Pi Custom Model

Adapted from David Ondrej's MIT-licensed pi-custom-model at commit 11dee2ebc2d045806b686ba0b57746f1e3d7e331. Current installed Pi help and documentation override this pinned workflow.

This workflow edits user-level Pi configuration. Do not run it implicitly.

Preflight

Before writing anything:

1. Run pi --version, pi --help, and pi --list-models <search>.
2. Locate and read the installed Pi documentation for models and custom providers. Confirm the current models.json schema and reload behavior.
3. Confirm the requested provider/model ID against the provider's current authoritative catalog. Preserve spelling, punctuation, and variant suffixes exactly.
4. Determine whether the provider is built in, needs a models.json entry, or requires a custom provider extension for a non-standard API.
5. Check only whether authentication is available through Pi's supported login or environment mechanism. Never print or copy credential values from auth.json.
6. Check project settings as well as user settings when the problem occurs only in one repository.

If the installed registry already resolves the exact ID, do not add a duplicate custom model. Diagnose the selected provider, project override, or launch arguments instead.