Struct SimpleStreamOptions 

pub struct SimpleStreamOptions {

    pub api_key: Option<String>,
    pub timeout: Option<Duration>,
    pub max_retries: Option<u32>,
    pub max_retry_delay: Option<Duration>,
    pub headers: Option<BTreeMap<String, String>>,
    pub metadata: Option<BTreeMap<String, String>>,
    pub cache_retention: CacheRetention,
    pub session_id: Option<String>,
    pub signal: CancellationToken,
    pub reasoning: Option<ThinkingLevel>,
    pub max_tokens: Option<u64>,
    pub temperature: Option<f64>,
    pub thinking_budgets: Option<ThinkingBudgets>,
}

Options for a “simple” (non-agent) stream call — what compaction / smoke tests / bare provider usage pass. The agent loop wraps this with its richer equivalent when it drives a real turn (M2).

Fields

api_key: Option<String>

API key. A provider may fall back to an env var when None (M3 anthropic reads ANTHROPIC_API_KEY); faux ignores it.

timeout: Option<Duration>

Request timeout. Defaults to 600 seconds; None uses the same provider fallback and can be used by patches to clear a caller override.

max_retries: Option<u32>

Max retry attempts on retryable errors (408/409/429/>=500). None = 0.

max_retry_delay: Option<Duration>

Cap on the total backoff between retries. None = 60s.

headers: Option<BTreeMap<String, String>>

Extra HTTP headers to send (merged over provider defaults).

metadata: Option<BTreeMap<String, String>>

Free-form provider metadata (request id, trace id, …). Anthropic reads user_id for abuse tracking; faux ignores it.

cache_retention: CacheRetention

Cache-retention hint for prompt caching.

session_id: Option<String>

Optional session id (faux uses it to simulate prompt-cache hits).

signal: CancellationToken

Cancellation token. Providers must poll this between chunks and emit an Error { reason: Aborted } terminal event when cancelled. Defaults to a fresh un-cancelled token so Default-constructed options never abort.

reasoning: Option<ThinkingLevel>

Extended-thinking level. None/Off → thinking disabled (provider may still emit thinking: { type: "disabled" } when the model reasons). Mirrors TS SimpleStreamOptions.reasoning.

max_tokens: Option<u64>

Optional caller override for the requested output-token ceiling. None lets the provider fit the thinking budget inside Model::max_tokens. Mirrors TS SimpleStreamOptions.maxTokens.

temperature: Option<f64>

Sampling temperature. Only applied when the model supports it AND thinking is disabled (Anthropic rejects temperature under thinking). Mirrors TS StreamOptions.temperature.

thinking_budgets: Option<ThinkingBudgets>

Per-level reasoning token budgets for budget-based thinking models. Faux ignores this; anthropic uses it in adjust_max_tokens_for_thinking.

Implementations

impl SimpleStreamOptions

pub fn new() -> Self

pub fn with_api_key(self, key: impl Into<String>) -> Self

pub fn with_signal(self, signal: CancellationToken) -> Self

pub fn reasoning_level(&self) -> ThinkingLevel

Resolve the effective reasoning level, defaulting to Off when unset.

pub fn request_timeout(&self) -> Duration

Resolve the effective request timeout, including the provider fallback.

Trait Implementations

impl Clone for SimpleStreamOptions

fn clone(&self) -> Self

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for SimpleStreamOptions

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

Formats the value using the given formatter.

impl Default for SimpleStreamOptions

fn default() -> Self

Returns the “default value” for a type.