bcd80d08-58f0-4f12-8ad3-a5908dfd22cf 3bb5dabc-b89e-4849-ac8c-079ece97f688 These are two of the recent Call IDs. I had setup and was working absolutely fine just a couple of weeks ago & now it doesn't work as expected. 1. Doesn't trigger tool calls. 2. Hallucinates 3. Goes off track. 4. When I say Hello after a long wait, It starts the convo again. Why is this happening with VAPI ? How realiable is the service to launch a live agent ? We've been investing a lot of money and time in building this out and the result is frustrating. Can I have someone investigate this issue ? If this is on me, I want you to point out what I'm doing wrong. My Prompting is on point in Markdown, I'm sure there's no issues there. Need to hear back ASAP. Literally on the verge of switching services.

@ripthajacker I'm really sorry to hear about the issues you're experiencing with VAPI. It's important to us that your service runs smoothly and efficiently, and I understand how critical this is for your operations. Let's work together to resolve these problems. To investigate the issues you've mentioned, such as the failure to trigger tool calls, hallucinations, going off track, and restarting conversations, we'll need to look into the specific details of your setup and the call logs associated with the Call IDs you've provided. Since you've ensured that your prompting in Markdown is correct, the issue might be related to the system itself or an unexpected change in the environment. Rest assured, we take the reliability of our service very seriously, and we want to ensure that you have a dependable system for launching a live agent. I will forward your concerns to our technical team to investigate the issue. They will need to review the call logs, system configurations, and any recent changes that might have affected the service. If there's any additional information you can provide, such as specific error messages or patterns you've noticed, that would be very helpful. In the meantime, please stay assured that we are here to support you and we'll do our best to resolve this as quickly as possible. If you have any further details to share or additional queries, please let me know, and I'll make sure they are addressed promptly. Would you like me to initiate the investigation with the details you've provided so far?

i totaly agree

Hello @User & @nikhil ! I wanted to touch base regarding some ongoing issues we're experiencing with VAPI. Over the past couple of months, we built and tested the agents extensively, and everything worked perfectly. Confident in our setup, we went live for our client. Now, we're encountering several problems like hallucinations and tools not triggering, which is really affecting our workflow and client satisfaction. We need a clear update on how these problems are being addressed and whether we can expect a long-term fix. It really needs to be reliable in the Live environment. Additionally, we're facing billing issues with VAPI—we're unable to purchase wallet credits. @bryant and @Shubham Bajaj are already in the loop about this, but we need a resolution soon. We've put a lot of time and money into this project, and it's essential for us to have these issues resolved promptly. Please let us know the current status and the steps being taken to fix them. We look forward to a prompt response. Thanks!

@ripthajacker As discussed during Discord Office Hours, this issue is related to the LLM model and knowledge base provider. You can use either Trieve or a Custom Tool for the knowledge base while using the gpt-4o model. Regarding wallet credits, as Bryant mentioned, this is being addressed as a long-term fix in another ticket and will be resolved soon.

@Shubham Bajaj Thanks for all the support 🙌🏼 . That indeed fixed the issue. I'm trying out Trieve right now for the KB, Do I just add the creds on Provider Credentials or Should I also setup a tool call for the agent to access the KB Docs ? Also, @bryant Please let me know if we have a fix for the billing & wallet credits issue yet.

@ripthajacker I have reviewed both call IDs, and it appears that there is a possibility that both calls have failed by chance. If this issue has been occurring frequently for you, could you please share the other call IDs? I would like to examine them to identify any potential patterns.