AI Providers

GrillMyCode supports four AI providers via the ai_provider input. All providers use the same OpenAI-compatible chat completions API shape internally — swapping providers requires only a change to a small number of inputs.

Quick comparison

Provider	ai_provider value	Requires api_key	Requires azure_endpoint	Default
GitHub Models	github-models	No (uses github_token)	No	✓
OpenAI	openai	Yes	No
OpenRouter	openrouter	Yes	No
Azure OpenAI	azure-openai	Yes	Yes

---

GitHub Models (default)

Uses the GitHub Models inference endpoint. Authentication is handled automatically with the built-in GITHUB_TOKEN — no secrets need to be created.

When to use: The default for all workflows. No setup cost. Suitable for most classroom deployments.

Required inputs

Input	Value
ai_provider	github-models (or omit)
github_token	${{ secrets.GITHUB_TOKEN }} (or omit — it is the default)

Optional inputs

Input	Notes
ai_model	Defaults to gpt-4o. Any model listed on the GitHub Models marketplace is valid. Examples: gpt-4o-mini, Phi-3-mini-128k-instruct
api_key	Leave empty to use github_token automatically. Supply an instructor PAT here to authenticate calls under the instructor's account — see Using an instructor token below

Using an instructor token

By default, the action authenticates GitHub Models API calls with the built-in GITHUB_TOKEN. Because GITHUB_TOKEN represents the repository owner — in a GitHub Classroom context, that is the student's personal account — the rate limit tier applied is the one attached to the student's GitHub plan (typically the free tier).

Supplying an instructor's Personal Access Token via api_key changes whose account is billed:

Authentication	Rate limit tier used	Quota shared across…
GITHUB_TOKEN (default)	Student's plan	That student only
Instructor PAT	Instructor's plan	Every repo using the same PAT

Practical effect for a classroom:

• Under the default, each student's workflow runs against that student's own quota. The limits are low on free accounts, but they are isolated — one student hitting their limit does not affect others.
• Under an instructor PAT, all students share the instructor's single quota. If the instructor has a higher-tier plan (Team, Enterprise, or an active Copilot subscription), the per-request limit and context window may be larger. The trade-off is that a surge of simultaneous submissions pools all requests against one account.

How to set it up:

1. Generate a fine-grained Personal Access Token from the instructor's GitHub account. No repository permissions are needed — GitHub Models only requires the token to be valid.
2. Add the token as an organisation-level secret (or a secret on each student repository) named e.g. INSTRUCTOR_GITHUB_TOKEN.
3. Pass it via api_key:

- uses: NSCC-ITC-Assessment/GrillMyCode@v1
  with:
    github_token: ${{ secrets.GITHUB_TOKEN }}
    api_key: ${{ secrets.INSTRUCTOR_GITHUB_TOKEN }}

note

github_token is still required for GitHub API operations (posting PR comments, creating issues, etc.). Only the GitHub Models inference call is authenticated with api_key when it is supplied.

Minimal example

- uses: NSCC-ITC-Assessment/GrillMyCode@v1
  with:
    github_token: ${{ secrets.GITHUB_TOKEN }}

With a different model

- uses: NSCC-ITC-Assessment/GrillMyCode@v1
  with:
    github_token: ${{ secrets.GITHUB_TOKEN }}
    ai_provider: 'github-models'
    ai_model: 'gpt-4o-mini'

---

OpenAI

Uses the OpenAI API directly.

When to use: When you need access to the full OpenAI model catalogue, higher rate limits, or a specific model not available on GitHub Models.

Required secrets

Create the following secret in Settings → Secrets and variables → Actions:

Secret name	Description
OPENAI_API_KEY	Your OpenAI API key

Inputs

Input	Value
ai_provider	openai
api_key	${{ secrets.OPENAI_API_KEY }}
ai_model	Any valid OpenAI model identifier. Examples: gpt-4o, gpt-4-turbo, gpt-4o-mini

Example

- uses: NSCC-ITC-Assessment/GrillMyCode@v1
  with:
    github_token: ${{ secrets.GITHUB_TOKEN }}
    ai_provider: 'openai'
    ai_model: 'gpt-4o'
    api_key: ${{ secrets.OPENAI_API_KEY }}

---

OpenRouter

Uses the OpenRouter unified inference API, which provides access to models from many different providers through a single key.

When to use: When you want to use a model from a provider other than OpenAI or Azure (e.g. Anthropic Claude, Google Gemini, Meta Llama) without setting up a separate account for each.

Required secrets

Secret name	Description
OPENROUTER_API_KEY	Your OpenRouter API key

Inputs

Input	Value
ai_provider	openrouter
api_key	${{ secrets.OPENROUTER_API_KEY }}
ai_model	Any model identifier supported by OpenRouter, in provider/model-name format. See the OpenRouter model list. Examples: openai/gpt-4o, anthropic/claude-3-5-sonnet, meta-llama/llama-3.1-70b-instruct

Example

- uses: NSCC-ITC-Assessment/GrillMyCode@v1
  with:
    github_token: ${{ secrets.GITHUB_TOKEN }}
    ai_provider: 'openrouter'
    ai_model: 'anthropic/claude-3-5-sonnet'
    api_key: ${{ secrets.OPENROUTER_API_KEY }}

---

Azure OpenAI

Uses a model deployed in your own Azure OpenAI resource.

When to use: When institutional policy requires data to stay within an Azure tenant, or when a specific Azure deployment is already provisioned.

Required secrets

Secret name	Description
AZURE_OPENAI_API_KEY	API key for your Azure OpenAI resource
AZURE_OPENAI_ENDPOINT	Full endpoint URL for your deployment, e.g. https://my-resource.openai.azure.com/openai/deployments/my-deployment

Inputs

Input	Value
ai_provider	azure-openai
api_key	${{ secrets.AZURE_OPENAI_API_KEY }}
azure_endpoint	${{ secrets.AZURE_OPENAI_ENDPOINT }}
ai_model	Your Azure deployment name (not the underlying model name — use whatever you named the deployment)

Example

- uses: NSCC-ITC-Assessment/GrillMyCode@v1
  with:
    github_token: ${{ secrets.GITHUB_TOKEN }}
    ai_provider: 'azure-openai'
    ai_model: 'my-gpt4o-deployment'
    api_key: ${{ secrets.AZURE_OPENAI_API_KEY }}
    azure_endpoint: ${{ secrets.AZURE_OPENAI_ENDPOINT }}