Hi, what happens when

pipeline-error-anthropic-llm-failed

occurs? I think we had case when it happend and voicebot went silent in the middle of conversation. In such case, is there possibility to make the voicebot tell verbally an error occured before conversation ends?

Hi there! 👋 Based on what I found, the

pipeline-error-anthropic-llm-failed

error typically occurs due to issues with how tool results are structured when using Anthropic models like Claude 3.5. The error happens when there's a mismatch in the message sequence - specifically when a

tool_use

block isn't properly followed by a corresponding

tool_result

block [(source)](https://discord.com/channels/1211482211119796234/1211483291191083018/1339232387279622154). Your observation about the voicebot going silent makes sense - when this error occurs, it likely disrupts the normal conversation flow. To verbally notify users before ending the conversation: You can use the

say

method with the

endCallAfterSpoken

parameter to deliver an error message before terminating the call [(source)](https://discord.com/channels/1211482211119796234/1211483291191083018/1335788708063553549). This would allow you to create a more graceful exit experience. For better user experience when handling this error: 1. Implement fallback mechanisms that specify alternative voices or models if the primary one fails [(source)](https://docs.vapi.ai/voice-fallback-plan) 2. Ensure your tool results are processed in the correct sequence and format 3. Use alphanumeric properties for tool schemas to avoid errors with property keys 4. Set up error detection conditions that trigger appropriate verbal responses This approach should help maintain a more natural conversation flow even when errors occur, rather than leaving users hanging with unexpected silence. Sources: - [Discord Message about error causes](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1339232387279622154) - [Discord Message about call termination](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1335788708063553549) - [Voice Fallback Plan Documentation](https://docs.vapi.ai/voice-fallback-plan)

@User would be great if you could take a look into that! thanks

Hey! To help track down this issue, could you share: - The call ID This would really help us figure out what went wrong!

Hey @Shubham Bajaj I think it was: 0d627631-0440-4d4d-ae2a-16003df0b755. Is there any way to make the but say something when such errors occur? If not, do you plan to introduce such feature?

When the

pipeline-error-anthropic-llm-failed

occurs, it typically means there's a problem with the AI model processing the request. Unfortunately, this can lead to the voicebot going silent. You can improve user experience by implementing fallback mechanisms in your assistant by adding idle messages, such as asking the user to repeat their query or offering a pre-defined message to inform them of the issue.

@Shubham Bajaj my main concern is how to make the bot say something on such errors (it can be predefined message), I guess this is where it is described, right? https://docs.vapi.ai/assistants/assistant-hooks

Yup.