Hey, this sounds like a known limitation with structuredOutputs in Vapi AI where certain fields have a hard character cap that isn’t affected by token limits, so long-call summaries get truncated. In production, this usually happens when everything is pushed into one summary field. A common fix is splitting the output into multiple fields or moving long summaries back to artifact/analysis outputs and handling them in the webhook. I’ve helped teams resolve this by redesigning the schema and adjusting post-call processing. Quick question: are you currently using one large summary field, or multiple structured fields in your config?
@Janko