Getting started¶

Choose an independent ESP-IDF project first. Components present in a directory, functions defined in a source file, and functions actually called at startup are different things.

1. Choose a project¶

The entry is main/AIoTNode.cpp in each project. See projects and versions for component and copy differences.

2. Build and run¶

The three sdkconfig files, the AI dependency lock and CI use ESP-IDF 6.0. In a terminal with that toolchain activated:

cd CODE/AIoTNode-TinyAuton-DSP
idf.py build
idf.py -p YOUR_SERIAL_PORT flash monitor

Replace YOUR_SERIAL_PORT with your device port. Existing configurations target ESP32-S3; use idf.py set-target when changing targets and save your custom configuration first. Review storage, PSRAM and component settings in idf.py menuconfig. The project CMake files, component directories and dependency locks define the build.

3. Select a verification run¶

• DSP: set the entry's TEST_TINY_DSP_* selectors. The default transform group calls tiny_fft_test(), tiny_dwt_test_all() and tiny_ica_test_all().
• Math: select calls inside middleware/tiny_math/mat/tiny_matrix_test.cpp. Most A–F calls are commented out; historical stage outputs do not imply that the current startup runs those stages.
• AI: set TEST_TINY_AI_MLP/CNN/ATTENTION. Training can occupy the CPU for extended periods. This example entry calls esp_task_wdt_deinit().

FFT tests and results explain inputs, criteria and output; the matrix test overview indexes historical runs by stage.

4. Record reproduction conditions¶

Record the project, Git commit, hardware, ESP-IDF version, configuration, input, selectors, acceptance criteria and complete serial output. Missing historical dates or devices stay marked as unrecorded. Source dates are not execution dates, and a documentation build is not a hardware test.