Struct LiveModelProfile 

#[non_exhaustive]
pub struct LiveModelProfile {
    pub thinking: bool,
    pub thinking_level_required: bool,
    pub affective_dialog_flag: bool,
    pub proactivity_flag: bool,
    pub vertex_async_tools: bool,
    pub context_update: bool,
    pub code_execution: bool,
}

What a Live model accepts in its setup message, as far as this crate knows.

The model catalog moves faster than releases, so this is keyed on the model name and errs toward sending what the caller set: only models known to reject or ignore a field have it removed. Anything unrecognised gets LiveModelProfile::DEFAULT — every field passes through. Platform rules that hold for every model (Google AI has no proactivity, for one) live on SessionConfig, not here.

let p = LiveModelProfile::of(&ModelId::LIVE_3_8);
assert!(!p.thinking && !p.affective_dialog_flag && p.vertex_async_tools);
assert!(p.context_update);
let x = LiveModelProfile::of(&ModelId::LIVE_3_8_EXTENDED_THINKING);
assert!(x.thinking && x.thinking_level_required);
let native = LiveModelProfile::of(&ModelId::LIVE_2_5_FLASH_NATIVE_AUDIO);
assert_eq!(native, LiveModelProfile::GEMINI_2_5_NATIVE_AUDIO);
assert!(!native.code_execution && !p.code_execution);

Fields (Non-exhaustive)

Non-exhaustive structs could have additional fields added in future. Therefore, non-exhaustive structs cannot be constructed in external crates using the traditional Struct { .. } syntax; cannot be matched against without a wildcard ..; and struct update syntax will not work.

thinking: bool

Accepts thinkingConfig. Vertex AI Live strips it for every model except those that require a thinking level.

thinking_level_required: bool

Requires thinkingConfig.thinkingLevel: the setup is refused without it (“Thinking level must be specified for this model”).

affective_dialog_flag: bool

Accepts the enableAffectiveDialog flag. Gemini 3.8 Live has affective dialogue always on and closes the session (1007) when the flag is sent.

proactivity_flag: bool

Accepts the proactivity config. Always on in Gemini 3.8 Live, whose docs say not to send it.

vertex_async_tools: bool

Accepts behavior / scheduling on Vertex AI (Google AI always does).

context_update: bool

Accepts a contextUpdate message, which replaces the declared tools and the system instruction mid-session. Gemini 2.5 Live closes the session over one (1007, “Request contains an invalid argument”).

code_execution: bool

Accepts the built-in codeExecution tool on Google AI. Gemini 3.8 Live, 3.8 Live Extended Thinking and 2.5 native audio close the session at setup over it (1007, “Code Execution tool is not supported for this model”; measured 2026-10-11). Vertex AI was not measured.

Implementations

impl LiveModelProfile

pub const DEFAULT: Self

The profile of a model this crate has no special knowledge of.

pub const GEMINI_2_5_NATIVE_AUDIO: Self

Gemini 2.5 native audio: as DEFAULT, without the codeExecution tool.

pub const GEMINI_3_8_LIVE: Self

Gemini 3.8 Live. No thinking: thinkingLevel is refused and a thinking budget has no effect.

pub const GEMINI_3_8_LIVE_EXTENDED_THINKING: Self

Gemini 3.8 Live Extended Thinking: requires thinkingLevel, and answers a question that needs thought in a later turn (a spoken holding line first, interactionStatus: IN_PROGRESS).

pub fn of(model: &ModelId) -> Self

The profile for model, matched on its name (with or without a models/ or publisher path).

Trait Implementations

impl Clone for LiveModelProfile

fn clone(&self) -> LiveModelProfile

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Copy for LiveModelProfile

impl Debug for LiveModelProfile

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Eq for LiveModelProfile

impl PartialEq for LiveModelProfile

fn eq(&self, other: &LiveModelProfile) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl StructuralPartialEq for LiveModelProfile