# Generated by API Evangelist (build-phrasing.py). Our phrasing, not observed demand.
overlay: 1.0.0
info:
  title: API Evangelist conversational phrasing for Hugging Face Text Generation API
  version: 1.0.0
extends: openapi/hugging-face-text-generation-api-openapi.yml
actions:
- target: $.info
  update:
    x-apievangelist-phrasing:
      method: generated
      generated: '2026-10-01'
      generator: build-phrasing.py
      label: Generated by API Evangelist
      operations: 3
- target: $.paths['/v1/completions'].post
  update:
    x-apievangelist-phrasing:
      intent: Complete a prompt via the completions API
      effect: write
      questions:
      - Can I use the OpenAI-style completions format to get a model to finish a prompt?
      - Does the completions endpoint accept stop sequences, a seed and streaming?
      instructions:
      - text: Complete {prompt} with model {model}.
        slots:
          prompt: requestBody.prompt
          model: requestBody.model
      - text: Get a completion of {prompt} from {model} up to {max_tokens} tokens, stopping at {stop}.
        slots:
          prompt: requestBody.prompt
          model: requestBody.model
          max_tokens: requestBody.max_tokens
          stop: requestBody.stop
      method: generated
      generated: '2026-10-01'
- target: $.paths['/generate'].post
  update:
    x-apievangelist-phrasing:
      intent: Generate text in one response on TGI
      effect: write
      questions:
      - Can I send a raw prompt to TGI's native generate route and get all the text back at once?
      - What generation parameters does the non-streaming TGI generate call accept?
      instructions:
      - text: Generate text from {inputs} on the TGI server and return it all at once.
        slots:
          inputs: requestBody.inputs
      - text: Run TGI generate on {inputs} with parameters {parameters}.
        slots:
          inputs: requestBody.inputs
          parameters: requestBody.parameters
      method: generated
      generated: '2026-10-01'
- target: $.paths['/generate_stream'].post
  update:
    x-apievangelist-phrasing:
      intent: Stream generated tokens over SSE
      effect: write
      questions:
      - Can I receive tokens incrementally over server-sent events while TGI generates?
      - Which TGI route streams output instead of waiting for the full text?
      instructions:
      - text: Stream tokens for {inputs} from the TGI server as they are generated.
        slots:
          inputs: requestBody.inputs
      - text: Stream a generation of {inputs} over SSE with parameters {parameters}.
        slots:
          inputs: requestBody.inputs
          parameters: requestBody.parameters
      method: generated
      generated: '2026-10-01'