Decisions

OpenAI Decisions-compatible synchronous evaluation. Send model, input (a string, or user messages with input_text and inline base64 input_image parts) and typed questions (predicate, choice, score); answers come back in question order with probabilities and confidence; a question the model declines comes back as a refusal answer. A bare model name is OpenAI’s, e.g. gpt-6-luna, which is called natively and answers with OpenAI’s body unchanged. Other decision models are addressed by their full catalog ID, e.g. typesafe/jev-latest, opper/kev-4b or greenpt/metis (list available endpoints with GET /v3/models?type=evaluation); those are reached by translation to System One and accept text input only, except empiriolabs/aplomb-1, which takes the images too. GreenPT Metis receives all 1–32 questions together in one upstream request. Translation preserves question order, probabilities and confidence, but joins text-message boundaries and does not expose native impact, score legends or additional answer metadata; use the System One endpoint for the native contract. Billed on input tokens. Uses a project-scoped Opper API key, Comply policies and provider entitlements. Cost and generation ID are returned in headers. Streaming is unsupported.

POST /v3/compat/decisions

Decisions API

cURL

curl --request POST \
  --url https://api.opper.ai/v3/compat/decisions \
  --header 'Authorization: Bearer <token>' \
  --header 'Content-Type: application/json' \
  --data '
{
  "input": "<string>",
  "model": "gpt-6-luna",
  "questions": [
    {
      "instructions": "<string>",
      "type": "predicate",
      "name": "<string>"
    }
  ]
}
'

200

{
  "answers": [
    {
      "probability": 0.5,
      "type": "predicate",
      "name": "<string>"
    }
  ],
  "model": "<string>",
  "usage": {
    "input_tokens": 1,
    "output_tokens": 1,
    "total_tokens": 1
  }
}

See the classification and scoring guide for working examples, model discovery, and usage and billing behavior.

Authorizations

Authorization · string · header · required

API key authentication. Pass your API key as a Bearer token.

Headers

X-Opper-Name · string

Function name for tracing and project-level guardrail function-scope filtering.

X-Opper-Parent-Span-Id · string<uuid>

Parent span ID for distributed tracing context.

X-Opper-Tags · string

Comma-separated key:value usage-attribution tags (e.g. tenant:acme,project:demo, max 8). Recorded on the generation's billing/metrics rows; group spend by any key via GET /v2/analytics/usage?group_by=. Header-borne twin of the /v3/session URL prefix tags (which win per key when both are present); opper.-prefixed keys and session_id are reserved. Malformed values return 400.

Body

application/json

input · string object[] · required

model · string · required

Minimum string length: 1

Example:

"gpt-6-luna"

questions · object[] · required

Minimum array length: 1

Option 1

Option 2

Option 3

Show child attributes

safety_identifier · string

Accepted for compatibility with OpenAI's API, then dropped: Opper neither forwards nor stores it. Requests to OpenAI carry Opper's own identifier for your project.

Response

200

application/json

Successful response

answers · object[] · required

One answer per question, in question order.

Option 1

Option 2

Option 3

Option 4

Show child attributes

model · string · required

usage · object · required

Show child attributes