Multiple Assistant response threads being spun up ...
# support
c
I've been running into an issue over the last week or so where the assistant seems to have two response threads being spun up and adding to the response stack at the exact same time. Causing the words from 2 responses to be added in asynchronously which makes it sound incoherent. This is only happening when a KB file is attached to the assistant - I'm assuming this is where the issue lies. Here is the call ID: 82df357c-cee5-4951-9ccc-05a88ff35b80 I've also attached the mp3 file https://cdn.discordapp.com/attachments/1247786716178284594/1247786716475953186/82df357c-cee5-4951-9ccc-05a88ff35b80-1717467696024-bb3cdb18-07a7-45f3-b52f-601429e034f6-stereo.wav?ex=66614b6f&is=665ff9ef&hm=d964a847819705e3ad582aeea9a3273011b5c0bb30edcc59669c5cada842b74c&
v
Hi @Connor Davis | Outbox AI, I'm an AI assistant for Vapi. I'll analyze available sources and get back to you shortly.
----------------------------- **While I generate an answer, here are some sources that might be helpful**: * [Creating Custom Knowledge Bases for Your Voice AI Assistants]() * [Create Assistant]() * [List Assistants]() * [Update Assistant]() -----------------------------
### Answer It sounds like you're encountering an issue related to multiple responses being generated simultaneously when a Knowledge Base is attached to an assistant, which could potentially be due to concurrency or synchronization issues within the assistant's configuration or the Knowledge Base processing. Unfortunately, the information available in the provided documentation does not directly address problems about multiple response threads being initiated concurrently when a Knowledge Base is attached. However, this could be related to how the assistant's settings are configured to handle knowledge retrieval and integration with the conversation flow. To resolve this issue, it might be helpful to review the assistant's configuration regarding the Knowledge Base and the handling of async operations. If possible, you could experiment with adjusting the concurrency settings or the way Knowledge Base queries are processed. For deeper diagnostics and resolution, I recommend checking the specific configurations of your assistant and how the Knowledge Base is integrated. If the problem persists, reviewing the logs associated with the provided call ID and consulting with the technical support or development team responsible for maintaining your Vapi environment might be necessary. If you have access to the system where you can modify the assistant or its Knowledge Base settings, reviewing those areas for potential concurrency handling or response synchronization issues might be beneficial.
Got it. Tagging @User for additional help.
n
Having this same problem
v
Got it. Tagging @User for additional help.
s
We are aware about this issue and we are working on it.
c
Perfect thanks!
g
Hi @Sahil I'm having this issue too. Does this mean we can't trust the KB for now or is there a work around?
s
Right now, No. But, it will be fixed by monday.
g
Thanks @Sahil it's good to get regular updates like this. Actually I was desparate to use it for a demo I'm doing, and I found that it is quite stable with Groq/Llama 70b at a temp of 0.1 When testing, it only started this "incoherent speech" once, and when I interrupted it, the speech became normal for the rest of the session. So I am comfortable sending the demo to the client with these settings