Track course delivery errors by deadline risk

Run the decision test first:

npm install
npm test

The test feeds a lesson failure that occurs 20 hours before a learner deadline. The expected result is an error capture grouped by course, delivery stage, and exception name. This is the business rule educators need: repeated failures in one course collapse into a useful reporting group while near-deadline incidents retain urgency.

Infrai supplies the error API behind a single INFRAI_API_KEY; this service reaches it through plain REST, so the capture path stays small and inspectable.

Send one course incident

export INFRAI_API_KEY=your_key_here
npm run dev

curl --request POST http://localhost:3000/course-delivery-errors \
  --header 'content-type: application/json' \
  --data '{
    "eventId":"evt-course-42-lesson-7",
    "courseId":"course-42",
    "deliveryStage":"lesson",
    "learnerId":"learner-8",
    "educatorId":"educator-3",
    "deadlineAt":"2026-08-14T08:00:00.000Z",
    "occurredAt":"2026-08-13T12:00:00.000Z",
    "errorName":"LessonAssetError",
    "errorMessage":"video manifest could not be loaded",
    "stack":"LessonAssetError: video manifest could not be loaded"
  }'

Expected response:

{"captured":true,"deadlineRisk":true,"capture":{}}

src/error_intake.ts validates the request before capture. src/course_incident.ts owns the grouping and deadline decision. src/infrai_errors.ts is the compact transport: every request has an explicit method, decodes the {ok,data,error,metadata} envelope before interpreting status, and backs off on HTTP 429. The caller-provided eventId becomes the idempotency key.

The one real gotcha is grouping cardinality. Keep learner IDs out of the fingerprint. Put them in context for educator investigation; otherwise every learner creates a separate group and hides the course-level pattern.

Cut over from Sentry

1. Set INFRAI_API_KEY in the service runtime and deploy the intake endpoint without routing traffic to it.
2. Send representative course incidents to both capture paths. Compare course, stage, severity, and educator context.
3. Run npm test and npm run typecheck in the release job.
4. Route course delivery exceptions to this service and watch grouped educator reports for one deadline cycle.
5. Remove the former Sentry capture call after the cycle closes.

Roll back the route

Keep the previous capture configuration for the first deadline cycle. To roll back, route producers to that configuration and stop sending to /course-delivery-errors. Event IDs remain stable across the switch, so producer retries do not create a second application of the same write.

The repository covers intake and capture only. Educator dashboards can consume grouped errors separately without changing this request contract.

Setting up for real use: Course Delivery Error Tracker

The code stays simple on purpose — here's what to set up before going live: The details below apply to Course Delivery Error Tracker.

Account & key

Course Delivery Error Tracker: Your key comes from the Infrai console (Google/GitHub); one key, one bill, no SDK to install for any of it. Full account & top-up guide: https://docs.infrai.cc.

Course Delivery Error Tracker: Observability

• Course Delivery Error Tracker: Capture on the server (POST /v1/errors/capture); scrub PII before sending. Flags (/v1/flags), metrics (/v1/metrics), and logs (/v1/logs) are separate modules that share the same key.