[Bug] Regression: Groq API requests routed to OpenAI (401 Unauthorized)

[konprtk]

Describe the bug

The AI proofreading feature suddenly returns a 401 Unauthorized error when using a valid Groq API Key.

Despite having worked perfectly for over 15 days, the error message now explicitly points to https://platform.openai.com, indicating that the keyboard is incorrectly routing requests to OpenAI's endpoint instead of Groq's.

To Reproduce

1. Open HeliboardL Settings > AI Settings.
2. Input a valid Groq API Key (starting with gsk_).
3. Select a Groq-supported model (e.g., llama-3.3-70b-versatile).
4. Attempt to use the Proofreading feature on any text field.
5. Observed Result: A popup appears with API error (401) and a message directing the user to OpenAI's dashboard.

Expected behavior

The app should route requests to the Groq endpoint (https://api.groq.com/openai/v1) when a Groq key/model is selected. It should not default to OpenAI's servers.

Technical Analysis & Validation

I have verified that the API key is fully functional and the Groq service is active by manually executing the following curl command from the same device:

curl -X POST "[https://api.groq.com/openai/v1/chat/completions](https://api.groq.com/openai/v1/chat/completions)" \
     -H "Authorization: Bearer [REDACTED_GROQ_KEY]" \
     -H "Content-Type: application/json" \
     -d '{
       "model": "llama-3.3-70b-versatile",
       "messages": [{"role": "user", "content": "hi"}]
     }'

Result: The command returns a successful 200 OK with a valid JSON response from Groq. This confirms that the issue is a routing regression within the app and not a provider-side failure.

App version:

• v3.7.2 (standard)

Device Information

• Model: Redmi Note 13 Pro 5G
• OS: Xiaomi HyperOS 2.0.207.0 (Android 15)

[LeanBitLab]

I have fixed the issue where Groq AI requests were being incorrectly routed to OpenAI if a custom endpoint had been previously saved. Expect the next release with fix after a week, i am waiting for the fdroid merge

Temporary Workaround Instructions:

1. Switch AI Provider: In AI Settings, set the provider to HF/OpenAI-compatible.
2. Manually Set Endpoint: Tap API Endpoint and paste the official Groq URL:
   https://api.groq.com/openai/v1/chat/completions
3. Enter Groq Key: Paste the Groq API key (gsk_...) into the API Token field.
4. Set Model: Manually enter the Groq model name (e.g., llama-3.3-70b-versatile).

option 2: you can just clear data and use groq only

to reproduce the bug:

1. Select OpenAI Provider: In AI Settings, switch the provider to HF/OpenAI-compatible.
2. Save Endpoint: Open the API Endpoint setting and press OK. This saves the OpenAI URL (https://api.openai.com/v1/chat/completions) into the app's persistent preferences.
3. Switch to Groq: Switch the provider back to Groq.
4. Trigger Bug: Attempt to use proofreading