Struct ChatCompletionRequest 

pub struct ChatCompletionRequest {

    pub model: String,
    pub messages: Vec<Message>,
    pub temperature: Option<f32>,
    pub top_p: Option<f32>,
    pub max_tokens: Option<u32>,
    pub max_completion_tokens: Option<u32>,
    pub stream: bool,
    pub stream_options: Option<Value>,
    pub tools: Vec<Tool>,
    pub tool_choice: Option<ToolChoice>,
    pub response_format: Option<ResponseFormat>,
    pub stop: Vec<String>,
    pub presence_penalty: Option<f32>,
    pub frequency_penalty: Option<f32>,
    pub n: Option<u32>,
    pub seed: Option<i64>,
    pub user: Option<String>,
    pub parallel_tool_calls: Option<bool>,
    pub reasoning_effort: Option<String>,
    pub tt_extras: HashMap<String, Value>,
    pub extra: HashMap<String, Value>,
}

Fields

model: String

messages: Vec<Message>

temperature: Option<f32>

top_p: Option<f32>

max_tokens: Option<u32>

max_completion_tokens: Option<u32>

Newer OpenAI spend cap for reasoning models (o3, o4-mini, …). When a client sets this it MUST be honored end-to-end — dropping it silently removes the caller’s output ceiling and changes spend.

stream: bool

stream_options: Option<Value>

OpenAI stream_options (e.g. { "include_usage": true }). Kept as an opaque value so the full object passes through to OpenAI-shaped upstreams unchanged.

tools: Vec<Tool>

tool_choice: Option<ToolChoice>

response_format: Option<ResponseFormat>

stop: Vec<String>

presence_penalty: Option<f32>

frequency_penalty: Option<f32>

n: Option<u32>

seed: Option<i64>

user: Option<String>

parallel_tool_calls: Option<bool>

Whether the model may emit tool calls in parallel (OpenAI parallel_tool_calls). Forwarded to OpenAI-shaped upstreams.

reasoning_effort: Option<String>

Reasoning-effort hint for reasoning models ("low"/"medium"/"high"). Materially changes output and cost, so it must reach the upstream when supported.

tt_extras: HashMap<String, Value>

TokenTrimmer-internal extras (cache config, route hints, etc.) that are stripped before forwarding to the provider.

extra: HashMap<String, Value>

Genuinely-unknown / newer OpenAI fields not modeled above. Captured via #[serde(flatten)] so they passthrough to OpenAI-shaped upstreams instead of being silently dropped on deserialize. Never includes the named fields above (serde consumes those first).

Trait Implementations

impl Clone for ChatCompletionRequest

fn clone(&self) -> ChatCompletionRequest

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for ChatCompletionRequest

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

Formats the value using the given formatter.

impl Default for ChatCompletionRequest

fn default() -> ChatCompletionRequest

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for ChatCompletionRequest

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for ChatCompletionRequest

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.