small-model-agent.yml

Run a small, self-hosted (or OpenAI-compatible API endpoint) AI agent against a pull request, enforcing bounded retries against pre-check repository verification gates (YAML/Markdown linting, PHI tripwire scans) (gha#436, gha#415).

Division of labor (wai#39): wai explains, ai-config configures, gha runs. Companion to Morrison-Lab/ai-config#1292.

Inputs

Input	Type	Default	Description
pr-number	string	''	PR number to run agent against.
endpoint-url	string	''	OpenAI-compatible API endpoint URL (e.g. http://localhost:11434/v1).
model	string	'small-model'	Model name override passed to the agent runner.
max-iterations	string	'5'	Maximum loop iterations / bounded retries.
setup-r	boolean	false	Install R before running verification gates.
install-quarto	boolean	false	Install Quarto before running verification gates.
apt-packages	string	''	Extra apt packages to install before setup.
pip-packages	string	''	Python packages to install with pip3.
checkout-submodules	boolean	false	Check out git submodules.
run-gates	boolean	true	Run repository verification gates.
dry-run	boolean	false	Run in dry-run verification mode without network calls.

Secrets

Secret	Required	Description
API_KEY	no	Optional API key for remote endpoint authentication.
SUBMODULES_TOKEN	no	Optional token to check out private submodules.

Permissions

Grant contents: read and pull-requests: write.

Concurrency

Do not declare a top-level concurrency: block in your caller workflow. small-model-agent.yml manages per-PR concurrency internally on its agent job (group: small-model-agent-<PR>). Adding a top-level concurrency: block in the caller with a PR-scoped group name causes GitHub Actions to detect a deadlock between the top-level workflow and the nested job, cancelling the run immediately (gha#437).

Example

# Copy to .github/workflows/small-model-agent.yml in your repo.
name: Small Model Agent

on:
  workflow_dispatch:
    inputs:
      pr_number:
        description: 'Pull request number'
        type: string
        required: false
      endpoint_url:
        description: 'OpenAI-compatible API endpoint URL'
        type: string
        required: false

jobs:
  agent:
    permissions:
      contents: read
      pull-requests: write
    uses: Morrison-Lab/gha/.github/workflows/small-model-agent.yml@v2
    with:
      pr-number: ${{ inputs.pr_number }}
      endpoint-url: ${{ inputs.endpoint_url }}