gproxy-protocol 4.1.14

//! Native Realtime wire DTOs. Source definitions are cited on each object.
//! Verified with Realtime.md and openai-python snapshot
//! e12b81d3bbf644ec7045e152d69bc4b68d69cd48 (2026-09-13).
//! https://github.com/openai/openai-python/tree/e12b81d3bbf644ec7045e152d69bc4b68d69cd48/src/openai/types/realtime
use super::RealtimeAudioFormats;
use super::wire::present_nullable;
use crate::Rest;
use serde::{Deserialize, Serialize};

/// Source: `openai/types/audio/transcription_language.py`, `TranscriptionLanguage`.
#[derive(Debug, Clone, PartialEq, Serialize, Deserialize, gproxy_protocol_macros::WireBuilder)]
#[cfg_attr(not(feature = "exhaustive"), non_exhaustive)]
#[derive(gproxy_protocol_macros::DeclaredFields)]
#[cfg_attr(feature = "schema", derive(schemars::JsonSchema))]
pub struct TranscriptionLanguage {
    pub code: String,
    #[serde(default, flatten, skip_serializing_if = "serde_json::Map::is_empty")]
    pub rest: Rest,
}

/// Source: `openai/types/realtime/log_prob_properties.py`, `LogProbProperties`.
#[derive(Debug, Clone, PartialEq, Serialize, Deserialize, gproxy_protocol_macros::WireBuilder)]
#[cfg_attr(not(feature = "exhaustive"), non_exhaustive)]
#[derive(gproxy_protocol_macros::DeclaredFields)]
#[cfg_attr(feature = "schema", derive(schemars::JsonSchema))]
pub struct LogProbProperties {
    pub token: String,
    pub bytes: Vec<i64>,
    pub logprob: serde_json::Number,
    #[serde(default, flatten, skip_serializing_if = "serde_json::Map::is_empty")]
    pub rest: Rest,
}

/// Source: `openai/types/realtime/realtime_transcription_session_create_request.py`, `RealtimeTranscriptionSessionCreateRequest`.
#[derive(Debug, Clone, PartialEq, Serialize, Deserialize, gproxy_protocol_macros::WireBuilder)]
#[cfg_attr(not(feature = "exhaustive"), non_exhaustive)]
#[derive(gproxy_protocol_macros::DeclaredFields)]
#[cfg_attr(feature = "schema", derive(schemars::JsonSchema))]
pub struct RealtimeTranscriptionSessionCreateRequest {
    #[serde(rename = "type")]
    pub type_: RealtimeTranscriptionSessionCreateRequestType,
    #[serde(
        default,
        deserialize_with = "present_nullable",
        skip_serializing_if = "Option::is_none"
    )]
    pub audio: Option<Option<RealtimeTranscriptionSessionAudio>>,
    #[serde(
        default,
        deserialize_with = "present_nullable",
        skip_serializing_if = "Option::is_none"
    )]
    pub include: Option<Option<Vec<RealtimeTranscriptionSessionCreateRequestIncludeItem>>>,
    #[serde(default, flatten, skip_serializing_if = "serde_json::Map::is_empty")]
    pub rest: Rest,
}

/// Source: `openai/types/realtime/audio_transcription.py`, `AudioTranscription`.
#[derive(Debug, Clone, PartialEq, Serialize, Deserialize, gproxy_protocol_macros::WireBuilder)]
#[cfg_attr(not(feature = "exhaustive"), non_exhaustive)]
#[derive(gproxy_protocol_macros::DeclaredFields)]
#[cfg_attr(feature = "schema", derive(schemars::JsonSchema))]
pub struct AudioTranscription {
    #[serde(
        default,
        deserialize_with = "present_nullable",
        skip_serializing_if = "Option::is_none"
    )]
    pub delay: Option<Option<AudioTranscriptionDelay>>,
    #[serde(
        default,
        deserialize_with = "present_nullable",
        skip_serializing_if = "Option::is_none"
    )]
    pub keywords: Option<Option<Vec<String>>>,
    #[serde(
        default,
        deserialize_with = "present_nullable",
        skip_serializing_if = "Option::is_none"
    )]
    pub language: Option<Option<String>>,
    #[serde(
        default,
        deserialize_with = "present_nullable",
        skip_serializing_if = "Option::is_none"
    )]
    pub languages: Option<Option<Vec<String>>>,
    #[serde(
        default,
        deserialize_with = "present_nullable",
        skip_serializing_if = "Option::is_none"
    )]
    pub model: Option<Option<String>>,
    #[serde(
        default,
        deserialize_with = "present_nullable",
        skip_serializing_if = "Option::is_none"
    )]
    pub prompt: Option<Option<String>>,
    #[serde(default, flatten, skip_serializing_if = "serde_json::Map::is_empty")]
    pub rest: Rest,
}

/// Source: `openai/types/realtime/realtime_transcription_session_audio.py`, `RealtimeTranscriptionSessionAudio`.
#[derive(Debug, Clone, PartialEq, Serialize, Deserialize, gproxy_protocol_macros::WireBuilder)]
#[cfg_attr(not(feature = "exhaustive"), non_exhaustive)]
#[derive(gproxy_protocol_macros::DeclaredFields)]
#[cfg_attr(feature = "schema", derive(schemars::JsonSchema))]
pub struct RealtimeTranscriptionSessionAudio {
    #[serde(
        default,
        deserialize_with = "present_nullable",
        skip_serializing_if = "Option::is_none"
    )]
    pub input: Option<Option<RealtimeTranscriptionSessionAudioInput>>,
    #[serde(default, flatten, skip_serializing_if = "serde_json::Map::is_empty")]
    pub rest: Rest,
}

/// Source: `openai/types/realtime/noise_reduction_type.py`, `NoiseReductionType`.
#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
#[cfg_attr(not(feature = "exhaustive"), non_exhaustive)]
#[derive(gproxy_protocol_macros::DeclaredFields)]
#[cfg_attr(feature = "schema", derive(schemars::JsonSchema))]
pub enum NoiseReductionType {
    #[serde(rename = "near_field")]
    NearField,
    #[serde(rename = "far_field")]
    FarField,
}

/// Source: `openai/types/realtime/realtime_transcription_session_audio_input.py`, `RealtimeTranscriptionSessionAudioInput`.
#[derive(Debug, Clone, PartialEq, Serialize, Deserialize, gproxy_protocol_macros::WireBuilder)]
#[cfg_attr(not(feature = "exhaustive"), non_exhaustive)]
#[derive(gproxy_protocol_macros::DeclaredFields)]
#[cfg_attr(feature = "schema", derive(schemars::JsonSchema))]
pub struct RealtimeTranscriptionSessionAudioInput {
    #[serde(
        default,
        deserialize_with = "present_nullable",
        skip_serializing_if = "Option::is_none"
    )]
    pub format: Option<Option<RealtimeAudioFormats>>,
    #[serde(
        default,
        deserialize_with = "present_nullable",
        skip_serializing_if = "Option::is_none"
    )]
    pub noise_reduction: Option<Option<RealtimeTranscriptionSessionAudioInputNoiseReduction>>,
    #[serde(
        default,
        deserialize_with = "present_nullable",
        skip_serializing_if = "Option::is_none"
    )]
    pub transcription: Option<Option<AudioTranscription>>,
    #[serde(
        default,
        deserialize_with = "present_nullable",
        skip_serializing_if = "Option::is_none"
    )]
    pub turn_detection: Option<Option<RealtimeTranscriptionSessionAudioInputTurnDetection>>,
    #[serde(default, flatten, skip_serializing_if = "serde_json::Map::is_empty")]
    pub rest: Rest,
}

/// Source: `openai/types/realtime/realtime_transcription_session_audio_input.py`, `NoiseReduction`.
#[derive(Debug, Clone, PartialEq, Serialize, Deserialize, gproxy_protocol_macros::WireBuilder)]
#[cfg_attr(not(feature = "exhaustive"), non_exhaustive)]
#[derive(gproxy_protocol_macros::DeclaredFields)]
#[cfg_attr(feature = "schema", derive(schemars::JsonSchema))]
pub struct RealtimeTranscriptionSessionAudioInputNoiseReduction {
    #[serde(rename = "type")]
    #[serde(
        default,
        deserialize_with = "present_nullable",
        skip_serializing_if = "Option::is_none"
    )]
    pub type_: Option<Option<NoiseReductionType>>,
    #[serde(default, flatten, skip_serializing_if = "serde_json::Map::is_empty")]
    pub rest: Rest,
}

/// Source: `openai/types/realtime/realtime_transcription_session_audio_input_turn_detection.py`, `RealtimeTranscriptionSessionAudioInputTurnDetection`.
#[derive(Debug, Clone, PartialEq, Serialize, Deserialize)]
#[cfg_attr(not(feature = "exhaustive"), non_exhaustive)]
#[serde(untagged)]
#[derive(gproxy_protocol_macros::DeclaredFields)]
#[cfg_attr(feature = "schema", derive(schemars::JsonSchema))]
pub enum RealtimeTranscriptionSessionAudioInputTurnDetection {
    RealtimeTranscriptionSessionAudioInputTurnDetectionServerVad(
        RealtimeTranscriptionSessionAudioInputTurnDetectionServerVad,
    ),
    RealtimeTranscriptionSessionAudioInputTurnDetectionSemanticVad(
        RealtimeTranscriptionSessionAudioInputTurnDetectionSemanticVad,
    ),
    Variant3(()),
}

/// Source: `openai/types/realtime/realtime_transcription_session_audio_input_turn_detection.py`, `ServerVad`.
#[derive(Debug, Clone, PartialEq, Serialize, Deserialize, gproxy_protocol_macros::WireBuilder)]
#[cfg_attr(not(feature = "exhaustive"), non_exhaustive)]
#[derive(gproxy_protocol_macros::DeclaredFields)]
#[cfg_attr(feature = "schema", derive(schemars::JsonSchema))]
pub struct RealtimeTranscriptionSessionAudioInputTurnDetectionServerVad {
    #[serde(rename = "type")]
    pub type_: RealtimeTranscriptionSessionAudioInputTurnDetectionServerVadType,
    #[serde(
        default,
        deserialize_with = "present_nullable",
        skip_serializing_if = "Option::is_none"
    )]
    pub create_response: Option<Option<bool>>,
    #[serde(
        default,
        deserialize_with = "present_nullable",
        skip_serializing_if = "Option::is_none"
    )]
    pub idle_timeout_ms: Option<Option<i64>>,
    #[serde(
        default,
        deserialize_with = "present_nullable",
        skip_serializing_if = "Option::is_none"
    )]
    pub interrupt_response: Option<Option<bool>>,
    #[serde(
        default,
        deserialize_with = "present_nullable",
        skip_serializing_if = "Option::is_none"
    )]
    pub prefix_padding_ms: Option<Option<i64>>,
    #[serde(
        default,
        deserialize_with = "present_nullable",
        skip_serializing_if = "Option::is_none"
    )]
    pub silence_duration_ms: Option<Option<i64>>,
    #[serde(
        default,
        deserialize_with = "present_nullable",
        skip_serializing_if = "Option::is_none"
    )]
    pub threshold: Option<Option<serde_json::Number>>,
    #[serde(default, flatten, skip_serializing_if = "serde_json::Map::is_empty")]
    pub rest: Rest,
}

/// Source: `openai/types/realtime/realtime_transcription_session_audio_input_turn_detection.py`, `SemanticVad`.
#[derive(Debug, Clone, PartialEq, Serialize, Deserialize, gproxy_protocol_macros::WireBuilder)]
#[cfg_attr(not(feature = "exhaustive"), non_exhaustive)]
#[derive(gproxy_protocol_macros::DeclaredFields)]
#[cfg_attr(feature = "schema", derive(schemars::JsonSchema))]
pub struct RealtimeTranscriptionSessionAudioInputTurnDetectionSemanticVad {
    #[serde(rename = "type")]
    pub type_: RealtimeTranscriptionSessionAudioInputTurnDetectionSemanticVadType,
    #[serde(
        default,
        deserialize_with = "present_nullable",
        skip_serializing_if = "Option::is_none"
    )]
    pub create_response: Option<Option<bool>>,
    #[serde(
        default,
        deserialize_with = "present_nullable",
        skip_serializing_if = "Option::is_none"
    )]
    pub eagerness:
        Option<Option<RealtimeTranscriptionSessionAudioInputTurnDetectionSemanticVadEagerness>>,
    #[serde(
        default,
        deserialize_with = "present_nullable",
        skip_serializing_if = "Option::is_none"
    )]
    pub interrupt_response: Option<Option<bool>>,
    #[serde(default, flatten, skip_serializing_if = "serde_json::Map::is_empty")]
    pub rest: Rest,
}

#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
#[cfg_attr(not(feature = "exhaustive"), non_exhaustive)]
#[derive(gproxy_protocol_macros::DeclaredFields)]
#[cfg_attr(feature = "schema", derive(schemars::JsonSchema))]
pub enum RealtimeTranscriptionSessionCreateRequestType {
    #[serde(rename = "transcription")]
    Transcription,
}

#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
#[cfg_attr(not(feature = "exhaustive"), non_exhaustive)]
#[derive(gproxy_protocol_macros::DeclaredFields)]
#[cfg_attr(feature = "schema", derive(schemars::JsonSchema))]
pub enum RealtimeTranscriptionSessionCreateRequestIncludeItem {
    #[serde(rename = "item.input_audio_transcription.logprobs")]
    ItemInputAudioTranscriptionLogprobs,
}

#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
#[cfg_attr(not(feature = "exhaustive"), non_exhaustive)]
#[derive(gproxy_protocol_macros::DeclaredFields)]
#[cfg_attr(feature = "schema", derive(schemars::JsonSchema))]
pub enum AudioTranscriptionDelay {
    #[serde(rename = "minimal")]
    Minimal,
    #[serde(rename = "low")]
    Low,
    #[serde(rename = "medium")]
    Medium,
    #[serde(rename = "high")]
    High,
    #[serde(rename = "xhigh")]
    Xhigh,
}

#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
#[cfg_attr(not(feature = "exhaustive"), non_exhaustive)]
#[derive(gproxy_protocol_macros::DeclaredFields)]
#[cfg_attr(feature = "schema", derive(schemars::JsonSchema))]
pub enum RealtimeTranscriptionSessionAudioInputTurnDetectionServerVadType {
    #[serde(rename = "server_vad")]
    ServerVad,
}

#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
#[cfg_attr(not(feature = "exhaustive"), non_exhaustive)]
#[derive(gproxy_protocol_macros::DeclaredFields)]
#[cfg_attr(feature = "schema", derive(schemars::JsonSchema))]
pub enum RealtimeTranscriptionSessionAudioInputTurnDetectionSemanticVadType {
    #[serde(rename = "semantic_vad")]
    SemanticVad,
}

#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
#[cfg_attr(not(feature = "exhaustive"), non_exhaustive)]
#[derive(gproxy_protocol_macros::DeclaredFields)]
#[cfg_attr(feature = "schema", derive(schemars::JsonSchema))]
pub enum RealtimeTranscriptionSessionAudioInputTurnDetectionSemanticVadEagerness {
    #[serde(rename = "low")]
    Low,
    #[serde(rename = "medium")]
    Medium,
    #[serde(rename = "high")]
    High,
    #[serde(rename = "auto")]
    Auto,
}