We have set { "llmRequestDelaySeconds": null, "llmRequestNonPunctuatedDelaySeconds": null, "startSpeakingPlan": { "transcriptionEndpointingPlan": { "onNumberSeconds": 1.1 } } } But vapi still cutting messages into several parts. Call id: 5fab5d56-5730-4409-bc1b-cd1cfbbe3c12

@Darkhan I understand you're experiencing issues with VAPI cutting messages into several parts despite setting the

transcriptionEndpointingPlan

. To assist you better, I'll need to check the current configuration and logs for the call ID you've provided. Please give me a moment to look into this. In the meantime, could you please confirm if there have been any network issues or if the audio quality might be affecting the transcription? This information can help in diagnosing the problem more accurately.

@Darkhan can you tell me the exact duration in the audio recording

00:41-00:44

It was because of your LLM response it added due to which the message was cut. You can see these were the voice response which was sent. 🔵 17:39:41:387 [user LOG] Voice input: Hmm... 🔵 17:39:47:070 [user LOG] Voice input: CouldThankCould you please provide the preferred time and the number of guests for your reservation on Friday, the 4th of October?

{ "time": "2024-09-26T17:40:01.895Z", "callId": "5fab5d56-5730-4409-bc1b-cd1cfbbe3c12", "orgId": "8458d17c-e496-4cf2-a379-5335178f65e5", "log": "Transcriber output: Eleventh.", "level": "LOG" }, { "time": "2024-09-26T17:40:03.011Z", "callId": "5fab5d56-5730-4409-bc1b-cd1cfbbe3c12", "orgId": "8458d17c-e496-4cf2-a379-5335178f65e5", "log": "Transcriber output: Eleventh of October.", "level": "LOG" } { "time": "2024-09-26T17:40:03.912Z", "callId": "5fab5d56-5730-4409-bc1b-cd1cfbbe3c12", "orgId": "8458d17c-e496-4cf2-a379-5335178f65e5", "log": "Transcriber output: Eleventh of October. It's not", "level": "LOG" }, # HERE { "time": "2024-09-26T17:40:04.812Z", "callId": "5fab5d56-5730-4409-bc1b-cd1cfbbe3c12", "orgId": "8458d17c-e496-4cf2-a379-5335178f65e5", "log": "Transcriber output: Eleventh of October. It's not of", "level": "LOG" }, # MODEL DIVIDED TO TWO CHUNKS { "time": "2024-09-26T17:40:05.851Z", "callId": "5fab5d56-5730-4409-bc1b-cd1cfbbe3c12", "orgId": "8458d17c-e496-4cf2-a379-5335178f65e5", "log": "Transcriber output: October.", "level": "LOG" },

Just to confirm you are referring related to partial transcript which is sent to custom llm right or are you talking about voice is getting cut out?

transcript is CUT OUT

Hey @Darkhan, can you please ask this question to Nikhil? He will be the right guy for this question.