Struct Request 

pub struct Request {

    pub provider: Provider,
    pub api_key: String,
    pub base_url: String,
    pub model: String,
    pub system_message: Option<String>,
    pub messages: Vec<Message>,
    pub reminder: Option<String>,
    pub tools: Vec<ToolDefinition>,
    pub tool_choice: Option<ToolChoice>,
    pub temperature: Option<f32>,
    pub max_tokens: Option<u32>,
    pub reasoning_effort: Option<ReasoningEffort>,
    pub response_format: Option<ResponseFormat>,
    pub extra_body: Map<String, Value>,
    pub max_retries: u32,
    pub retry_delay_ms: u64,
}

A self-contained, provider-agnostic chat-completion request.

Carries everything needed to hit an LLM API: provider, credentials, model, messages, tools, and tuning parameters.

Call stream() or complete() with a shared reqwest::Client to send the request.

Fields

provider: Provider

Which provider to use.

api_key: String

API key / token.

base_url: String

Base URL override. If empty, uses Provider::default_base_url.

model: String

Model identifier (e.g. "deepseek-chat", "gpt-4o").

system_message: Option<String>

Optional system prompt.

messages: Vec<Message>

Conversation history.

reminder: Option<String>

Dynamic runtime context appended after the stable conversation prefix.

tools: Vec<ToolDefinition>

Tools the model may call.

tool_choice: Option<ToolChoice>

How the model should select tools.

temperature: Option<f32>

Sampling temperature.

max_tokens: Option<u32>

Maximum tokens to generate.

reasoning_effort: Option<ReasoningEffort>

Reasoning effort hint. None means leave provider default in place; Some(ReasoningEffort::None) explicitly disables thinking on providers that support it.

response_format: Option<ResponseFormat>

Constrain the output format.

extra_body: Map<String, Value>

Arbitrary extra top-level JSON fields merged into the provider’s raw request body (e.g. prefix, thinking).

max_retries: u32

Maximum retries for transient errors. Default: 3.

retry_delay_ms: u64

Initial retry delay in milliseconds. Default: 1000.

Implementations

impl Request

pub fn new(provider: Provider, api_key: impl Into<String>) -> Self

Create a new request for the given provider and API key.

Sets sensible defaults: provider’s default base URL and model, 3 retries with 1 s initial delay, no system prompt.

pub fn deepseek(api_key: impl Into<String>) -> Self

Shortcut for Request::new(Provider::DeepSeek, api_key).

pub fn openai(api_key: impl Into<String>) -> Self

Shortcut for Request::new(Provider::OpenAI, api_key).

pub fn anthropic(api_key: impl Into<String>) -> Self

Shortcut for Request::new(Provider::Anthropic, api_key).

pub fn gemini(api_key: impl Into<String>) -> Self

Shortcut for Request::new(Provider::Gemini, api_key).

pub fn kimi(api_key: impl Into<String>) -> Self

Shortcut for Request::new(Provider::Kimi, api_key).

pub fn glm(api_key: impl Into<String>) -> Self

Shortcut for Request::new(Provider::Glm, api_key).

pub fn minimax(api_key: impl Into<String>) -> Self

Shortcut for Request::new(Provider::Minimax, api_key).

pub fn mimo(api_key: impl Into<String>) -> Self

Shortcut for Request::new(Provider::Mimo, api_key).

pub fn grok(api_key: impl Into<String>) -> Self

Shortcut for Request::new(Provider::Grok, api_key).

pub fn openrouter(api_key: impl Into<String>) -> Self

Shortcut for Request::new(Provider::OpenRouter, api_key).

pub fn base_url(self, url: impl Into<String>) -> Self

Override the base URL.

pub fn model(self, m: impl Into<String>) -> Self

Set the model.

pub fn system_prompt(self, p: impl Into<String>) -> Self

Set the system prompt.

pub fn reminder(self, p: impl Into<String>) -> Self

Set dynamic runtime context for this request only.

pub fn message(self, m: Message) -> Self

Append a message to the conversation.

pub fn user(self, text: impl Into<String>) -> Self

Append a user text message (convenience).

pub fn messages(self, msgs: Vec<Message>) -> Self

Set the full message history.

pub fn tools(self, tools: Vec<ToolDefinition>) -> Self

Set the tool definitions.

pub fn temperature(self, t: f32) -> Self

Set the temperature.

pub fn max_tokens(self, n: u32) -> Self

Set the max tokens.

pub fn reasoning_effort(self, e: ReasoningEffort) -> Self

Set the reasoning-effort hint. ReasoningEffort::None explicitly disables thinking on providers that support the toggle.

pub fn text(self) -> Self

Set the response format to plain text (the default).

pub fn json_schema( self, name: impl Into<String>, schema: Value, strict: bool, ) -> Self

Constrain output to a named JSON Schema (OpenAI json_schema mode).

Use schemars::schema_for!(T) to generate the schema:

let schema = serde_json::to_value(schemars::schema_for!(MyStruct)).unwrap();
let req = Request::openai(key).json_schema("my_struct", schema, true);

pub fn json(self) -> Self

Set the response format to JSON object mode.

The model will be constrained to emit a valid JSON object. You must also instruct the model to produce JSON in your system prompt or user message — the format flag alone is not sufficient for most providers.

pub fn retries(self, max: u32, initial_delay_ms: u64) -> Self

Set retry parameters.

pub fn extra_body(self, extra: Map<String, Value>) -> Self

Merge extra JSON fields into the request body.

pub fn effective_base_url(&self) -> &str

Resolve the effective base URL (custom or provider default).

pub async fn stream( &self, http: &Client, ) -> Result<BoxStream<'static, LlmEvent>, ApiError>

Send a streaming request and return a stream of LlmEvents.

pub async fn complete( &self, http: &Client, ) -> Result<CompleteResponse, ApiError>

Send a non-streaming request and return the complete response.

Trait Implementations

impl Clone for Request

fn clone(&self) -> Request

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for Request

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

Formats the value using the given formatter.