Even though we followed the design of the Korvo board for the speaker and microphone, getting them to work is another thing. Though the MicroPython build that includes esp-idf and esp-adf works (see https://github.com/espressif/esp-adf/tree/master/micropython_adf if you want to know how to do it), importing the audio library does nothing: cannot record audio from microphone to SD card, and cannot play music from SD card to speaker.

We hit a turning point in solving this problem when we decided to just skip the SD card, and just play music to the speaker via a socket stream. It worked.

Moving to getting the microphones to work is particularly harder than we thought. Everytime we turned on the microphone object and started recording to the SD card, we got core panicked error from ESP32. Digging deeper into the problem shows that it's not the speaker or the microphone that are causing the problem, it's the micro SD card all along!

The micro SD card interface used in Korvo is SDIO, whereas we have been trying to use the SD card with SPI. Switching to SDIO solved the problem.