ReadonlyconfigReadonlyserviceStable, human-readable name for the service. Must be set explicitly on each subclass
because this.constructor.name gets mangled by minification in production builds,
which breaks tracing/observability span names.
ProtectedloggerLangSmith dataset to evaluate against, e.g. "Chart Review Run Evals".
OptionalexperimentPrefix?: stringOptional prefix for the LangSmith experiment name.
OptionalpatientId?: PatientIdWhen set, scope the experiment to this patient's examples — all review states, incoming included.
Optionalsplits?: string[]Dataset splits to evaluate, literally (union). Omit (with patientId unset) to run the golden split.
ProtectedderiveDerive the queue retentionSeconds from the effective dedup TTL. The queue overloads
retentionSeconds — it's BOTH the message lifetime AND the idempotency dedup window
(min(retentionSeconds, 24h)). So:
ttlHours·3600), capped at 24h (the window can't exceed it);delaySeconds + every retry),
else a short TTL would expire the message mid-retry — and the queue hard-rejects
delaySeconds > retentionSeconds.
ttlHours <= 0 means "no dedup" → undefined: the idempotency key is dropped (see
performLater), so there's no dedup window to bound — this just leaves the queue's default
message lifetime in place.Extract an entityId from params for job_executions tracking. Override in subclasses when applicable.
LangSmith dataset to evaluate against, e.g. "Chart Review Run Evals".
OptionalexperimentPrefix?: stringOptional prefix for the LangSmith experiment name.
OptionalpatientId?: PatientIdWhen set, scope the experiment to this patient's examples — all review states, incoming included.
Optionalsplits?: string[]Dataset splits to evaluate, literally (union). Omit (with patientId unset) to run the golden split.
Derive a deduplication key from the params. Return null to skip deduplication (every call enqueues).
LangSmith dataset to evaluate against, e.g. "Chart Review Run Evals".
OptionalexperimentPrefix?: stringOptional prefix for the LangSmith experiment name.
OptionalpatientId?: PatientIdWhen set, scope the experiment to this patient's examples — all review states, incoming included.
Optionalsplits?: string[]Dataset splits to evaluate, literally (union). Omit (with patientId unset) to run the golden split.
LangSmith dataset to evaluate against, e.g. "Chart Review Run Evals".
OptionalexperimentPrefix?: stringOptional prefix for the LangSmith experiment name.
OptionalpatientId?: PatientIdWhen set, scope the experiment to this patient's examples — all review states, incoming included.
Optionalsplits?: string[]Dataset splits to evaluate, literally (union). Omit (with patientId unset) to run the golden split.
Enqueue this job for async processing via Vercel Queues. Checks for deduplication via job_executions before sending.
LangSmith dataset to evaluate against, e.g. "Chart Review Run Evals".
OptionalexperimentPrefix?: stringOptional prefix for the LangSmith experiment name.
OptionalpatientId?: PatientIdWhen set, scope the experiment to this patient's examples — all review states, incoming included.
Optionalsplits?: string[]Dataset splits to evaluate, literally (union). Omit (with patientId unset) to run the golden split.
Optionaloptions: PerformLaterOptionsLangSmith dataset to evaluate against, e.g. "Chart Review Run Evals".
OptionalexperimentPrefix?: stringOptional prefix for the LangSmith experiment name.
OptionalpatientId?: PatientIdWhen set, scope the experiment to this patient's examples — all review states, incoming included.
Optionalsplits?: string[]Dataset splits to evaluate, literally (union). Omit (with patientId unset) to run the golden split.
Run the job with status tracking. Used by the queue handler. Errors propagate (not caught by safe()) so the RetryHandler can decide.
parentSpanContext (the enqueuer's span, recovered from the queue envelope) is attached
as a span LINK and the run span is started as a new trace root, so each job runs in its
own trace rather than collapsing a fan-out into one unbounded trace.
LangSmith dataset to evaluate against, e.g. "Chart Review Run Evals".
OptionalexperimentPrefix?: stringOptional prefix for the LangSmith experiment name.
OptionalpatientId?: PatientIdWhen set, scope the experiment to this patient's examples — all review states, incoming included.
Optionalsplits?: string[]Dataset splits to evaluate, literally (union). Omit (with patientId unset) to run the golden split.
OptionalparentSpanContext: SpanContextProtectedspanOptional hook for subclasses to attach extra non-PII span attributes derived from params.
The returned attributes appear on the service.${serviceName} span (and, for jobs,
on job.enqueue.* and job.run.* spans too).
Safe ID-like keys on params (e.g. patientId, templateIds, limit, cursor) are
already auto-extracted via toSpanAttributes("service.input", params) — you don't need
to enumerate them here. Use this hook for derived/computed attributes (counts, flags,
computed names) or for fields you've explicitly verified safe but that aren't in the
global allowlist. Hook output overrides auto-extracted attributes on key collision.
Safe by default: returns {} when not overridden. Only declare keys you've confirmed
are non-PII. Never include emails, names, dob, free-text, etc.
LangSmith dataset to evaluate against, e.g. "Chart Review Run Evals".
OptionalexperimentPrefix?: stringOptional prefix for the LangSmith experiment name.
OptionalpatientId?: PatientIdWhen set, scope the experiment to this patient's examples — all review states, incoming included.
Optionalsplits?: string[]Dataset splits to evaluate, literally (union). Omit (with patientId unset) to run the golden split.
ProtectedspanOptional hook for subclasses to attach extra non-PII span attributes derived from the
successful service result. Only called when run resolves successfully; on error, the
span is marked via markSpanError instead.
Safe by default: returns {} when not overridden. Only declare keys you've confirmed
are non-PII (e.g. counts, status flags, computed summaries). Never include emails,
names, dob, free-text, etc.
Runs a LangSmith offline experiment: each dataset example (one chart-review question) is run against the eval-branch clone and scored against its graded ground truth. Stamps the git commit as the "application version" so experiments stay comparable across prompt/config changes.
maxAttempts: 1— an experiment is not idempotent; a retry would re-run every question and create a duplicate experiment.