This is a small example of Edge Impulse + Redpanda Connect working together for an easy to use Edge AI streaming pipeline.
This is the output of myself going through the sound classification tutorial and outputting the resulting project as a C++ library. The C++ library is unpacked into the my_edge_impulse_project/ directory. I have then built a small integration point using Redpanda's C++ SDK (found here) using the edge impulse library. You can see the code and build system at src/transform.cc and CMakeLists.txt respectively.
The Data Transform takes in a JSON array representing the raw feature data for the classifier, and outputs the score for each label.
The following command will compile a wasm32-wasi artifact at build/app. The artifact can be deployed to Redpanda or Redpanda Connect.
docker run -v `pwd`:/src -w /src ghcr.io/webassembly/wasi-sdk /bin/bash -c 'cmake -Bbuild && cmake --build build'
The simplest way to run this is using the Redpanda Connect configuration file that can be found at connect.yaml (make sure you've built the SDK first!).
The follow command sends some example data (redpanda labelled and noise labelled respectively) into connect to be classified and reported on stdout.
cat example_data.txt | docker run --rm -i -v $(pwd)/connect.yaml:/connect.yaml -v $(pwd)/build/app:/app.wasm docker.redpanda.com/redpandadata/connect run