Library integration and API

This message will disappear after all relevant tasks have been resolved.

Semantic MediaWiki

There are 1 incomplete or pending task to finish installation of Semantic MediaWiki. An administrator or user with sufficient rights can complete it. This should be done before adding new data to avoid inconsistencies.

STM32Cube AI Studio generates a set of C source/header files and a static runtime library for seamless integration.

1. File structure

Generated artifacts are placed under the AI sub-folder of your exported project and typically include:

App/AI: Network sources and headers, application code, glue code
Middlewares/: AI runtime library integration

Middleware and application content may vary based on toolchain, project variant whether application or validation, and project settings.

2. Integration teps

Include headers: Include stai.h and the generated network header(s).
Runtime initialization: Calls stai_runtime_init().
Network context: Declares the context with STAI_NETWORK_CONTEXT_DECLARE(...) and initialize with stai_network_init().
Activation buffers: Allocates and binds with stai_network_set_activations().
Inputs/outputs: Binds buffers using stai_network_set_inputs() and stai_network_set_outputs().
Inspect/run: Optionally calls stai_network_get_info(), then run with stai_network_run().
Deinit: Calls stai_network_deinit() and stai_runtime_deinit().

3. Multi-network support

Use the ST‑AI (stai) embedded client API multi-network support for projects with multiple models.

4. Code and data placement

.text/.rodata: Place in internal flash.
.data/.bss: Place in SRAM.
External memory: Use for large models automatically or as configured.

5. Debugging

Black box: No internal state introspection.
Error API: Use the ST‑AI (stai) error reporting for integration issues.

6. Related ST Edge AI Core documentation

The embedded inference API is documented in ST Edge AI Core: