> ## Documentation Index > Fetch the complete documentation index at: https://impossibl.com/docs/llms.txt > Use this file to discover all available pages before exploring further. # Answer typed questions about an input (OpenAI Decisions) > Mirrors OpenAI's native `POST /v1/decisions` exactly, so the official OpenAI SDKs work by pointing `OPENAI_BASE_URL` at this API. Send an `input` (text, or user messages with `input_text`/base64 `input_image` parts) plus an array of typed `questions` (predicate, choice, score), each with a unique `name`, and receive one typed answer per question with probabilities. Requires openai/gpt-6-luna-decisions; other models return 400. Billed on `usage.input_tokens` only ($0.10/M, $0.20/M above 272K input tokens); output and cache tokens are never billed. OpenAI's own validation errors are returned as 400. ## OpenAPI ````yaml /openapi.yaml post /v1/decisions openapi: 3.1.0 info: title: Impossibl AI API version: 0.0.1 description: > The Impossibl AI API is one API for all AI providers, OpenAI-compatible across OpenAI, Anthropic, Google, xAI, and more (`GET /v1/models` lists the full catalog), billed from a prepaid credit balance. Usage is billed at the model rates published by GET /v1/models with no usage markup. Prepaid credit purchases charge a separate fixed 5% platform fee before tax. **Get started** Create a workspace key at https://impossibl.com/dashboard, then call a model with `Authorization: Bearer ` and a full model ID from `GET /v1/models`. The walkthrough is at https://impossibl.com/docs/quickstart. Agents can create a funded account and initial key with `POST /v1/accounts`. The current API continues to use `/v1`; there is no `/v2` base URL. Conversation history is managed by your application. Provider-specific SDK endpoints and fields outside this specification are not guaranteed compatible. servers: - url: https://api.impossibl.com description: Production - url: http://localhost:8787 description: Local dev security: - BearerKey: [] paths: /v1/decisions: post: summary: Answer typed questions about an input (OpenAI Decisions) description: >- Mirrors OpenAI's native `POST /v1/decisions` exactly, so the official OpenAI SDKs work by pointing `OPENAI_BASE_URL` at this API. Send an `input` (text, or user messages with `input_text`/base64 `input_image` parts) plus an array of typed `questions` (predicate, choice, score), each with a unique `name`, and receive one typed answer per question with probabilities. Requires openai/gpt-6-luna-decisions; other models return 400. Billed on `usage.input_tokens` only ($0.10/M, $0.20/M above 272K input tokens); output and cache tokens are never billed. OpenAI's own validation errors are returned as 400. operationId: createDecision requestBody: required: true content: application/json: schema: $ref: '#/components/schemas/DecisionsRequest' responses: '200': description: >- One answer per question, in order, each echoing its question's name, plus token usage. content: application/json: schema: $ref: '#/components/schemas/DecisionsResponse' '400': $ref: '#/components/responses/BadRequest' '401': $ref: '#/components/responses/Unauthorized' '402': $ref: '#/components/responses/InsufficientCredits' '404': $ref: '#/components/responses/UnknownModel' '502': $ref: '#/components/responses/UpstreamError' '503': $ref: '#/components/responses/ProviderNotConfigured' components: schemas: DecisionsRequest: type: object required: - model - input - questions properties: model: type: string example: openai/gpt-6-luna-decisions input: description: >- Text, or user messages whose content holds `input_text` and `input_image` (base64 data URL) parts. oneOf: - type: string - type: array items: type: object questions: type: array minItems: 1 items: type: object required: - type - name - instructions properties: type: type: string enum: - predicate - choice - score name: type: string description: Unique within the request; the answer echoes it. instructions: type: string choices: type: array minItems: 2 description: Required for `choice`. items: type: object required: - value properties: value: type: string description: type: string levels: type: array description: Required for `score`, ordered lowest to highest. items: type: object required: - label properties: label: type: string description: type: string DecisionsResponse: type: object required: - model - answers - usage properties: model: type: string example: gpt-6-luna answers: type: array items: type: object required: - type - name description: >- `predicate` carries `probability` (0..1). `choice` carries `choice`, `confidence` and `probabilities` [{value, probability}]. `score` carries `score` (probability-weighted 0-based level index), `confidence` and `probabilities` [{value, label, probability}]. `refusal` carries only `name`. properties: type: type: string enum: - predicate - choice - score - refusal name: type: string probability: type: number minimum: 0 maximum: 1 choice: type: string score: type: number confidence: type: number minimum: 0 maximum: 1 probabilities: type: array items: type: object usage: type: object required: - input_tokens - output_tokens properties: input_tokens: type: integer minimum: 0 description: The billed count. output_tokens: type: integer minimum: 0 description: Always 0. total_tokens: type: integer minimum: 0 Error: type: object properties: error: type: object properties: message: type: string type: type: string description: >- Common types include authentication_error, invalid_request_error, insufficient_quota, billing_required, permission_error, upstream_error, and provider_not_configured. responses: BadRequest: description: Malformed request (bad JSON, or missing required fields). content: application/json: schema: $ref: '#/components/schemas/Error' example: error: message: '`model` and `messages` are required' type: invalid_request_error Unauthorized: description: Missing or invalid API key. content: application/json: schema: $ref: '#/components/schemas/Error' example: error: message: Invalid API key type: authentication_error InsufficientCredits: description: >- Available credits or postpaid headroom cannot cover the request's balance check, which can include estimated input cost. Top up or reduce the request size. content: application/json: schema: $ref: '#/components/schemas/Error' example: error: message: Insufficient credits. Top up to continue. type: insufficient_quota UnknownModel: description: The requested `model` id is not in the registry (see GET /v1/models). content: application/json: schema: $ref: '#/components/schemas/Error' example: error: message: 'Unknown model: foo/bar' type: invalid_request_error UpstreamError: description: The upstream provider returned an error. content: application/json: schema: $ref: '#/components/schemas/Error' example: error: message: 'permission-denied: ...' type: upstream_error ProviderNotConfigured: description: The provider for this model has no API key configured on the gateway. content: application/json: schema: $ref: '#/components/schemas/Error' example: error: message: 'Provider not configured: missing OPENAI_API_KEY' type: provider_not_configured securitySchemes: BearerKey: type: http scheme: bearer description: Gateway API key (`imp-rt-...`) from POST /v1/accounts or /v1/keys. ```` This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.