Struct SemanticProgressSink 

pub struct SemanticProgressSink { /* private fields */ }

Open the sink file (append, create) on first event, cache the handle in a Mutex. We deliberately accept the cost of a Mutex over every event because the JSONL stream is several events per batch, not per row — even a 50ms batch wall-time dwarfs the lock cost.

Implementations

impl SemanticProgressSink

pub fn open(tier: &str, embedder_id: &str) -> Self

Open a sink for the given tier+embedder. Returns a no-op sink when the env var is unset, so callers can always emit events unconditionally without branching.

pub fn disabled() -> Self

Sink that never writes — kept as an explicit factory so callers can default to a sink without consulting the env var (e.g. tests that don’t care about telemetry).

pub fn is_active(&self) -> bool

True if the sink is actively writing (env var set + file opened). Callers can branch on this to skip building expensive SemanticProgressFields when no one will read them.

pub fn emit(&self, event: SemanticProgressEvent, fields: SemanticProgressFields)

Emit one event. Best-effort: a write failure logs at debug and returns Ok — telemetry never bubbles errors into the backfill.

pub fn emit_bare(&self, event: SemanticProgressEvent)

Convenience: emit an event with no extra fields.

pub fn pid(&self) -> u32

Process-pid for cross-correlation. Stable for the life of the sink.