AI Action

GitHub Action to interact with different AI model providers.

Usage

In order to use this action, you need to

1. create an API KEY for the AI Gateway
2. pick one of the supported models

Basic Text Generation

name: Basic text generation example
on:
  push:
    branches:
      - main

jobs:
  generate-text:
    runs-on: ubuntu-latest
    steps:
      - uses: vercel/ai-action@v2
        id: prompt
        with:
          prompt: 'Why is the sky blue?'
          model: 'openai/gpt5'
          api-key: ${{ secrets.AI_GATEWAY_API_KEY }}
      - run: echo ${{ steps.prompt.outputs.text }}

Text Generation with System Message

You can provide a system message to set the behavior or context for the AI model:

name: Text generation with system message
on:
  push:
    branches:
      - main

jobs:
  generate-text:
    runs-on: ubuntu-latest
    steps:
      - uses: vercel/ai-action@v2
        id: prompt
        with:
          system: 'You are a kindergarten teacher getting questions by 5 year old students'
          prompt: 'Why is the sky blue?'
          model: 'openai/gpt5'
          api-key: ${{ secrets.AI_GATEWAY_API_KEY }}
      - run: echo ${{ steps.prompt.outputs.text }}

Structured JSON Generation

When you provide a JSON schema, the action will generate structured data that conforms to your schema:

name: Structured data generation example
on:
  push:
    branches:
      - main

jobs:
  generate-recipe:
    runs-on: ubuntu-latest
    steps:
      - uses: vercel/ai-action@v2
        id: recipe
        with:
          prompt: 'Generate a lasagna recipe'
          schema: |
            {
              "$schema": "https://json-schema.org/draft/2020-12/schema",
              "type": "object",
              "properties": {
                "recipe": {
                  "type": "object",
                  "properties": {
                    "name": {"type": "string"},
                    "ingredients": {
                      "type": "array",
                      "items": {"type": "string"}
                    },
                    "steps": {
                      "type": "array",
                      "items": {"type": "string"}
                    }
                  },
                  "required": ["name", "ingredients", "steps"],
                  "additionalProperties": false
                }
              },
              "required": ["recipe"],
              "additionalProperties": false
            }
          model: 'openai/gpt-4.1'
          api-key: ${{ secrets.AI_GATEWAY_API_KEY }}
      - name: Use structured output
        run: |
          echo "Generated recipe JSON:"
          echo '${{ steps.recipe.outputs.json }}'
          
          # Parse and use specific fields
          echo "Recipe name: ${{ fromJson(steps.recipe.outputs.json).recipe.name }}"

Inputs

prompt

Required. The input prompt to generate the text from.

api-key

Required. An API KEY for the AI Gateway.

model

Required. An identifier from the list of provider models supported by the AI Gateway: https://vercel.com/ai-gateway/models

schema

Optional. A valid JSON Schema for structured output generation. When provided, the action will use generateObject to generate structured JSON data that conforms to the schema. The schema should be a valid JSON Schema (draft 2020-12 or compatible).

system

Optional. A system message to set the behavior or context for the AI model. This is useful for defining the role, personality, or instructions for the AI assistant. The system message is supported by both generateText() and generateObject() methods.

Outputs

text

The generated text by the model. When using structured generation with a schema, this contains the JSON string.

json

The generated JSON object when using structured generation with a schema. This output is only available when the schema input is provided.

Examples

• Issue triaging
• Calculate version bumps and release notes
• Detect spam in README changes
• add yours

How it works

The action is utilizing the AI SDK to send requests to the AI Gateway.

• Text Generation: Uses generateText() for basic text generation
• Structured Generation: Uses generateObject() when a JSON schema is provided, ensuring the output conforms to your specified structure

Contributing

contributing.md

License

MIT