Inference API
Every model runs behind one base URL, with the same key and the same request format:
https://api.olaverselab.com- Requests and responses are JSON. Send
Content-Type: application/json. - Every
POSTneeds your key as a bearer token:Authorization: Bearer YOUR_API_KEY.GETendpoints are open. See Authentication. - There is no warm-up step. Models start on demand, so you can call any endpoint straight away.
Endpoints
Section titled “Endpoints”| Capability | Endpoint | Batch | Batch limit |
|---|---|---|---|
| Decisions | POST /v1/decide |
POST /v1/decide/batch |
32 cases |
| Document extraction | POST /v1/extract |
POST /v1/extract/batch |
16 documents |
| Language identification | POST /v1/language-id |
POST /v1/language-id/batch |
256 texts |
| Diacritic restoration | POST /v1/diacritize |
POST /v1/diacritize/batch |
64 texts |
| Question generation | POST /v1/questions/generate |
POST /v1/questions/generate/batch |
64 passages |
| Model list | GET /v1/models |
The same models are also served at POST /v1/chat/completions (OpenAI format) and POST /v1/systemone (System One format). See OpenAI & OpenRouter clients.
Choosing a model
Section titled “Choosing a model”Each endpoint has a default model, and takes an optional model field to pick another. Model ids can be given as olaverse/<name> too, in any case: olaverse/MIST-Extract-1.0 and mist-extract-1.0 are the same model. An unknown model returns 400. The model catalog lists them all.
Batches
Section titled “Batches”Batch endpoints take a list in place of a single input and return results in the same order. In decision and extraction batches, give each item an id to match results back. Each item succeeds or fails on its own, so one bad item doesn’t sink the batch, and only successful items are billed.
Every billable response carries a usage block with what it was charged on: tokens, texts, characters or documents, depending on the model.
"usage": {"input_tokens": 312, "output_tokens": 0, "decisions": 3}See Pricing for how each unit is counted.
Timeouts
Section titled “Timeouts”Give your HTTP client a timeout of about 2 minutes. Most calls return in about a second, but long documents and large batches take longer. See Retries and timeouts.