I want to check if the LLM is actually generating responses with flush syntax, so I’m trying to extract the raw LLM response. I tested with voice.chunkPlan and voice.chunkPlan.formatPlan.enabled = false, but I didn’t see any tags in the assistant’s output. https://docs.vapi.ai/assistants/voice-formatting-plan#customizing-the-formatting-plan I’ve enabled the required configurations for flush syntax, but the results are inconsistent across calls. https://docs.vapi.ai/assistants/flush-syntax LLM: OpenAI GPT5 Mini Voice: ElevenLabs Has anyone figured out a reliable way to confirm if the LLM is outputting flush tags in its raw response?

Hi orthogenius, To get the LLM to use flush syntax, follow these steps: 1. **Flush Configuration**: Make sure to use

<flush />

,

<flush>

, or

</flush>

in the LLM outputs where you want immediate audio feedback. 2. **Voice Configuration**: Check that

voice.chunkPlan

has

"enabled": true

and

"formatPlan": { "enabled": true }

. 3. **Extract Raw LLM Output**: Disable other formatting to see flush tags in the raw output by setting

voice.chunkPlan.formatPlan.enabled = false

. 4. **Troubleshooting**: If flush tags aren't recognized, check for syntax errors or config issues. You can find more details [here](https://docs.vapi.ai/assistants/flush-syntax). Test with different scenarios to see how flush syntax works. If issues continue, review your configurations or simulate responses to pinpoint the problem.

Thanks for the reply, Pravin After multiple trials, the LLM did output tags — I verified this by setting voice.chunkPlan.formatPlan.enabled = false. Later, I switched to voice.chunkPlan.enabled = true and formatPlan.enabled = true to see how the flush syntax works. But in the call logs during testing, I noticed "flush": false at the points where I had predefined tags in the system prompt setup.

Could you please provide a call id where that occurred?

Sure Call IDs where flush configs were enabled but returned "flush": false in call logs: a19d9b74-36a3-466a-b335-a02faa463817 9f9b5691-4415-4ba1-805f-77d54ad7efcd

Thank you for the call ids. We will look into this and get back to you