kevin-presalytics / cvops-hub Goto Github PK

Prerequisitives:

• Create minio bucket with workspace

Upload model process:

• Create record in models collection for the the workspace
• Create presigned Url for models/{id} = [ ] Upload via python client? "upload_model" method

Deploy

• deplooy via python client "upload_and_deploy_to_device"
• Run model checks
• create download presigned url
• Send message over mqtt to to device with presigned url
• Device to download model
• Device to start inference
• Device to delete old model

Command to start minio for testing:

docker run -p 9000:9000 -d -p 9001:9001 -e "MINIO_ROOT_USER=hub" -e "MINIO_ROOT_PASSWORD=Password@1" --name minio quay.io/minio/minio server /data --console-address ":9001"

Guidelines for Events:

1. Any device can create an event, so long at the hub can process the event (write to db).

Deployment Sequence:

1. Device creates a deployment on the workspace/{workspaceId}/deployments channel with DeploymentCreateRequest.
2. Hub validates the deployment request. Do the requested devices support the supplied model type? Do the requested devices exist inthe workspce?
3. Hub Response to the with the deployment created platform event on the events/workspace/{workspaceId}channel. For an invalid request, use the DeploymentFailed Status.
4. Where device id matches deploymentInitiatorId, device sends an and UploadUrlRequest message workspace/{workspaceId}/deployments
5. Hub Reponds to the Repsonse topic, typically device/{deviceId}/commands with and UploadUrlRepsonse containing a signed Url for storage
6. device emits a DeploymentUpdated platform event with status ModelUploading.
7. Device uploads to to Url. When upload is complete, device emits a DeploymentUpdated platform event with status ModelReady.
8. Target devices request a model download Url.
9. Target device emits a DeploymentUpdated with ModelDownloading status
10. Target deice emits a DeploymentUpdated with Active status