Support Model

[nathro]

Add optional model selection and update Augment Agent reference

• Add a new optional model input to the composite action and forward it to augmentcode/augment-agent
• Document the model input in the README with an example (e.g., openai/gpt-4o-mini)
• Update the Augment Agent action reference from a pinned commit to augmentcode/augment-agent@v0

Why

• Allow users to choose the underlying model for cost/latency/quality trade-offs without changing defaults
• Keep compatibility with the latest v0 action series that accepts the model parameter

Notes

• This is backward compatible; if model is not provided, the agent’s default model is used
• Using the v0 tag follows GitHub’s recommended major-version pinning; teams that require strict pinning can switch back to a specific SHA

Usage example

- name: Run Augment Agent
  uses: augmentcode/augment-agent@v0
  with:
    augment_session_auth: ${{ secrets.AUGMENT_SESSION_AUTH }}
    github_token: ${{ secrets.GITHUB_TOKEN }}
    pull_number: ${{ github.event.pull_request.number }}
    repo_name: ${{ github.repository }}
    custom_context: ${{ steps.custom_context.outputs.context }}
    model: openai/gpt-4o-mini

---

🤖 This description was generated automatically. Please react with 👍 if it's helpful or 👎 if it needs improvement.

[github-actions]

Thanks for the update adding optional model passthrough and documenting it. Overall looks good. I have one suggestion around supply-chain security and build reproducibility.

---

🤖 Automated review complete. Please react with 👍 or 👎 on the individual review comments to provide feedback on their usefulness.

[github-actions]

Security and reproducibility nit: consider pinning the action to a commit SHA instead of a floating tag. Using @v0 can pick up changes unexpectedly, which makes builds harder to reproduce and increases supply-chain risk.

For example, pin to a known-good commit (the repo previously used a pinned SHA):

uses: augmentcode/augment-agent@6f08f56902d6728a0ffbbb2004c136c06fec8104

If you still prefer auto-updates, consider at least a specific minor version tag (e.g., v0.1.x) and monitor with Dependabot/Actions review, acknowledging the trade-offs.

[mtpauly]

The example should probably be a model that's publicly available (sonnet4 or gpt5)