Call ID: 019c27ea-1252-7775-9dce-b89b74927e01 Transcript shows only messages voiced 1min 40s into the conversation. This affects all of our calls starting from January 1st and is an ongoing issue.

Hi, I can help investigate and fix this issue so your full call transcripts show correctly. I’ve handled similar ongoing transcription problems before and can start right away. @Argonautas

Thanks for flagging this, losing the first ~1:40 of transcripts is a serious issue, especially if it’s affecting all calls since Jan 1. This usually points to a transcript stream initialization or provider-side buffering delay rather than the call itself. I’ve seen this happen when the transcription service only starts after the assistant fully stabilizes. We should check whether the transcript is tied to call.start vs assistant.speech.start events. Can you confirm which transcription provider you’re using and if this started after any VAPI config or provider change? @Argonautas

Hey — we reviewed the logs for call ID 019c27ea-1252-7775-9dce-b89b74927e01 and confirmed the transcript is missing the first ~100 seconds of the conversation. The cause is your squad’s handoff configuration using

contextEngineeringPlan: { type: 'lastNMessages', maxMessages: 8 }

. While this limits LLM context during handoffs (as expected), it also unintentionally truncates the final call record. Recommended fix: Add

fullMessageHistoryEnabled: true

to at least one assistant in your squad (ideally the final assistant), for example:

{
  "name": "Paulius_Booking_v3",
  "artifactPlan": {
    "fullMessageHistoryEnabled": true
  }
}

This preserves the full conversation history in call records while still keeping handoff context limited. Alternative: You can switch handoff destinations to

contextEngineeringPlan: { type: 'all' }

, but note this will increase token usage.