Good morning, I'm trying to understand which element of my assistant is causing such high latencies in the responses to the interlocutor but, when I use the API call: get calls I only have the duration values ​​for the Vapi and Transcriber components. the part of json that refers to the llm model and the voice does not contain this information. How can I get this information to study what to intervene on? Thanks Mattia https://cdn.discordapp.com/attachments/1329829621679784067/1329829621918597212/ex_duration.PNG?ex=678bc3d8&is=678a7258&hm=e10ff8c6677b2ae25744e6f63fbace9eaaada29fe3967632c0634bc626110bb7&

@Mattia Good morning, Mattia! To investigate the latency issues you're experiencing with your assistant, it's important to consider several factors that could contribute to the delay. Since you've mentioned that the JSON from the

get calls

API call doesn't contain duration values for the LLM model and the voice component, you might need to enable more detailed logging or request this feature if it's not currently available. In the meantime, you can analyze the latencies based on the configurable elements you have control over, such as

assistant.silenceTimeoutSeconds

,

assistant.responseDelaySeconds

, and

assistant.llmRequestDelaySeconds

. Adjusting these settings might help reduce the overall latency. If you need more detailed information on the latencies for the LLM model and the voice component, you may need to reach out to the support team for further assistance. They might provide you with additional tools or methods to measure and optimize these specific components. Please let me know if you need assistance with adjusting the current settings or if there's anything else I can help you with!

@User Hello, can you tell me why the "minutes" field is missing for the model and the voice? Is there a way to get this information?

Hey @Mattia To help track down this issue, could you share: - The 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!

Good morning, the assistant ID used is: 0ad5cc69-da53-4480-bcbc-806b1ab95153 As you can see from the attached json example, the latency times of the "model" and "voice" components are not present as well as the cost. Why? I'm trying to improve the latency of the first assistant response which is terrible at the moment. It takes 5-10 seconds to respond to the customer (see json) and I would like to understand which components are creating the problem. https://cdn.discordapp.com/attachments/1329829621679784067/1330829310487171113/Test.json?ex=678f66e1&is=678e1561&hm=bf9a9e6bcb3b825e004772838b31e468d8e34da8c0c7de4a5495c83114bed81b&

@Mattia You are experiencing user input being recognized later because of endpointing. I suggest trying out the call with Deepgram's endpointing set to 300ms and startSpeakingPlan waitSeconds to 0.8 seconds. The real-time model is still in beta, so you can expect minor hiccups. DO let me know how it goes.

thank you for the suggestions. The problem that the cost and duration values ​​of the model and the voice are not valued in the json remains pending. Can you help me with this? Mattia

@Mattia it is available under costs which you can see in the shared json file, correct me if I am wrong here.

As you can see from the photo extracted from the attached json in the cost section neither the cost is valued nor is the "minutes" attribute present for model and voice https://cdn.discordapp.com/attachments/1329829621679784067/1333341860403347530/Cost_Example.PNG?ex=67988adf&is=6797395f&hm=e5f7044341ff2cdd33cb9758bee899f9fe244bc124e9692432a0aad447536d38&