https://vapi.ai/ logo
Join Discord
Powered by
URGENT (production environment) Assistant did not ...
# support
t
For these 2 calls, 5b008756-6b10-4dbf-8e4e-54ce728d7143 bc48b6da-490c-42c2-a9b4-91ee68d5bbb8 Our educated guess is that the Anthropic model (Claude 3.5) did not answer or servers were overloaded We have 2 issues 1. MOST URGENT and IMPORTANT: voice received but no response was provided 2. QUESTION: how do we set fallback LLM, example user Anthropic Claude 3.5 as default and have a fallback model to be OpenAI GPT 4.1 for instance 3. Less urgents: we do not have access to the logs. When trying to retrieve them from our private S3, they cannot be found Thanks for your support, this is a production environment and we are encountering these issues frequently, ie the agent stopping responding to users
l
It looks like the Claude 3.5 model likely didn’t respond due to overload or connectivity issues. We can help set up a fallback LLM, for example GPT-4.1, to ensure the agent always replies. We should also verify your S3 log integration to make sure all requests are being captured. Can you confirm if fallback models are currently enabled in your setup?
t
Hello @Lordvickthor I can send you the 2 logs I found on our S3 but one does not exist Yes please help us setup the fallback model, I do not know how to check if a fallback model is setup How can we proceed
l
Hello! I can take a closer look at the two logs you found and help verify what went missing. I can also guide you step-by-step to set up a fallback LLM so Claude 3.5 defaults but GPT-4.1 (or another model) kicks in if needed. The easiest way is to review your current model configuration and adjust the fallback settings directly. We can continue this discussion privately so I can assist you through the setup and ensure everything works reliably.
@TheNewPenguin
t
Thanks for your quick reply @Lordvickthor Just sent you a PM
@User Can you provide support here?
@Vapi @Vapi Support Bot anyone?
v
To set up a fallback LLM, add the 
fallbackModels
property to your assistant's model configuration. For example:
Copy code
json
"model": {
  "provider": "anthropic",
  "model": "claude-3.5", 
  "fallbackModels": [
    { "provider": "openai", "model": "gpt-4.1" }
  ]
}
Replace the model names with your actual model IDs. This ensures your assistant will automatically switch to GPT-4.1 if Claude 3.5 fails to respond. [See documentation for details](https://docs.vapi.ai/quickstart/dashboard#set-the-llm-large-language-model) Source: - [Vapi Quickstart: Set the LLM (large language model)](https://docs.vapi.ai/quickstart/dashboard#set-the-llm-large-language-model)
t
This is not working I tried to add 'fallbackModels': [{'provider': 'anthropic', 'model': 'claude-3-7-sonnet-20250219'}] or 'fallbackModels': ['claude-3-7-sonnet-20250219'] In both cases I am receiving a 400 error: POST https://api.vapi.ai/assistant "HTTP/1.1 400 Bad Request"
Ok, I will share my findings with the community on undocumented behavior of fallbackModels. It does not work as bot suggested above I user 2 providers: openai and anthropic fallbackModels only work with openai and throws an error if you try to use anthropic @User I need a proper fallbackModels with other providers as we prefer the way Anthropic Claude behaves but as Anthropic is regularely too busy, it just throws an error and the agents stops being responsive... A shame... This forces us to use openai to have a fallback plan SUCCESS: curl -X POST 'https://api.vapi.ai/assistant' \ -H "Authorization: Bearer $VAPI_PRIVATE_KEY" \ -H 'Accept: application/json' \ -H 'Content-Type: application/json' \ --data-binary '{ "name": "Alex", "voice": { "voiceId": "Elliot", "provider": "vapi" }, "model": { "model": "gpt-4.1", "provider": "openai", "fallbackModels": [ "gpt-4o" ], "messages": [ { "role": "system", "content": "You are a helpful assistant." } ] }, "transcriber": { "model": "nova-2", "language": "en", "provider": "deepgram" } }'
ERROR (with syntax suggested by bot): '{ "name": "Alex", "voice": { "voiceId": "Elliot", "provider": "vapi" }, "model": { "model": "gpt-4.1", "provider": "openai", "fallbackModels": [ { "model": "gpt-4o", "provider": "openai" }, { "model": "claude-3-7-sonnet-20250219", "provider": "anthropic" } ], "messages": [ { "role": "system", "content": "You are a helpful assistant." } ] }, "transcriber": { "model": "nova-2", "language": "en", "provider": "deepgram" } }' {"message":["model.each value in fallbackModels must be one of the following values: gpt-5, etc.
## Error using anthropic primary curl -X POST 'https://api.vapi.ai/assistant' \ -H "Authorization: Bearer $VAPI_PRIVATE_KEY" \ -H 'Accept: application/json' \ -H 'Content-Type: application/json' \ --data-binary '{ "name": "Alex", "voice": { "voiceId": "Elliot", "provider": "vapi" }, "model": { "model": "claude-3-5-sonnet-20241022", "provider": "anthropic", "fallbackModels": [ "claude-3-7-sonnet-20250219" ], "messages": [ { "role": "system", "content": "You are a helpful assistant." } ] }, "transcriber": { "model": "nova-2", "language": "en", "provider": "deepgram" } }' {"message":["model.property fallbackModels should not exist"],"error":"Bad Request","statusCode":400}%
r
Hi, I’m Romaric from Mentoria Guru. We tested Vapi AI, but the French language support isn’t working well (the agent doesn’t understand our requests and transcriptions are inaccurate).
c
Please configure your transcriber to 
fr
for French instead of 
en
for English.
Copy code
"transcriber": {
"model": "nova-2",
"language": "en", // change to fr
"provider": "deepgram"
t
@kyle anything about my original issue? Do you plan to have fallback plans for Anthropic?
c
We apologize for the delay in responses lately. We are in process of making some changes to our workflow in how we handle support requests. The good news is that we are starting to get back on track of having a 48 hour response time and we would like to start cleaning up anything old. With that being said, just wanted to check in with you to see if you have made any progress or changes to your project since support has last reached out.
4 Views
PreviousNext
Powered by