Build Observation Processor from Source#
Prerequisites#
A working Docker Engine and Docker Compose installation.
A clone of the SceneScape repository.
Generated service secrets before deployment. From the repository root, run
make init-secretsif they have not already been created.
Build the image#
From the repository root, build the Observation Processor image:
make observation_processor
This target builds the SceneScape common base image first, then builds the
Observation Processor runtime image. The service Makefile can also build the
image directly with make -C observation_processor when the common base image
is already available.
Run with Docker Compose#
Start the Scene Controller and Observation Processor profiles:
make deploy DEPLOY_PROFILES="controller observation-processor"
The service listens on port 50052 inside the Compose network with TLS, using
the network address observation-processor.scenescape.intel.com. make init-secrets generates the scenescape-observation-processor server
certificate mounted by the Compose profile. Deploying the profile starts the
service; the Scene Controller and Tracker service send detections to it only
when the experimental observation processor flag is set and the service host and
port are configured. See
Experimental Observation Processor.
For deployment credentials, environment variables, and a sample request, see the Observation Processor service README.
Tests#
The unit and loopback gRPC tests are in
tests/sscape_tests/observation_processor/. After setting up the repository test
environment, run them with the same PYTHONPATH the root unit suite uses for
the spatial performance libraries:
PYTHONPATH="spatial_performance_primitives/frustum_gaussian/python:spatial_performance_primitives${PYTHONPATH:+:$PYTHONPATH}" \
tests/.venv/bin/python -m pytest tests/sscape_tests/observation_processor -q
To run the full Scenescape unit suite (includes this package):
make run_sscape_unit_suite
Kubernetes#
A Helm chart for the Observation Processor has not been created yet. The service cannot currently be deployed with the Scenescape Kubernetes chart.