API reference
The planned endpoint surface, so you can design against it. Reach out to join the closed beta or push back on the shape.
Base URL
https://api.jeantechnologies.com/v1
Authentication
Every request carries a bearer token.
curl https://api.jeantechnologies.com/v1/tokenizers \
-H "Authorization: Bearer $JEAN_API_KEY"
Keys are issued per engagement. There is no self-serve signup.
Endpoints
Serving. Call a pretrained model directly.
| Method | Path | Purpose |
|---|---|---|
GET | /models | List the models your key can reach |
GET | /models/{id} | Model card: version, context, adaptation state |
POST | /models/{id}/predict | Distribution over what a person does next |
POST | /models/{id}/embed | Interaction history to vector |
Semantic IDs. Give a catalog a vocabulary the model can generate.
| Method | Path | Purpose |
|---|---|---|
POST | /tokenizers | Fit a semantic ID tokenizer on your catalog |
GET | /tokenizers/{id} | Status and fit diagnostics |
POST | /semantic-ids | Assign semantic IDs to items |
POST | /generate | Recommend from your catalog by decoding semantic IDs |
The typical order is fit, poll until ready, assign, then generate. /predict and /embed need
no tokenizer and work the moment you have a key. See Models for which
endpoint answers which question, and Semantic IDs for the reasoning
behind the design.
Status codes
| Code | Meaning |
|---|---|
200 | Success |
202 | Accepted. The fit is queued and running asynchronously. |
400 | Malformed request body |
401 | Missing or invalid API key |
404 | Tokenizer not found |
422 | Schema validation error |
500 | Internal error |