Struct GenerateRequest 

pub struct GenerateRequest {
    pub prompt: String,
    pub model: Option<String>,
    pub params: GenerateParams,
    pub context: Option<String>,
    pub tools: Option<Vec<Value>>,
    pub images: Option<Vec<ContentBlock>>,
    pub messages: Option<Vec<Message>>,
    pub cache_control: bool,
    pub response_format: Option<ResponseFormat>,
    pub intent: Option<IntentHint>,
}

A text generation request.

Default is derived so call sites can mutate just the fields they care about: GenerateRequest { prompt: "...".into(), ..Default::default() }. The default prompt = "" is useless on its own — callers always override it — but the ..Default::default() shorthand stops the per-call-site mechanical churn every time a new optional field lands (closes #109).

Fields

prompt: String

The prompt to complete (first user message for single-turn).

model: Option<String>

Optional model override.

params: GenerateParams

Generation parameters.

context: Option<String>

Optional memory context to prepend to the prompt. When provided, this is injected as a system-level context block before the user prompt, grounding the model’s response.

tools: Option<Vec<Value>>

Optional tool definitions for structured tool_use. When provided, the model may return tool_calls instead of text. Each tool is a JSON object with: name, description, parameters (JSON Schema).

images: Option<Vec<ContentBlock>>

Optional images for vision models. When provided with a single-turn prompt, these are included as image content blocks in the user message. For multi-turn with messages, use UserMultimodal variants instead.

messages: Option<Vec<Message>>

Optional multi-turn conversation history. When provided, the backend builds a proper multi-turn message array instead of a single user message. The prompt field is ignored when messages are present.

cache_control: bool

Enable prompt caching for Anthropic API. When true, system prompt and tools are marked with cache_control breakpoints, enabling cache reuse across parent/child agent calls sharing the same prefix.

response_format: Option<ResponseFormat>

Constrain output to JSON (optionally schema-validated). See ResponseFormat for the per-provider mapping. Defaults to None — free-form text.

intent: Option<IntentHint>

Caller-supplied routing intent. None preserves the existing adaptive vs. pinned-model behavior. When Some, the adaptive router uses the hint to filter candidates (hard require), override task selection, and bias the score profile (prefer_local). See crate::intent::IntentHint.

Trait Implementations

impl Clone for GenerateRequest

fn clone(&self) -> GenerateRequest

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for GenerateRequest

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

Formats the value using the given formatter.

impl Default for GenerateRequest

fn default() -> GenerateRequest

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for GenerateRequest

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for GenerateRequest

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

where __S: Serializer,

Serialize this value into the given Serde serializer.