if you're using the Web Chat integration for your assistant, make sure to check the settings in the Integrations view of your assistant. You can customize various settings there, including the chat persistence, which determines whether the assistant remembers previous conversations with the user.
For more information, you can refer to the following documents:
1. Voiceflow Developer Guide
2. Voiceflow Knowledge Base
GPT 4 is slow. We typically see response times of around 10 seconds.