Developer access
Read API
A read-only HTTP interface over the same classified corpus the site shows. Every classification carries its confidence and rationale, so downstream systems can reproduce and contest a rating rather than inherit it blindly.
Authentication
Every request needs a key issued from the staff console, sent as a bearer token. Keys are stored only as fingerprints — the value is shown once at creation and cannot be recovered.
curl -H "Authorization: Bearer gnip_…" \ "https://<your-domain>/api/public/v1/articles?limit=25&content_type=N"
Endpoints
GET /api/public/v1/articles
Classified items, newest first, with content type, framing, confidences, rationale and headline fidelity.
- limit
- 1–200, default 50
- content_type
- N | E | A | UT
- orientation
- liberal | moderate | conservative | unknown
- topic
- exact topic label
- since
- ISO timestamp
GET /api/public/v1/stories
Multi-source clusters with each member item, so you can compare framing across outlets.
- limit
- 1–100, default 40
GET /api/public/v1/publishers
The monitored source registry with each outlet's reliability record and sample size.
Terms of interpretation
- Framing orientation is descriptive and jurisdiction-relative. It is never a measure of accuracy or trust, and must not be presented as one.
- The untrustworthy designation appears only where a named human reviewer applied it. It is never produced automatically.
- Confidence values are part of every classification. Ratings below the review threshold are provisional and may change.
- Only headlines, summaries, links and derived analysis are served. Full article text is never stored or redistributed.
Responses
All endpoints return JSON with a count and an items array. A missing, inactive or unknown key returns 401. Query failures return 500 with an error code and no detail.