SDK for TESAIoT Dev Kit
API reference & tutorials (ModusToolbox)
Loading...
Searching...
No Matches
Header facts that are easy to get wrong
  • The shipped copies of ai_engine.h (proj_cm55/modules/ai_models/ and lib/edge_ai/include/) are byte-identical.
  • ai_engine_register(), ai_engine_dyn_count() and ai_engine_dyn_capacity() are in api.txt and callable — run-time registration is documented in ai_engine.h, with an example in proj_cm55/examples/edge_ai/05_register_model.c.
  • ai_result_t has no valid, label or confidence field. The real reduction is the inline ai_result_top_positive() in ai_engine.h, which returns the strongest positive class (index 0 is the negative class by registry contract) as 0..100; the example in proj_cm55/modules/ai_models/README.md uses it.
  • ai_engine_stack_words() returning 0 means the inference task was never created; every reading on every page below is vacuous until it is non-zero.