Struct RetryPolicy 

pub struct RetryPolicy {
    pub max_attempts: u32,
    pub initial_backoff: Duration,
    pub max_backoff: Duration,
    pub total_deadline: Duration,
}

Bounds on a retry sequence.

Two independent limits apply, and whichever trips first wins: max_attempts caps how many times the work is tried, total_deadline caps the wall-clock time the whole sequence may consume. The deadline is what keeps a per-attempt timeout from multiplying — a 600 s send timeout retried four times must not become a 40-minute hang.

Fields

max_attempts: u32

Total attempts, including the first. 1 disables retrying.

initial_backoff: Duration

Backoff base for the first retry; doubles each subsequent retry.

max_backoff: Duration

Ceiling on any single delay, including a server-supplied Retry-After.

total_deadline: Duration

Ceiling on the wall-clock time the whole sequence may consume.

Implementations

impl RetryPolicy

pub fn from_env() -> Self

RetryPolicy::default with any environment overrides applied.

Resolved once per process: these are operator settings, and re-reading them per request would cost a syscall on every completion for a value that cannot meaningfully change mid-run.

An unset or unparseable variable leaves that field at its default — a typo degrades to standard behaviour rather than disabling retry.

pub fn disabled() -> Self

A policy that makes one attempt and never retries.

What the CLI’s --no-retry resolves to, and what a caller wanting strictly one-shot behaviour should ask for by name rather than by knowing that max_attempts: 1 means “off”.

fn with_env_overrides<F, E>(self, lookup: &F) -> Self

where F: Fn(&'static str) -> Result<String, E>,

Apply overrides read through lookup, which is the process environment in production and a fixture in tests.

Trait Implementations

impl Clone for RetryPolicy

fn clone(&self) -> RetryPolicy

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for RetryPolicy

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

Formats the value using the given formatter.

impl Default for RetryPolicy

fn default() -> Self

Defaults tuned for the LLM completion path.

Deliberately modest: the proxy already absorbs ModelLoading server-side, so a client-side retry is covering startup contention, which by definition means something upstream has already waited a long time. A larger budget here would stack on top of that and turn an unlucky request into a multi-minute hang.

impl PartialEq for RetryPolicy

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

Tests for self and other values to be equal, and is used by ==.

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

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl Copy for RetryPolicy

impl Eq for RetryPolicy

impl StructuralPartialEq for RetryPolicy