BoundaryML/baml · error · io::Error

WASM history boundary {} was not begun

Error message

WASM history boundary {} was not begun

What it means

append_log_body in the WASM runs store requires a history boundary that was previously begun via begin(). The call looks up the BoundaryId in the boundaries map and, when absent, returns an io::Error of kind NotFound saying the boundary 'was not begun'. It means the caller is appending a log body to a run whose lifecycle was never started (or already finished/removed).

Source

Thrown at baml_language/crates/bridge_wasm/src/runs.rs:114

        value_writer.append_run_started(&RunStartedRecord {
            request: start.request.clone(),
            created_at_ms: start.created_at_ms,
            time_anchor: start.time_anchor,
        })?;
        self.boundaries
            .insert(start.boundary_id, WasmHistoryBoundary { value_writer });
        Ok(())
    }

    fn append_log_body(
        &mut self,
        boundary_id: BoundaryId,
        event: LogEventRecord,
        codec: ValueCodec,
        body: Vec<u8>,
    ) -> io::Result<ValueWriteOutcome> {
        let boundary = self.boundaries.get_mut(&boundary_id).ok_or_else(|| {
            io::Error::new(
                io::ErrorKind::NotFound,
                format!(
                    "WASM history boundary {} was not begun",
                    boundary_id.to_wire_string()
                ),
            )
        })?;
        boundary.value_writer.append_log_body(codec, body, event)
    }

    fn append_capture_loss(
        &mut self,
        boundary_id: BoundaryId,
        record: &CaptureLossRecord,
    ) -> io::Result<()> {
        let boundary = self.boundaries.get_mut(&boundary_id).ok_or_else(|| {
            io::Error::new(
                io::ErrorKind::NotFound,

View on GitHub (pinned to bd85ce9dee)

Solutions

  1. Ensure begin() is called for the boundary before any append_log_body call
  2. Verify the BoundaryId is the one returned by begin(), not reconstructed from wire bytes of a closed run
  3. Check ordering in your JS glue code so bodies are appended before the run is finished/dropped
  4. Handle the io NotFound error in the JS caller and re-begin the boundary if replaying history

Example fix

// before
runs.append_log_body(boundaryId, event, codec, body);
// after
if (!runs.hasBoundary(boundaryId)) boundaryId = runs.begin(meta);
runs.append_log_body(boundaryId, event, codec, body);
Defensive patterns

Strategy: validation

Validate before calling

function canAppendBody(runs, boundaryId) {
  return typeof runs.hasBoundary === 'function' && runs.hasBoundary(boundaryId);
}

Try / catch

try {
  runs.append_log_body(boundaryId, event, codec, body);
} catch (e) {
  if (String(e.message).includes('was not begun')) {
    boundaryId = runs.begin(meta); // re-begin and retry once
    runs.append_log_body(boundaryId, event, codec, body);
  } else throw e;
}

Prevention

When it happens

Trigger: Calling append_log_body with a BoundaryId that was never created, that was already completed, or from a stale reference after the boundary map dropped it.

Common situations: Happens in WASM test harnesses and replay paths (e.g. warm_history_replays_log_payload_and_body_without_live_runstore) when log bodies are appended after the run was torn down, or when boundary IDs are recreated/reused across page reloads.

Understand the failure class

Background: Record Not Found Errors: "not found", RecordNotFound, and "was not found" — what they mean and how to fix them — this error's family across 28 libraries.

Related errors

AI-assisted analysis of BoundaryML/baml@bd85ce9dee (2026-09-12). Data as JSON: /api/errors/b0b3f32e6f2952d0. Report an issue: GitHub.