Detect wake word model from index.json

[78]

This pull request introduces several improvements and refactoring to the build system, audio service initialization, and board configurations, with a focus on simplifying wake word selection, improving maintainability, and updating default configurations. The most significant changes are grouped below:

Build System Improvements and Refactoring

• Added a reusable CMake function find_component_by_pattern to dynamically locate components by pattern, replacing repeated code blocks for finding components such as esp_emote_gfx, espressif__esp-sr, and xiaozhi-fonts in main/CMakeLists.txt. This reduces redundancy and improves maintainability. [1] [2] [3]
• Updated the project version from 2.0.1 to 2.0.2 in CMakeLists.txt.

Wake Word and Audio Service Refactoring

• Refactored wake word selection logic in audio_service.cc to choose the appropriate wake word engine at runtime based on the platform and available models, instead of compile-time macros. This enables more flexible and correct wake word detection on different hardware. [1] [2] [3] [4] [5] [6]
• Updated the logic for appending wake word source files in main/CMakeLists.txt to match the new runtime selection approach.

Board Configuration and Documentation Updates

• Removed explicit flash size and partition table settings from multiple board config.json files and corresponding documentation, relying on improved defaults and reducing user configuration burden. [1] [2] [3] [4] [5] [6] [7] [8] [9] [10]
• Changed the default partition table for ESP32 builds to partitions/v2/4m.csv in sdkconfig.defaults.esp32.

Font and Board Defaults

• Updated default font assignments for several boards to use font_puhui_basic_14_1 instead of font_puhui_14_1 for improved consistency. [1] [2]

Other Minor Improvements

• Increased the delay before displaying board information in WiFi config mode from 1 second to 1.5 seconds for better user experience.
• Simplified power save timer initialization in several board classes by removing conditional logic based on wake word configuration. [1] [2] [3]

[tyoh-love]

@78 thank you so much for your good work!!!

It is not clear, the step to enable custom wake word feature. could you give us example and guide doc to make it working?

[78]

@78 thank you so much for your good work!!!

It is not clear, the step to enable custom wake word feature. could you give us example and guide doc to make it working?

Try these steps in menuconfig:

0. Select "Flash Default Assets"
1. Disable "Wake Word Detection (AFE)"
2. Enable "Custom Wake Word Detection"
3. Fill in these forms:

4. Deselect all default wake words in "Load Multiple Wake Words"
5. Select a "Chinese Speech Commands Model" or "English Speech Commands Model", typically "mn7_cn" or "mn7_en"

Build and Flash.

Check the serial logs if the multinet model has been initialized successfully.