openapi: 3.0.1
info:
  title: Predibase Adapters Inference API
  description: 'Specification of the Predibase API surfaces documented at https://docs.predibase.com. Two planes are covered: (1) the inference data plane on https://serving.app.predibase.com, which exposes an OpenAI-compatible chat/completions interface plus native generate / generate_stream text-generation endpoints, scoped per tenant and deployment; and (2) the control plane on https://api.app.predibase.com, which manages fine-tuning jobs, adapter repositories, deployments, datasets, and base models. All endpoints authenticate with a Predibase API token sent as an HTTP Bearer token.'
  termsOfService: https://predibase.com/terms-of-service
  contact:
    name: Predibase Support
    email: support@predibase.com
    url: https://docs.predibase.com
  version: '2.0'
servers:
- url: https://serving.app.predibase.com/{tenant}/deployments/v2/llms/{model}
  description: Inference (serving) base. tenant is your Predibase tenant ID (Settings > My Profile); model is the deployment name (Deployments page). The OpenAI-compatible routes live under the /v1 suffix of this base.
  variables:
    tenant:
      default: TENANT_ID
      description: Predibase tenant ID.
    model:
      default: DEPLOYMENT_NAME
      description: Deployment name (base model deployment).
- url: https://api.app.predibase.com/v2
  description: Control plane base for fine-tuning, adapters, deployments, datasets, and models.
security:
- bearerAuth: []
tags:
- name: Inference
paths:
  /v1/chat/completions:
    post:
      operationId: createChatCompletion
      tags:
      - Inference
      summary: OpenAI-compatible chat completion.
      description: Creates a chat completion against the deployment named in the server URL. OpenAI Chat Completions v1 compatible. Set model to an adapter ID ("repo"/"version") to serve a fine-tuned LoRA on top of the base model, or leave it empty to use the base model. Set stream to true to receive Server-Sent Events.
      servers:
      - url: https://serving.app.predibase.com/{tenant}/deployments/v2/llms/{model}
        variables:
          tenant:
            default: TENANT_ID
          model:
            default: DEPLOYMENT_NAME
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/ChatCompletionRequest'
      responses:
        '200':
          description: A chat completion (or an SSE stream when stream is true).
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/ChatCompletionResponse'
            text/event-stream:
              schema:
                type: string
  /v1/completions:
    post:
      operationId: createCompletion
      tags:
      - Inference
      summary: OpenAI-compatible text completion.
      description: Creates a text completion against the deployment named in the server URL. OpenAI Completions v1 compatible. Set model to an adapter ID to serve a fine-tuned LoRA.
      servers:
      - url: https://serving.app.predibase.com/{tenant}/deployments/v2/llms/{model}
        variables:
          tenant:
            default: TENANT_ID
          model:
            default: DEPLOYMENT_NAME
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/CompletionRequest'
      responses:
        '200':
          description: A text completion.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/CompletionResponse'
  /generate:
    post:
      operationId: generate
      tags:
      - Inference
      summary: Native text generation.
      description: Generates text from a deployed model. The full URL is https://serving.app.predibase.com/{tenant}/deployments/v2/llms/{deployment}/generate. Optionally specify a LoRA adapter via parameters.adapter_id and parameters.adapter_source (pbase, hub, or s3).
      servers:
      - url: https://serving.app.predibase.com/{tenant}/deployments/v2/llms/{model}
        variables:
          tenant:
            default: TENANT_ID
          model:
            default: DEPLOYMENT_NAME
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/GenerateRequest'
      responses:
        '200':
          description: Generated text.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/GenerateResponse'
  /generate_stream:
    post:
      operationId: generateStream
      tags:
      - Inference
      summary: Native streaming text generation.
      description: Streams generated tokens from a deployed model as Server-Sent Events. The full URL is https://serving.app.predibase.com/{tenant}/deployments/v2/llms/{deployment}/generate_stream.
      servers:
      - url: https://serving.app.predibase.com/{tenant}/deployments/v2/llms/{model}
        variables:
          tenant:
            default: TENANT_ID
          model:
            default: DEPLOYMENT_NAME
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/GenerateRequest'
      responses:
        '200':
          description: An SSE stream of generated token events.
          content:
            text/event-stream:
              schema:
                type: string
components:
  schemas:
    ChatCompletionRequest:
      type: object
      required:
      - messages
      properties:
        model:
          type: string
          description: Adapter ID in the form "repo"/"version" to serve a fine-tuned LoRA, or empty to use the deployment's base model.
        messages:
          type: array
          items:
            $ref: '#/components/schemas/ChatMessage'
        max_tokens:
          type: integer
        temperature:
          type: number
        top_p:
          type: number
        stream:
          type: boolean
          description: When true, the response is delivered as Server-Sent Events.
    ChatMessage:
      type: object
      required:
      - role
      - content
      properties:
        role:
          type: string
          enum:
          - system
          - user
          - assistant
        content:
          type: string
    ChatCompletionResponse:
      type: object
      properties:
        id:
          type: string
        object:
          type: string
        created:
          type: integer
        model:
          type: string
        choices:
          type: array
          items:
            type: object
            properties:
              index:
                type: integer
              message:
                $ref: '#/components/schemas/ChatMessage'
              finish_reason:
                type: string
        usage:
          $ref: '#/components/schemas/Usage'
    CompletionRequest:
      type: object
      required:
      - prompt
      properties:
        model:
          type: string
          description: Adapter ID ("repo"/"version") or empty for the base model.
        prompt:
          type: string
        max_tokens:
          type: integer
        temperature:
          type: number
        stream:
          type: boolean
    GenerateResponse:
      type: object
      properties:
        generated_text:
          type: string
        details:
          type: object
    Usage:
      type: object
      properties:
        prompt_tokens:
          type: integer
        completion_tokens:
          type: integer
        total_tokens:
          type: integer
    GenerateRequest:
      type: object
      required:
      - inputs
      properties:
        inputs:
          type: string
          description: The prompt text. Include the fine-tuning prompt template when querying fine-tuned models.
        parameters:
          type: object
          properties:
            max_new_tokens:
              type: integer
            temperature:
              type: number
            adapter_id:
              type: string
              description: Adapter to apply, e.g. "my-repo/1".
            adapter_source:
              type: string
              enum:
              - pbase
              - hub
              - s3
              description: Where the adapter is loaded from.
    CompletionResponse:
      type: object
      properties:
        id:
          type: string
        object:
          type: string
        choices:
          type: array
          items:
            type: object
            properties:
              text:
                type: string
              index:
                type: integer
              finish_reason:
                type: string
        usage:
          $ref: '#/components/schemas/Usage'
  securitySchemes:
    bearerAuth:
      type: http
      scheme: bearer
      bearerFormat: Predibase API token
      description: 'Predibase API token sent as Authorization: Bearer <PREDIBASE_API_TOKEN>. Generate a token from Settings in the Predibase console.'