Stylome API documentation
Analyze text with the Stylome developer API at https://stylome.io. This reference, its Markdown versions, the OpenAPI schema, and the interactive reference are public. Reading them does not require an API key. Scoring requires a Bearer credential and prepaid API credit.
The API is in early alpha. Endpoints and models can change without a guaranteed notice period. API credit is separate from a personal extension subscription.
Start here
- Read authentication to create a key and choose its scopes and retention mode.
- Read scoring to prepare a request and interpret the result.
- Read operations for read endpoints, error handling, timeouts, and retries.
- Read billing and privacy before sending customer content.
Every page has an equivalent Markdown representation: append .md to its URL, or request Accept: text/markdown. Normal browser requests receive HTML. The documentation needs no JavaScript. The optional interactive reference uses Swagger UI and JavaScript.
Public endpoints
All paths below use the base URL https://stylome.io. Send JSON request bodies with Content-Type: application/json.
| Method and path | Purpose | Authorization |
|---|---|---|
GET /api/v1/models |
Public model IDs, limits, normalization, capabilities, and price | Public |
POST /api/v1/score |
Analyze one text with an explicit model | Bearer; score:create |
GET /api/v1/key |
Inspect the current key identity and retention policy | Bearer; no extra read scope |
GET /api/v1/limits |
Inspect account quota and prepaid balance | Bearer; account:read |
GET /api/v1/usage |
Query account request-ledger usage | Bearer; usage:read |
GET /api/v1/status |
API/database reachability and inference configuration | Public |
The public OpenAPI schema contains exactly these six paths, including full request and response schemas. Website session, billing checkout, and private internal endpoints are outside this developer contract.
Check available models without a key
curl --fail-with-body https://stylome.io/api/v1/models
The response is an object with object: "list" and a data array. Each model includes id, display_name, description, recommended, status, optional deprecation and retirement dates, input limits, max_request_bytes, length_unit, normalization, operating points, response_kind, retention support, price, and pricing_version.
Use stylome-1.2-large for the three authorship classes and relative confidence. stylome-1.0-large has a separate legacy numerical response. Retired IDs, including stylome-large-1.0, are rejected. Read the live catalog before assuming a model is available.
Interpret results with care
A classification is evidence to review, not proof of a person's conduct. Results depend on the source, register, length, and distribution of the input. Confidence is relative nearest-neighbor support, not a probability of correctness. Operating-point names refer to calibration false-positive-rate targets, not guaranteed error rates on new text.
For model measurements and their stated datasets and limits, see transparency and methodology. Compare systems at a matched false-positive rate. The scoring reference explains which analyses the public response supports.
For integration help, contact [email protected] with a request ID. Do not send secret keys or private text in support logs.