Compact reference for the Jev-compatible HTTP API. For a guided version with
examples, see Calling the API.
Base URL (default): http://127.0.0.1:8080.
| Method | Path | Purpose |
|---|
POST | /v1/systemone | Evaluate a state against typed questions. |
GET | /v1/models | List the served model. |
GET | /health | Readiness and model startup time. |
GET | /health/live | Liveness; independent of the model. |
| Field | Type | Required | Notes |
|---|
model | string | yes | Served name (default typed-lm) |
state | string or JSON | yes | The facts of the case |
questions | object | yes | Non-empty map of question name to question |
type | Required fields | Candidates |
|---|
noul | instructions | Optional criteria {"yes","no"} |
choice | instructions, criteria (object) | One entry per option |
score | instructions, criteria (array) | 2 to 10 levels, increasing |
| Field | Type | Notes |
|---|
model | string | Echoes the resolved model |
answers | object | One entry per question name |
usage.input_tokens | integer | Prompt tokens |
usage.output_tokens | integer | Decision positions read |
type | Fields |
|---|
noul | noul (0.0 to 1.0) |
choice | choice, probabilities, confidence |
score | score, legend, probabilities, confidence |
| Status | Situation |
|---|
200 | Success |
404 | Unknown model |
422 | Malformed body or question outside the contract |
500 | Inference failure |
Error envelope:
{ "error": { "message": "..." } }
{
"object": "list",
"data": [{ "id": "typed-lm", "object": "model", "owned_by": "typed-lm" }],
"models": [
{ "name": "typed-lm", "description": "typed-lm model served from context '...'", "release_date": "unknown" }
]
}
{ "status": "ok", "startup_seconds": 12.3 }
{ "status": "ok" }