@mzwing/pi-codex-downgrade-detector

A Pi extension that tells you when a turn was served by a model other than the one you selected. It compares the model the server names — the openai-model response header, or the model a relay echoes back — with the one Pi asked for.

Install

pi install npm:@mzwing/pi-codex-downgrade-detector

Usage

The footer keeps one glyph for the latest turn:

| Glyph | Meaning | | ----- | --------------------------------------------------- | | · | No turn has finished yet | | ✓ | The server confirmed your model | | ↓ | Something cheaper answered | | ↑ | Something else answered, still not what you picked | | ⚠ | A different model, or the reasoning effort differed | | ? | Nothing named a model, so nothing is confirmed |

When a turn diverges, a row above the editor names both models, and the first substitution of each pair raises a notification.

/codex-downgrade lists the turns this session observed; /codex-downgrade show prints the resolved config.

Configuration

Optional. ~/.pi/agent/extensions/pi-codex-downgrade-detector/config.json and .pi/extensions/pi-codex-downgrade-detector/config.json are merged, the project over the global one.

| Field | Default | Description | | ------------- | ------------------ | ------------------------------------------------------------------- | | providers | ["openai-codex"] | Provider ids to watch. [] watches every provider. | | tiers | {} | Extra slug → rank entries, higher meaning more capable. | | checkEffort | true | Also compare the reasoning effort Pi sent with your selected level. | | notify | true | Notify the first time a pair is substituted. |

Credits

https://t.me/c/2502727045/42320

Super thx for opening the detecting verdict out!

License

MIT