Hi all, I ve been testing Trieve the new kb provider, and i did not get good results with a transient assistent. I tried to integrate pinecone, with replit, as a knowledge based provider, is this setting still correct? I get an error. "knowledgeBase": { "provider": "custom-knowledge-base", "server": { "url": "https://xxxxxhd3qqady8o.wxxx.replit.dev/query", "secret": "your-secret-key-here" } And yes i checck the docs: https://docs.vapi.ai/knowledgebase Any help welcome,

@zamzam Hello! I'm glad to hear you're exploring different knowledge base providers for your voice application. The configuration snippet you've provided seems to be structured correctly for a custom knowledge base integration with VAPI. However, if you're encountering an error, there could be a few reasons: 1. **URL or Secret Key Issue**: Ensure that the URL and secret key are correct and match the ones provided by your knowledge base service (Pinecone in this case). The placeholders like

xxxxxhd3qqady8o

and

your-secret-key-here

should be replaced with actual values. 2. **Server Availability**: Verify that the server at the provided URL is up and running and that there are no connectivity issues. 3. **API Compatibility**: Check if the API endpoint of your knowledge base is compatible with the expected request and response format of VAPI. 4. **Error Details**: Look at the specific error message you're receiving for more clues. It might indicate whether the issue is with the request format, authentication, or server-side processing. If you've already checked the documentation and the above points, and the issue persists, please provide the error message or code you're receiving so I can assist you further. If the error message contains sensitive information, please redact it before sharing. If you need to check the server status or any other specific details, please let me know what information you require, and I'll do my best to assist you.

I tested the kb, either with a 'normal' assistant and a 'transient 'assistant, did not get very good results with pinecone. So i am using a tool, and pinecone as a vector database and Replit. Each time a question is made, the query is sent to Pinecone. Whereas i get good results with the fixed assistant, i have issues with the transient assistant, either with a transient tool, or with an id tool. The conversations are desynchronized, and async is selected as false. Do you have any idea of what is going on? Another issue i noticed is that conversations recordings are wrong now, and only part of the conversation is recorded. call id: 4c07ddcc-aad4-4a78-ac1e-17f28109fb7b Thanks !!

@zamzam Sorry for the late response. Are you still experiencing this problem? If so, could you share: - The recent call ID - When exactly this happened (the timestamp) - What response you expected to get - What response you actually got instead This would really help me figure out what went wrong!