Inferless Inference API

openapi: 3.0.1
info:
  title: Inferless Inference API
  description: Specification of the Inferless serverless GPU inference platform. Covers the per-deployment model inference endpoint (KServe v2 style inputs[] payload) and the workspace-scoped REST management API for model settings and logs. The inference host is auto-generated per deployment (m-<id>.<region>.model-v1.inferless.com); management APIs are served from https://api.inferless.com. All requests authenticate with a workspace API key passed as a Bearer token in the Authorization header.
  termsOfService: https://www.inferless.com/terms
  contact:
    name: Inferless Support
    url: https://www.inferless.com
  version: '1.0'
servers:
- url: https://api.inferless.com
  description: Inferless management API base URL.
- url: https://{modelId}.{region}.model-v1.inferless.com
  description: Per-deployment inference host. The full URL is auto-generated and shown on the model's API page in the Inferless console.
  variables:
    modelId:
      default: m-xxxxxxxx
      description: The deployed model identifier prefix.
    region:
      default: default
      description: The deployment region segment.
security:
- bearerAuth: []
tags:
- name: Inference
paths:
  /v2/inference/{model_name}/infer:
    post:
      operationId: runInference
      tags:
      - Inference
      summary: Run inference against a deployed model.
      description: Sends an inference request to a deployed model's auto-generated endpoint. The request body uses the KServe v2 inputs[] structure (name, shape, datatype, data). This path is served from the per-deployment inference host, not from api.inferless.com.
      servers:
      - url: https://{modelId}.{region}.model-v1.inferless.com
        variables:
          modelId:
            default: m-xxxxxxxx
          region:
            default: default
      parameters:
      - name: model_name
        in: path
        required: true
        schema:
          type: string
        description: The model name segment of the auto-generated inference URL.
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/InferenceRequest'
            example:
              inputs:
              - name: prompt
                shape:
                - 1
                datatype: BYTES
                data:
                - What is AI?
      responses:
        '200':
          description: Inference result.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/InferenceResponse'
        '401':
          description: Missing or invalid API key.
components:
  schemas:
    InferenceResponse:
      type: object
      description: Model output. Inferless returns the dictionary produced by the model's inference function, optionally wrapped in an outputs[] array.
      properties:
        outputs:
          type: array
          items:
            $ref: '#/components/schemas/InferenceTensor'
    InferenceRequest:
      type: object
      required:
      - inputs
      properties:
        inputs:
          type: array
          items:
            $ref: '#/components/schemas/InferenceTensor'
    InferenceTensor:
      type: object
      required:
      - name
      - shape
      - datatype
      - data
      properties:
        name:
          type: string
          description: Input parameter name as defined in the model input schema.
          example: prompt
        shape:
          type: array
          items:
            type: integer
          description: Tensor dimensions; use [-1] for variable-length inputs.
          example:
          - 1
        datatype:
          type: string
          description: Data type of the input.
          enum:
          - BYTES
          - STRING
          - INT8
          - INT32
          - FP32
          example: BYTES
        data:
          type: array
          items: {}
          description: The input values matching the declared shape and datatype.
          example:
          - What is AI?
  securitySchemes:
    bearerAuth:
      type: http
      scheme: bearer
      description: Workspace API key passed as a Bearer token in the Authorization header.