Hyperbolic Discover Model And Complete

arazzo: 1.0.1
info:
  title: Hyperbolic Discover Model And Complete
  summary: List the catalog and run a base-model text completion against a non-instruct model.
  description: >-
    Targets Hyperbolic's legacy text completions endpoint, which serves base
    (non-instruct) models such as the Llama-3.1-405B base BF16 and FP8 variants
    unique to Hyperbolic. The workflow lists the live catalog to confirm the
    base model, then submits a prompt completion. Each step inlines its request
    and inline Authorization Bearer credential.
  version: 1.0.0
sourceDescriptions:
- name: modelsApi
  url: ../openapi/hyperbolic-models-api-openapi.yml
  type: openapi
- name: completionsApi
  url: ../openapi/hyperbolic-completions-api-openapi.yml
  type: openapi
workflows:
- workflowId: discover-model-and-complete
  summary: Confirm a base model from the catalog and generate a raw text completion.
  description: >-
    Reads the model catalog, then sends a prompt to the legacy completions
    endpoint for a base model. Useful for high-throughput base-model prompting
    where instruct chat formatting is not wanted.
  inputs:
    type: object
    required:
    - apiKey
    - model
    - prompt
    properties:
      apiKey:
        type: string
        description: Hyperbolic API key passed as a Bearer token.
      model:
        type: string
        description: Base model id (e.g. meta-llama/Meta-Llama-3.1-405B).
      prompt:
        type: string
        description: The raw text prompt to complete.
      maxTokens:
        type: integer
        description: Maximum number of tokens to generate.
      temperature:
        type: number
        description: Sampling temperature between 0 and 2.
  steps:
  - stepId: listCatalog
    description: >-
      List the live inference catalog so the base model can be confirmed before
      a completion is requested.
    operationId: listModels
    parameters:
    - name: Authorization
      in: header
      value: "Bearer $inputs.apiKey"
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      models: $response.body#/data
  - stepId: complete
    description: >-
      Submit the prompt to the base model and capture the generated text and
      token usage.
    operationId: createCompletion
    parameters:
    - name: Authorization
      in: header
      value: "Bearer $inputs.apiKey"
    requestBody:
      contentType: application/json
      payload:
        model: $inputs.model
        prompt: $inputs.prompt
        max_tokens: $inputs.maxTokens
        temperature: $inputs.temperature
        stream: false
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      completionId: $response.body#/id
      text: $response.body#/choices/0/text
      finishReason: $response.body#/choices/0/finish_reason
      totalTokens: $response.body#/usage/total_tokens
  outputs:
    models: $steps.listCatalog.outputs.models
    text: $steps.complete.outputs.text
    totalTokens: $steps.complete.outputs.totalTokens