Struct CompletionRequest 

pub struct CompletionRequest<'a> {
    pub messages: &'a [Message],
    pub system_prompt: &'a str,
    pub tools: &'a [ToolSchema],
    pub max_tokens: Option<u32>,
    pub tool_choice: Option<ToolChoice>,
    pub thinking: Option<ThinkingMode>,
    pub effort: Option<EffortLevel>,
    pub output_schema: Option<Value>,
    pub enable_caching: bool,
    pub fallback_model: Option<String>,
    pub temperature: Option<f64>,
}

A request to the LLM API.

Fields

messages: &'a [Message]

system_prompt: &'a str

tools: &'a [ToolSchema]

max_tokens: Option<u32>

tool_choice: Option<ToolChoice>

Tool selection constraint.

thinking: Option<ThinkingMode>

Thinking/reasoning configuration.

effort: Option<EffortLevel>

Effort level for the response.

output_schema: Option<Value>

JSON schema for structured output mode.

enable_caching: bool

Enable prompt caching.

fallback_model: Option<String>

Fallback model if primary is overloaded.

temperature: Option<f64>

Temperature override.

Implementations

impl<'a> CompletionRequest<'a>

pub fn simple( messages: &'a [Message], system_prompt: &'a str, tools: &'a [ToolSchema], max_tokens: Option<u32>, ) -> Self

Create a simple request with just messages and system prompt.