Struct CreateChatCompletionRequest 

pub struct CreateChatCompletionRequest {

    pub messages: Vec<ChatCompletionRequestMessage>,
    pub model: String,
    pub mm_processor_kwargs: Option<Value>,
    pub store: Option<bool>,
    pub reasoning_effort: Option<ReasoningEffort>,
    pub metadata: Option<Value>,
    pub frequency_penalty: Option<f32>,
    pub logit_bias: Option<HashMap<String, Value>>,
    pub logprobs: Option<bool>,
    pub top_logprobs: Option<u8>,
    pub max_tokens: Option<u32>,
    pub max_completion_tokens: Option<u32>,
    pub n: Option<u8>,
    pub modalities: Option<Vec<ResponseModalities>>,
    pub prediction: Option<PredictionContent>,
    pub audio: Option<ChatCompletionAudio>,
    pub presence_penalty: Option<f32>,
    pub response_format: Option<ResponseFormat>,
    pub seed: Option<i64>,
    pub service_tier: Option<ServiceTier>,
    pub stop: Option<Stop>,
    pub stream: Option<bool>,
    pub stream_options: Option<ChatCompletionStreamOptions>,
    pub temperature: Option<f32>,
    pub top_p: Option<f32>,
    pub tools: Option<Vec<ChatCompletionTool>>,
    pub tool_choice: Option<ChatCompletionToolChoiceOption>,
    pub parallel_tool_calls: Option<bool>,
    pub user: Option<String>,
    pub prompt_cache_key: Option<String>,
    pub function_call: Option<ChatCompletionFunctionCall>,
    pub functions: Option<Vec<ChatCompletionFunctions>>,
    pub web_search_options: Option<WebSearchOptions>,
}

Chat completion request with multimodal processor support.

Extends upstream CreateChatCompletionRequest with:

• mm_processor_kwargs: multimodal processor configuration (vLLM-specific)
• Uses our extended ChatCompletionRequestMessage (with reasoning, video/audio)
• Uses our extended ChatCompletionStreamOptions (with continuous_usage_stats)

Fields

messages: Vec<ChatCompletionRequestMessage>

model: String

mm_processor_kwargs: Option<Value>

Multimodal processor configuration (vLLM-specific)

store: Option<bool>

reasoning_effort: Option<ReasoningEffort>

metadata: Option<Value>

frequency_penalty: Option<f32>

logit_bias: Option<HashMap<String, Value>>

logprobs: Option<bool>

top_logprobs: Option<u8>

max_tokens: Option<u32>

👎Deprecated

max_completion_tokens: Option<u32>

n: Option<u8>

modalities: Option<Vec<ResponseModalities>>

prediction: Option<PredictionContent>

audio: Option<ChatCompletionAudio>

presence_penalty: Option<f32>

response_format: Option<ResponseFormat>

seed: Option<i64>

service_tier: Option<ServiceTier>

stop: Option<Stop>

stream: Option<bool>

stream_options: Option<ChatCompletionStreamOptions>

temperature: Option<f32>

top_p: Option<f32>

tools: Option<Vec<ChatCompletionTool>>

tool_choice: Option<ChatCompletionToolChoiceOption>

parallel_tool_calls: Option<bool>

user: Option<String>

prompt_cache_key: Option<String>

OpenAI cache-affinity hint: requests sharing a prompt prefix send the same key (Kimi Code CLI sends its session id on every request).

NOTICE: accepted and preserved only. Nothing in this crate or in Dynamo acts on it yet — Dynamo’s KV-aware router keys on prompt-prefix block hashes, not on this value.

function_call: Option<ChatCompletionFunctionCall>

👎Deprecated

functions: Option<Vec<ChatCompletionFunctions>>

👎Deprecated

web_search_options: Option<WebSearchOptions>

Implementations

impl CreateChatCompletionRequest

pub fn dynamic_system_tools(&self) -> impl Iterator<Item = &Value>

Kimi-style dynamic tools declared on system messages, in message order.

Kimi defines these as coexisting with the top-level tools list: a dynamic declaration keeps its position in the message history so the prompt prefix (and any KV cache built on it) stays intact. Do not fold them into tools; reason about the union with Self::has_effective_tools and Self::effective_tool_contains.

Only system messages can carry tools in the typed schema; the developer message is the upstream type and has no such field.

pub fn has_effective_tools(&self) -> bool

Whether the request declares any tool, either top-level or through a dynamic system-message declaration.

Gates that decide whether model output may be interpreted as tool calls must use this rather than tools alone, or a call to a dynamically declared tool is stripped from the response.

pub fn effective_tool_names(&self) -> impl Iterator<Item = &str>

Names of every tool the model can see: top-level tools first, then dynamic system-message tools in message order.

pub fn effective_tool_contains(&self, name: &str) -> bool

Whether name is declared anywhere in the effective tool set.

Use this to validate a named tool_choice so a forced call to a dynamically declared tool is not rejected as “not present in tools”.

Trait Implementations

impl Clone for CreateChatCompletionRequest

fn clone(&self) -> CreateChatCompletionRequest

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for CreateChatCompletionRequest

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

Formats the value using the given formatter.

impl Default for CreateChatCompletionRequest

fn default() -> CreateChatCompletionRequest

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for CreateChatCompletionRequest

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl PartialEq for CreateChatCompletionRequest

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

Equality operator ==.

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

Inequality operator !=.

impl Serialize for CreateChatCompletionRequest

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

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl StructuralPartialEq for CreateChatCompletionRequest