Release Notes#
Intel® Edge Insights System 2.0#
Intel® Edge Insights System 2.0 release is a major release with revamped architecture, introduction of new unified GUI with keycloak enabled identity management for personas, easy and improved AI tasks deployment, integrated Intel® Geti Server for efficient MLOPs flows etc. The Intel® Edge Insights System package is available for download at ESH site. For getting started guide, please refer link
Note: As a prerequisite, please run commands:
$ sudo rm -rf /opt/intel/eii
(please backup the data if anything is required here) andsudo rm -rf /opt/esh/
if you had previous instance of Intel® Edge Insights System already installed before running./edgesoftware install
.
Changes#
Intel® Edge Insights System re-architecture:
ZMQ based EIS message bus replaced by gRPC protocol for communication between microservices
License based microservice activation is been disabled
ConfigManager Agent:
Generate the new gRPC TLS certs/keys for the microservices communication
Dynamic configuration support and pipeline deployment w/ baseline and pipeline services architecture
Edge Video Analytics Microservice:
Enabled annotation overlay for clients.
Optimized client publishers with multi-threading.
Added gRPC support in EVAM standalone and EIS deployments.
Introduced mTLS support for secure communication.
Added option to disable LEM check.
Made REST server port configurable via environment variable.
Enforced minimum TLS version and updated cipher suites.
Added support for RGB frame format in GST UDF plugin.
Enhanced XIRIS app to support additional configuration parameters.
Bug Fixes:
Fixed expired JWT replacement for Model Registry Microservice in EVAM.
Resolved issue where multiple pipeline instances didn’t refer to their own data.
Removed hardcoded visual overlay for classification models.
Corrected model loading from Model Registry during EVAM startup.
Enabled SSL certificate verification for HTTPS requests to Model Registry.
Infrastructure & Updates:
Optimized docker image size for better performance.
Updated 3rd party versions and DLStreamer dependencies.
Made EVAM filesystem read-only for improved security.
Added API for downloading models from Model Registry into EVAM.
Data Store:
Added support to push the data to Azure cloud
Performance improvement for data read/write, export operations.
Integrated latest Influxdb v1.8.10 and MinIO v2024-10
Enabled profiling mode to display throughput performance metrics
Data Collection as a service:
Integrated gRPC mode of communication with other microservices
Introduced “read-only” flag for improved security in docker compose file.
Disabled auto annotation and simple storage.
Bug Fixes & Improvements:
Made some CVAT containers run as non-root
Updated retry logic for license launch failures and added option to disable LEM check.
Model Registry
Moved JWT Secret Key to
.env
file and updated Digital Signature Algorithm for enhanced security.Enabled HTTPS mode by default and enforced stronger TLS version/cipher suites.
Added
PUT /models
API for model metadata updates.Bug Fixes & Improvements:
Fixed version format issue in
/models
endpoint query.Resolved issue with
mraas_minio
unable to create CAs directory in volume mount.Upgraded dependencies (FastAPI, Uvicorn, MLflow, psycopg2).
Time Series Analytics Services (Telegraf and Kapacitor):
Removed ZMQ references and support for Golang UDFs
Visualizer:
Redesigned and removed overlay logic for better latency performance improvements
Enabled profiling mode to display throughput performance metrics
Unified UI:
Intel Spark Island based new Unified UI designs and workflows
Keycloak based identity management for users and personas access
Support for new camera configurations
Integrated Intel® Geti Server for efficient MLOPs flows
Enabled additional features for operator persona
Enabled DCaaS configurations for remote backup
Combined Custom UDFs and Intel® Geti UDFs pages into single Models page
Easy and improved AI tasks deployment
Known Issues#
For known issues and troubleshooting steps, please refer to Troubleshooting guide
Intel® Edge Insights System 1.6#
This release is majorly a security fix release on top of Intel® Edge Insights System 1.5. We recommended users to move to this version as it has security vulnerabilities fixes.
Note: As a prerequisite, please run commands:
$ sudo rm -rf /opt/intel/eii
(please backup the data if anything is required here) andsudo rm -rf /opt/esh/
if you had previous instance of Intel® Edge Insights System already installed before running./edgesoftware install
.
Changes#
Fixed the security vulnerabilities found in the latest security scans and no functional code changes done.
The documentation would remain same as that of Intel® Edge Insights System 1.5 which is accessible at https://eiidocs.intel.com/1.5/. Please download latest 1.6 variants SKU from ESH site.
Intel® Edge Insights System 1.5#
This release is majorly a bugfix release on top of Intel® Edge Insights System 1.0. Intel® Edge Insights System - Standard and Intel® Edge Insights System - Enhanced 1.5 SKUs are available for download at ESH site.
Changes#
Bundled the pre-built helm charts in Intel® Edge Insights System - Enhanced package, enables single node helm deployment on native Kubernetes cluster.
Integrated latest software license library for all applicable microservices like Config Manager Agent, Edge Video Analytics Microservice etc.
Config Manager Agent
Helm template changes added to enable license check and removed master node dependency
Trivy scan fixes done
Edge Video Analytics Microservice
Object tracking for UDFs implemented
Bug fixes:
AUTO plugin issue fixed
Force creation of EII user during build time even though the same user ID already exists in the image
Object tracking ID fixed
Helm template changes added to enable license check
Added tags to published metadata
Enabled Balluff unrestricted license only in Edge Insights System mode. Restriction (watermark post 200 frames) applies in standalone mode.
DL Streamer 2024.0.2 upgrade
Xiris App updates with support to multiple configurable options.
Fix EVA MQTT Publisher bug.
Improved the way gencamsrc handles defaulting of property values based on user input
Data Store
Helm template changes added to enable license check
Trivy scan fixes done
Bugfix: Fixed the frame drops issue and localhost access issue
Intel® Edge Data Collection
Bugfix: img extension reference in annotation json set to actual encoding
Time Series Analytics Services (Telegraf and Kapacitor)
Helm template changes added to enable license check
Trivy scan fixes done
Moved the config files volume mount to docker compose file
Removed conda dependency due to Anaconda license unavailability
Removed messagebus support for the timeseries analytics pipeline due to latest telegraf and kapacitor version upgrades
MultiModal Data Visualization Microservice
Helm template changes added to enable license check
Trivy scan fixes done
Fixed the GUI hang issue by adding the fix in Visualizer when it is not receiving frames from Edge Video Analytics Microservice
Updated draw_results flag behavior to enable/disable the processing of frames and drawing defects in config.json
Model Registry
Added capability to access the service’s current version
Bugfix: Incorrect created and last modified time returned by GET /models endpoint
Web Vision GUI
Trivy scan fixes, docker bench security fixes and checkmarx fixes done
Improvements:
Update message shown in camera preview screen
Changing error message text when license manager is not working
Enable user friendly names for camera
Stop camera preview containers
Check for camera usage not more than 1 in project
Added user config file for adding the login credentials. Please refer readme for details
Favicon’s ratio icon is changed
RTSP camera changes
Project name is fixed to 20 characters on display
Changing the RTSP and Xiris field name change
Geni Camera Pipeline changes, removed the hardcoded parameters like throughput-limit etc.,
Timer popup from LinearProgress to CircularProgress
Enabled “Auto” Device mode for UDF configuration
Bug fixes:
Fit all 4 inference previews in a window not requiring to scroll up
Fixed the operator login issue
Fixed the GUI hang issue when one of the camera streams is not working as expected
Fixed camera preview screen issue
Removed support of Intel® In-Band Manageability Framework
Known Issues#
Memory leak seen with Edge Video Analytics Microservice on long runs due to anomalib UDF
Edge Video Analytics Microservice is not compliant with Intel® Geti™ UDFs created using Intel® Geti™ version earlier than 1.8
Intel® Edge Insights System 1.0#
Intel® Edge Insights System is the new branding name for Intel® Edge Insights for Industrial. Intel® Edge Insights System - Standard and Intel® Edge Insights System - Enhanced SKUs are available for download at ESH site.
Changes#
Intel® Edge Insights System packaging:
Support for different SKUs – Standard and Enhanced
License based microservice activation
Addition of new services/features
ConfigManager Agent:
Persisting EtcdUI/GUI changes across module restarts
Add REST API to facilitate config updates from GUI
SW license lib integrated
Edge Video Analytics Microservice:
Support for Anomalib, new version of Geti (v1.8)
DLStreamer/OpenVINO version upgrade
SDK upgrade (Balluff and Xiris)
Support for secured MQTT communication
SW license lib integrated
Data Store:
Added REST API interfaces
Bulk update and delete APIs support (ZMQ, HTTP, gRPC)
Local storage support on disk
SW license lib integrated
Data Collection as a service:
Annotating classification for available data using CVAT (Computer Vision Annotation Tool)
Remote data storage at fixed interval or on-demand
SW license lib integrated
Time Series Analytics Services (Telegraf and Kapacitor):
SW license lib integration w/ minor enhancements
Visualizer:
SW license lib integration w/ minor enhancements
Web Vision GUI:
Allow users to configure a select features of the modules
Enable camera preview for all supported camera types
Enable “File Ingestion” as input for project creation
Enable use of “Model Registry” for managing Geti models
Enable Watchdog feature
Following modules are removed from the Intel® Edge Insights System package:
OPCUA Data Export microservice
REST Data Export microservice
Known Issues#
Memory leak seen with Edge Video Analytics Microservice on long runs due to anomalib UDF
Edge Video Analytics Microservice is not compliant with Intel® Geti™ UDFs created using Intel® Geti™ version earlier than 1.8
EII 4.1.0#
Intel® Edge Insights for Industrial (EII) v4.1.0 release done on the microservice-based architecture foundation of EII v4.0.0 release by integrating upgraded standalone microservices. The release includes newer versions of Config Manager Agent(CMA), Data Store, Edge Video Analytics Microservice(EVAM) and Visualizer along with the security & other bug fixes. With Geti integration, users now can deploy models at the edge using EII UDFs (User defined function). This will allow for any Geti deployment to take advantage of EII’s AI pipeline reducing effort, increase efficiency, create a seamless experience for Edge AI analytics.
Changes#
General security issues fixed. Recommendation is to use EII 4.1.0 instead of EII 4.0.0 release.
Integrated latest versions of Config Manager Agent, Data Store, Edge Video Analytics Microservice and Visualizer microservices
Data Store 4.2.1:
Moving out the DBs as separate containers for greater flexibility
Having single message bus and config manager interface at the Data Store level
Bringing in some configuration changes to better support configuration based on the data type (JSON/BLOB)
Demonstrate how easily one can add the new DB handler for new DB instance. By default, will have InfluxDB and MinIO
Adding support for read, write, update, list and delete functions over EII message bus request-response socket
Enable gRPC interface support just like EII message Bus
Added DS sample app - GUI app to exercise the functions in DS microservice
ConfigMgr Agent 4.1.1:
Moved the Etcd UI microservice into Config Manager Agent microservice (No separate EtcdUI microservice now)
Edge Video Analytics Microservice 1.1.1:
Upgrade to ubuntu 22.04 base image
Upgrade to DL Streamer 2022.3.0 version
Added GETi deployment sample, works with GETi v1.5.x SDK/server
Added metadata converter for GETi udf
Added unit tests and scripts for unit test integration
Consolidation of udf and gvapython scripts
Single pipeline configuration in EII mode
Timeseries: Telegraf and Kapacitor config changes to adapt DS v4.2.1(Docker Compose and Helm)
Following modules are removed from the EII package:
Universal Wellpad Controller (UWC)
VideoIngestion(VI) and VideoAnalytics(VA)
Custom UDF applications dependent on VI and VA.
Tools dependent on VI and VA like SWTriggerUtility, GigEConfig and JupyterNotebook
Removed custom download option in the ESH download page (https://edgesoftware.intel.com/industrialinsights) and moved the only component
Manageability
under custom download option to recommended download option
Known Issues#
Web Deployment Tool:
PCB filter and classifier configs goes away when we delete and re-import in configure page
“A task in already in progress” issue appears when one logouts out and logs back in and tries to re-build in configure page. The workaround is to run
sudo docker-compose down
command from the[WORK_DIR]/edge_insights_industrial/Edge_Insights_for_Industrial_4.1.0/IEdgeInsights/build
directory in the terminal.
EII 4.0.0#
Starting from EII 4.0.0 release onwards, EII codebase and the pre-built
docker images would be released under the
Corporate Non Disclosure Agreement Edge Software Hub (ESH) Intel proprietary license
only via Edge Software Hub at
https://edgesoftware.intel.com/industrialinsights. EII 4.0.0 onwards,
the source code will not be released to public github. Also, the docker images will
not be published at https://hub.docker.com/u/openedgeinsights.
Features#
Enabled independent building and deployment of microservices
Simpler deployments or New EII services integration (
This additional option is being provided only in DEV mode
)In this option, the module microservice does not depend on the ConfigManagerAgent (CMA) microservice for configuration. All the microservice configs are directly read from the microservices respective config.json files. Below are the two APIs added in the ConfigMgr library to enable this additional option.
Config file APIs
Config file watch
Complex deployments
This is the existing use case driven approach of launching multiple microservices by using the use case yml file with the list of services. This deployment is supported both in DEV and PROD mode like earlier.
Enabled ETCD watch capability for video and timeseries services to auto-restart microservices when microservices config/interface changes are done via the EtcdUI interface
Timeseries pipeline improvement - provided volume mount option for loading the python udfs and other required configs in DEV mode for easy udf dev
Enabled Datastore Microservice, it adds support for running databases: InfluxDB* (vision and timeseries metadata) and MinIO* (image data)
Enhanced Edge Video Analytics Microservice (EVAM), is now the default Video analytics pipeline, which supports ingestion from diverse cameras (GenICam GigE and USB3 Cameras, RTSP Cameras and USB cameras), Gstreamer based UDF loader to run custom UDFs, image ingestion from storage & Geti SDK integration that enables usage of GETi generated deployment folder for model inference. All future EII new development for video capabilities will focus on EVAM instead of VI VA.
Distributing Universal Wellpad Controller (UWC) recipe as a use case with EII 4.0.0 package via ESH. Added the simpler deployments capabilities to independently build and deploy microservices just like EII services.
Added Helm charts improvements:
Enabled single helm charts (no separate provision and deployment helm charts)
Added steps around usage of NFS based persistent volumes (helpful while accessing PVs across nodes in cluster) – needed to run in Anthos cluster environment
Added Web Deployment Tool improvements
New UI design
Added support for Edge Video Analytics Microservice for vision use case
Added support for time series use case
The following fixes are added to improve security hardening:
Fixed security-related findings from the Bandit* and the hadolint* tools.
Upgraded the third-party software components to the latest versions as appropriate.
Hardened the Docker container image by removing the usage of the privilege flag for the applicable Universal Wellpad Controller services and making the rootfs read-only for the Jupyter Notebook and the Config Manager Agent EII services.
Known Issues#
Device provisioning not working as expected on ThingsBoard® portal with Intel In-Band Manageability
Noticed Azure manifests deployment doesn’t work in proxy network
NativePclIngestion pod fails to detect the Realsense camera in k8s cluster
Need to update the HOST IP to show video frames in Visualizer UI for video use case
NativeOneAPIIngestion container fails to work due to blur sample removal in the latest
intel-basekit 2023.1.0
Web Deployment Tool UI issues:
Remote deployment always happens in DEV mode even when the user selects the PROD mode
During drag and drop of components in “Build and Configure” screen, one can notice little bit of alignment issues of the UI components
UWC use case install not supported in the PRC region from the ESH page
Web Deployment Tool modules are not available for download in the “Custom Configuration” flow from the ESH page