Pi Web API-key setup does not support Qwen Token Plan custom endpoint, causing misleading 401

[zackie1671-create]

Summary

Qwen Token Plan works correctly in Pi CLI after configuring it as a custom provider,
but entering the Token Plan API key directly in Pi Web's Models/API Key UI results in:

401 Invalid API-key provided

The API key itself is valid.

Environment

• Windows 11
• Pi: 0.84.x
• Pi Web: latest
• Provider: Alibaba Cloud Qwen Token Plan Personal
• Model: qwen3.8-max

What I initially did

In Pi Web, I entered my Qwen Token Plan API key (sk-sp-...) through the API-key UI.

Pi Web returned:

401 Invalid API-key provided

Root cause / workaround

Qwen Token Plan requires its dedicated OpenAI-compatible endpoint:

https://token-plan.cn-beijing.maas.aliyuncs.com/compatible-mode/v1

It cannot be used as a normal Qwen/DashScope API key.

I added the following custom provider to:

~/.pi/agent/models.json

{
"providers": {
"qwen-token-plan": {
"baseUrl": "https://token-plan.cn-beijing.maas.aliyuncs.com/compatible-mode/v1",
"api": "openai-completions",
"apiKey": "QWEN_TOKEN_PLAN_API_KEY",
"models": [
{
"id": "qwen3.8-max",
"name": "Qwen 3.8 Max - Token Plan",
"input": ["text", "image"]
}
]
}
}
}

After setting the environment variable and launching Pi/Pi Web with this configuration,
Qwen 3.8 Max works correctly in both CLI and Pi Web.

Expected behavior

It would be helpful if Pi Web could support custom API-key providers directly from the Models UI,
including:

• custom Base URL
• API compatibility type
• model ID
• environment-variable-backed API key

Alternatively, if a key such as sk-sp-... is entered for Qwen Token Plan,
the UI could explain that Token Plan requires a different endpoint instead of showing a generic
"Invalid API key" error.

Why this matters

The current UI makes it appear that the API key itself is invalid,
while the actual issue is the provider endpoint/configuration.

This creates a confusing difference between Pi CLI and Pi Web.