SystemOne
SystemOne is a decision endpoint. Unlike chat, it is not conversational and does not stream. You describe a situation once, ask one or more structured choice questions, and get back the selected choice, a confidence score, and per-option probabilities.
Endpoint
Section titled “Endpoint”- Method
POST- Path
/v1/systemone- Model
typesafe/jev- Streaming
false- Chat
false
Request body
Section titled “Request body”| Field | Type | Required | Description |
|---|---|---|---|
model |
string | Yes | The decision model. Use typesafe/jev. |
state |
string | Yes | A description of the situation the decision is made against. |
questions |
object | Yes | Questions keyed by an id you choose. Each key becomes the id in the response answers. |
Each value in questions is a choice question:
| Field | Type | Required | Description |
|---|---|---|---|
type |
string | Yes | The question type. Use "choice". |
instructions |
string | Yes | The question to answer, phrased as an instruction. |
options |
string[] | Yes | The options the model chooses between. |
criteria |
object | Yes | A record mapping each option to its decision criterion — the condition under which that option should be chosen. |
Example
Section titled “Example”Request:
curl https://gurouter.com/v1/systemone \ -H "Authorization: Bearer <API_KEY>" \ -H "Content-Type: application/json" \ -d '{ "model": "typesafe/jev", "state": "The production deploy finished cleanly with zero downtime and all health checks passing.", "questions": { "deploy_success": { "type": "choice", "instructions": "Did the deploy succeed?", "options": ["yes", "no"], "criteria": { "yes": "The state describes a deploy that completed without errors, downtime, or failing health checks.", "no": "The state describes a deploy that failed, had downtime, or had failing health checks." } } } }'Response (200 OK):
{ "model": "typesafe/jev", "answers": { "deploy_success": { "type": "choice", "choice": "yes", "confidence": 1, "probabilities": { "yes": 1, "no": 0 } } }, "usage": { "input_tokens": 335, "output_tokens": 32 }}Response body
Section titled “Response body”| Field | Type | Description |
|---|---|---|
model |
string | The model that produced the answers. |
answers |
object | Keyed by the same ids you sent in questions. |
answers.<id>.type |
string | The question type, "choice". |
answers.<id>.choice |
string | The selected option. |
answers.<id>.confidence |
number | Confidence in the selected choice, from 0 to 1. |
answers.<id>.probabilities |
object | A record mapping each option to its probability from 0 to 1. |
usage.input_tokens |
number | Input tokens billed for the request. |
usage.output_tokens |
number | Output tokens billed for the response. |
Billing
Section titled “Billing”SystemOne is billed on both usage.input_tokens and usage.output_tokens. Read those
fields from the response to account for the cost of each call.
Validation errors
Section titled “Validation errors”The endpoint validates the question structure. Two observed 400 responses:
| Case | Status | Message | Param |
|---|---|---|---|
Missing instructions |
400 |
Invalid input |
questions.<id>.instructions |
Missing criteria |
400 |
Invalid input: expected record, received undefined |
questions.<id>.criteria |
