> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://developers.deepgram.com/docs/flux/state/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://developers.deepgram.com/_mcp/server.

# Understanding the Flux State Machine

Emitted events adhere to the below state machine for managing turns:

```mermaid
stateDiagram-v2
  direction LR

  [*] --> Initial
  Initial --> TurnOngoing : StartOfTurn
  TurnOngoing --> AwaitingEnd : EagerEndOfTurn
  AwaitingEnd --> TurnOngoing : TurnResumed
  AwaitingEnd --> Initial : EndOfTurn
  TurnOngoing --> Initial : ForceEndTurn
  AwaitingEnd --> Initial : ForceEndTurn

  state Initial {
    [*] --> Ready
  }

  state TurnOngoing {
    [*] --> Speaking
  }

  state AwaitingEnd {
    [*] --> Processing
  }
```

1. `Update` messages are sent for approximately every 0.25 seconds of transcribed audio, regardless of transcript updates, unless a state change has occurred.
2. An `EagerEndOfTurn` message always contains a nonempty transcript.
3. A `TurnResumed` message can only follow a preceding `EagerEndOfTurn` message.
4. The `EndOfTurn` transcript will always match the immediately preceding `EagerEndOfTurn` transcript. If the transcript changes after an `EagerEndOfTurn`, a `TurnResumed` event will occur first.
5. The `turn_index` increments immediately following an `EndOfTurn` message.
6. When using `flux-general-multi`, all `TurnInfo` events include `languages` (detected languages sorted by word count) and `languages_hinted` (active language hints). See [Language Prompting](/docs/flux/language-prompting).
7. Every `EndOfTurn` event includes a `trigger` field — `model` (native detection), `manual` ([`ForceEndTurn`](/docs/flux/force-end-turn)), or `timeout` (`eot_timeout_ms`) — indicating what ended the turn.
8. You can end the current turn at any time by sending a [`ForceEndTurn`](/docs/flux/force-end-turn) message. Flux emits an `EndOfTurn` with `trigger: "manual"` from any in-turn state.

> **Info**
>
> **Configuring Event Behavior**: The `EagerEndOfTurn` and `TurnResumed` events are only triggered when you set the `eager_eot_threshold` parameter. The `EndOfTurn` event behavior is controlled by `eot_threshold` and `eot_timeout_ms` parameters. See the [End-of-Turn Configuration](/docs/flux/configuration) for details on tuning these thresholds for your use case.

> **Info**
>
> **Barge-in and audio quality**: Flux's `StartOfTurn` event is the recommended way to trigger barge-in — it's more reliable than an external VAD because every `StartOfTurn` is guaranteed to contain a non-empty transcript. For guidance on echo cancellation, noise suppression, and other audio preprocessing that affects turn detection, see [Audio Preprocessing & Barge-In](/guides/deep-dives/audio-preprocessing-barge-in).

## Turn Lifecycle Example

Here's how Flux processes a customer calling support saying "Hi I need to cancel my subscription please."

Notice how confidence builds up and how the `EagerEndOfTurn` event fires before the final `EndOfTurn`. With `EagerEndOfTurn`, your voice agent can begin preparing a response before the user has fully finished speaking. This allows you to send a synchronous request with early context, creating the effect of a faster, more natural reply.

```json
{
  "event": "Update",
  "turn_index": 0,
  "audio_window_start": 0.0,
  "audio_window_end": 0.2,
  "transcript": "",
  "words": [],
  "end_of_turn_confidence": 0.1
}

{
  "event": "Update",
  "turn_index": 0,
  "audio_window_start": 0.0,
  "audio_window_end": 0.5,
  "transcript": "",
  "words": [],
  "end_of_turn_confidence": 0.1
}

{
  "event": "StartOfTurn",
  "turn_index": 0,
  "audio_window_start": 0.0,
  "audio_window_end": 0.6,
  "transcript": "Hi I",
  "words": [
    {
      "word": "Hi",
      "confidence": 0.95,
      "start": 0.42,
      "end": 0.56
    },
    {
      "word": "I",
      "confidence": 0.92,
      "start": 0.56,
      "end": 0.6
    }
  ],
  "end_of_turn_confidence": 0.1
}

{
  "event": "Update",
  "turn_index": 0,
  "audio_window_start": 0.0,
  "audio_window_end": 0.8,
  "transcript": "Hi I need to",
  "words": [...],
  "end_of_turn_confidence": 0.1
}

{
  "event": "Update",
  "turn_index": 0,
  "audio_window_start": 0.0,
  "audio_window_end": 1.0,
  "transcript": "Hi I need to cancel my subscription.",
  "words": [...],
  "end_of_turn_confidence": 0.3
}

{
  "event": "EagerEndOfTurn",
  "turn_index": 0,
  "audio_window_start": 0.0,
  "audio_window_end": 1.1,
  "transcript": "Hi I need to cancel my subscription.",
  "words": [...],
  "end_of_turn_confidence": 0.3
}

{
  "event": "TurnResumed",
  "turn_index": 0,
  "audio_window_start": 0.0,
  "audio_window_end": 1.2,
  "transcript": "Hi I need to cancel my subscription please",
  "words": [...],
  "end_of_turn_confidence": 0.1
}

{
  "event": "Update",
  "turn_index": 0,
  "audio_window_start": 0.0,
  "audio_window_end": 1.4,
  "transcript": "Hi I need to cancel my subscription please.",
  "words": [...],
  "end_of_turn_confidence": 0.3
}

{
  "event": "EagerEndOfTurn",
  "turn_index": 0,
  "audio_window_start": 0.0,
  "audio_window_end": 1.5,
  "transcript": "Hi I need to cancel my subscription please.",
  "words": [...],
  "end_of_turn_confidence": 0.3
}

{
  "event": "Update",
  "turn_index": 0,
  "audio_window_start": 0.1,
  "audio_window_end": 1.6,
  "transcript": "Hi I need to cancel my subscription please.",
  "words": [...],
  "end_of_turn_confidence": 0.5
}

{
  "event": "EndOfTurn",
  "turn_index": 0,
  "audio_window_start": 0.0,
  "audio_window_end": 1.7,
  "transcript": "Hi I need to cancel my subscription please.",
  "words": [...],
  "end_of_turn_confidence": 0.7,
  "trigger": "model"
}
```

> Traditional STT+VAD requires you to build complex interruption logic. Flux handles this natively.