Voiceflow/Anthropic Reliability Concerns
We’re launching a chatbot next week for a major international brand using Voiceflow. All of our agents within the bot run on Claude 4, but last week we noticed a couple outages / issues within Voiceflow. At times the fallback LLM kicked in constantly, and today the AI-generated buttons and exit paths didn’t work at all for a certain time period. Both are critical for the usage of the bot.
Wanted to know, how confident are you guys in Claude 4’s reliability within Voiceflow going forward? And have you encountered similar outages in the past?
4 Replies
Hey @Jasper , our core API uptime is very very strong at 99.9% uptime, it's the underlying services that can have reliability which is why we made model fallback a PRO feature so more people could access it.
Our status page: https://status.voiceflow.com/
In terms of Anthropic, they can have outages from time to time so I recommend having a good fallback, but they do have the best models right now in my opinion
Congrats on the customer! 🙂
@Braden (Voiceflow CEO) , The thing here is we fully reply on the ai-generated buttons here. We couldn't get it to work to achieve the same ''button generation'' behavior with one prompt for both Claude and GPT as fallback. As far as i know there is no option for another routing when the fallback kicks-in, or I might be missing something?
Anyway, you guys are doing amazing work! Thanks
Anyway, you guys are doing amazing work! Thanks
Np! yeah sonnet is best for buttons in agents, i'd say they are pretty stable but in the odd chance they go down you may lose the buttons, but at least the agent won't go down