Hello Hiwonder Support Team,
ACHINE|MIC_TYPE: TonyPiPro|WonderEchoPro
ASR_LANGUAGE: English
VERSION: V2.0|2025-11-24
The issue is unchanged. While running the AI Large Model course, the RealTimeOpenAIASR class in speech.so connects to the microphone, but the WebSocket connection to the OpenAI Realtime API closes immediately with this error:
Disconnected from server. 4000 invalid_request_error.beta_api_shape_disabled
I have verified that:
- My OpenAI API key is valid and works with the official OpenAI REST API.
- The microphone is functioning properly.
- Internet connectivity is normal.
It appears that OpenAI has retired the beta Realtime API format used by the bundled speech.so. Since speech.so is a compiled binary, I cannot update it myself.
Could you please let me know:
1. Is there an updated speech.so (or a newer system image than V2.0 2025-11-24) that supports the current OpenAI Realtime API?
2. If so, where can I download it, and how should I install it?
3. If not, is there updated source code or an example project that replaces RealTimeOpenAIASR?
I would like to keep using the original Hiwonder framework without modifying the rest of the project.
Thank you again for your help.