← Amber English

Reading self-check: method & API

Use the original sentences to mark what you know, what you do not know, and what you want to leave for later.

Open self-check → · All tools & API guide

How contexts are selected

Amber uses its shared English vocabulary profile and NGSL coverage calculator. Unconfirmed or conflicting contexts come first, followed by reference-list gaps, repeated spellings, spelling order and original position. This order is a practical review order, not a difficulty score or a statistically representative sample. Each round shows 12 contexts; the API accepts 1–50.

Each distinct word and exact sentence is a separate item. Repeated identical sentences are shown once per spelling. Very long segments use an exact excerpt around the occurrence. Locations are paragraph-local UTF-16 offsets, with a source SHA-256 and method version. The reference is NGSL 1.2, CC BY-SA 4.0; see the shared profiler for counting rules and data provenance.

What your results mean

“I know this use” and “I do not know this use” are self-reports tied to the displayed context. They are not scored answers. Skipped items stay unmeasured; opening a page, clicking next or leaving is not an incorrect response. Knowing one use of bank does not confirm another use. No vocabulary-size score or whole-book comprehension estimate is produced.

Earlier records retain their source, time and granularity. Only exact contexts match, with book identity respected when provided; matching does not certify an imported record. Opposing statements remain available as conflicts in the exported report. New self-reports have source self-check and context granularity. The full report includes selected unmarked contexts and their pending IDs. Record downloads contain actual records only; skipped items are not saved as learning evidence. JSON preserves complete records; CSV protects spreadsheet formulas with an apostrophe.

No account or model call is required. Public inputs and responses are not stored on the server or saved to your account. Download before closing or refreshing. In Amber, explicit marks are synced to the current account for later snapshots; they do not unlock assessment eligibility or change already accepted reading versions.

API for agents and developers

POST JSON to /api/tools/vocabulary-self-check. Fields: text (required string), previous (optional word lines, CSV or record array), choices (optional array), limit (optional integer 1–50, default 12). First request selects contexts. Submit the same text, previous and limit with choices containing the returned item id, status known or unknown, and at as Unix milliseconds. Duplicate or stale choice IDs and unrecognized fields fail. Only submit choices explicitly provided by the person; Agent execution itself is not evidence of knowledge.

{"text":"The otter rested by the river bank.","previous":[],"limit":12}

Response schema vocabulary-self-check/1, method amber-vocabulary-self-check/1.0: source, reference, counts, items (selection-time state and exact source), submitted (new self-reports), merge (retained records, changes and conflicts). counts.pending describes the full candidate pool before this submission; counts.reported counts this request’s choices. Repeating identical input, including timestamps, produces identical output and has no saved side effects. Complete reproducible request and result.

Limits: 300,000 UTF-16 text characters; 600,000 characters for serialized prior records; 10,000 prior or merged records; 50 choices; 1,800,000 total request bytes. HTTP 400: correct input; 413: split input; 415: use application/json; 500: retry the same input. Results use no-store and noindex. No generated quiz or calibrated measurement is included. Method updated October 10, 2026.

Continue with record merging or spelling coverage.