Flowstate Service

The Pose Estimators developed can be run as a service on Intrinsic Flowstate solutions to build applications with robotic systems.

First, build another Docker image that wraps the pose estimator Docker image generated previously with some environment variables to communicate with the Intrinsic Platform.

# Replace POSE_ESTIMATOR_DOCKER_TAG with tag of the pose estimator image generated previously.
# Replace "bpc_pose_estimator_fs:team_name" with a different tag if needed.
cd ~/bpc_ws/bpc/flowstate
docker buildx build -t bpc_pose_estimator_fs:team_name \
  --file Dockerfile.service \
  --build-arg BASE_IMAGE=<POSE_ESTIMATOR_DOCKER_TAG> \
  .

Next, save this Docker image locally. Note, the name of the output file must remain bpc_pose_estimator.tar.

docker save bpc_pose_estimator_fs:example -o bpc_pose_estimator.tar

Then build a Flowstate Service bundle.

cd ~/
git clone https://github.com/intrinsic-ai/sdk.git
cd sdk
bazel run //intrinsic/tools/inbuild -- service bundle \
  --manifest ~/bpc_ws/src/bpc/services/flowstate/bpc_pose_estimator.manifest.textproto \
  --oci_image ~/bpc_ws/src/bpc/services/flowstate/bpc_pose_estimator.tar \
  --output ~/bpc_ws/src/bpc/services/flowstate/bpc_pose_estimator.bundle.tar

Finally install the Service in a running solution.

bazel run //intrinsic/tools/inctl -- service install ~/bpc_ws/src/bpc/services/flowstate/bpc_pose_estimator.bundle.tar \
  --cluster CLUSTER_ID \
  --org ORG \
  --registry REGISTRY \
  --skip_direct_upload

To uninstall the Service,

bazel run //intrinsic/tools/inctl -- service uninstall ai.intrinsic.bpc_pose_estimator \
  --solution SOLUTION_ID \
  --org ORG