Learning evidence from the first action.
Seneca Careers converts supported learning actions into xAPI statement data when they are received. It retains activity time, links related events, and labels how the evidence arrived. It is a capture and export service, not a certified Learning Record Store (LRS).
Three ways in
| Route | How it works | What the record says |
|---|---|---|
| Instrument a learning app | Its server sends attempts, questions and revisions through a scoped source key. | Source submission, unverified until a provider attestation process exists. |
| Import Seneca history | The learner uploads their own xAPI export. | Self-imported, unverified. Original date, ID reference and content hash retained. |
| Add evidence | The learner adds a work excerpt and context. | Self-reported; no claim of independent observation. |
Server-side SDK
Create a source key in Data connections after enabling capture. Store it on the learning application's server. Every key is bound to one learner; never use one key for a class or organization. This pilot does not implement partner OAuth enrollment or automatic connection to Zipper.
import { createSenecaCapture } from './capture-sdk.mjs';
const capture = createSenecaCapture({sourceKey: process.env.SENECA_SOURCE_KEY});
// Persist this event in your app's transactional outbox before sending.
const event = {
id: crypto.randomUUID(),
verb: 'revised',
title: 'Market research recommendation',
evidence: 'Replaced a single estimate with a sensitivity analysis after feedback.',
occurredAt: new Date().toISOString(),
assistance: 'human',
rightsConfirmed: true
};
await capture.record(event);
// On retry, reuse the SAME id and event body.
The API contract
POST https://vindicaseneca.com/api/careers/capture accepts
the event above with
Authorization: Bearer <scoped-source-key>. The
endpoint returns the persisted statement. Identical ID/content retries
return the original record; changed content under the same ID returns
409. This endpoint accepts a Seneca event envelope, not arbitrary LRS
statements. It does not advertise the LRS HTTP contract or an xAPI
version header.
A same-origin, signed-in Seneca session can also add activity through the workspace. Sources must not manufacture learner actions from inferred personality, page visits, hiring interactions or tutor behavior. An application must actually instrument a meaningful learning action; xAPI does not collect data by itself.
Supported learning vocabulary
| Action | Meaning |
|---|---|
| attempted | The learner attempted a task. It does not imply correctness. |
| asked | The learner asked a question. |
| was-corrected | Feedback or correction addressed the learner's work. |
| revised | The learner revised earlier work. |
| produced | The learner produced an artifact. |
| abandoned | An activity was left unfinished; no trait inference follows. |
Optional episodeId groups related actions using a UUID.
relatedStatementId links an existing activity owned by the
same learner. A subsequent revision should carry a new ID and a
reference to the original, never overwrite it. Voiding creates a
separate StatementRef correction; exports retain correction history.
Identity, versions and provenance
Actors use the existing Seneca opaque account pseudonym, derived server-side from the verified Seneca identity. The browser cannot choose another learner. This pilot uses the xAPI 1.0.3 statement shape to interoperate with the existing Seneca export builder. xAPI 2.0 is the current specification; a future LRS delivery adapter must negotiate and test its transport version separately.
Native capture does not set stored, authority,
or version: these belong to the receiving LRS. The service
stores its own received_at alongside the statement. An LRS
credential never belongs in a browser.
Custom vocabulary lives under
https://vindicaseneca.com/xapi/careers/. The
profiles/learning-process/v1 category identifies this
application profile; activity-types/work describes the
learner's task. activities/{id} identifies a private task
and resolves here without disclosing it. Extension names are
provenance, source, assistance,
source-url, related-statement-id,
original-statement-id, and original-sha256.
Assistance is none, ai, human, mixed, or unspecified; none means none
reported, not independently verified.
What an import does
The importer accepts one learner's Seneca learning-process statements, up to 50 per request. It projects them to a new deterministic statement ID and the current learner's account, and adds an unverified-import label plus original statement ID and hash. It never silently rewrites an existing statement under its original ID. Only the supported activity verbs are accepted; assessment scores, arbitrary statement types, email actors and unknown properties are rejected. Importing a file is not proof that its contents are true.
Employer access is separate
A job application shares only the profile and introductory note. Sharing learning evidence requires a selection of existing statements, a named reviewing employer account, a role purpose and an expiry of at most 90 days. New activities are excluded. Revocation and withdrawal stop subsequent access. Screenshots or previously seen information cannot be recalled.
What still needs a source integration
Zipper hooks, third-party learning apps, LMS webhooks and Drive revisions each need an adapter, authorization and a semantic mapping. No source is automatically connected by creating an account here. The SDK is ready to be integrated, but it does not install itself into those products. This first deployment supports one employer reviewer account per company and does not include billing, applicant ranking, automated hiring decisions or attachment processing.