1.0.82 Regression: Unknown command: /model with BYOK

Describe the bug

As of 1.0.81/82 the /model command apparently does nto work anymore if the model is set via environment variables for BYOK (see https://docs.github.com/en/copilot/how-tos/copilot-cli/customize-copilot/use-byok-models)

On e.g. Azure AI Foundry multiple models can be hosted and I used the /model command to switch between models (like gpt-5.6-luna/terra/sol) based on requirements, this does not work anymore.

Affected version

1.0.81 and 1.0.82, it works with 1.0.80 on windows x64, verified this by manually downgrading the cli version.

Steps to reproduce the behavior

1. setup azure ai foundry, deploy a few gpt-5.6 models
2. run github copilot cli like this:

SET COPILOT_PROVIDER_BASE_URL=https://my-azure-ai-foundry.com/openai/v1
SET COPILOT_PROVIDER_API_KEY=key

SET COPILOT_PROVIDER_TYPE=azure
SET COPILOT_PROVIDER_WIRE_API=responses
SET COPILOT_MODEL=gpt-5.6-luna
copilot --reasoning-effort max

Expected behavior

I can select the model after starting copilot with BYOK settings.

Additional context

No response

首先，使用文件中所述的 COPILOT_PROVIDER_* 和 COPILOT_MODEL 環境變數重現該故障，然後比較 1.0.80 與 1.0.81/82 版本中的 /model 行為。確認使用 Azure AI Foundry BYOK 設定啟動 copilot 後可以選取模型，包括 responses wire API。