Edge Conductor Release Notes
2.5.1 (Jul 29, 2025)
Bugs
- Scheduling: Resolved issues with the training/deployment schedule, including an icon that wasn't updating correctly and the removal of the meridiem (AM/PM) display.
- Workspace: Fixed the search function within the Member table in the Workspace menu.
- Inference: Resolved an issue that prevented users from viewing inference results, output, and logs.
2.5.0 (Jul 21, 2025)
Feature
- Enhanced User Experience
- A significantly improved web interface with a major UX/UI overhaul designed for increased usability and efficiency.
- Workspace Management
- Workspaces enable improved scalability and resource utilization by allowing multiple independent projects to run within a single Edge Conductor deployment. This feature provides logical isolation and prevents interference between projects.
Performance
- Database Optimization: Improved database processing speed and efficiency through the implementation of asynchronous methods
2.4.2 (Apr 23, 2025)
Etc
- Update of AWS Python Boto3 Library
Bugs
- Fixed issue with not processing files larger than 5GB in S3 Dataset
- UI
- When relabeling based on the inference results of the foundation model, it was incorrectly displayed as Unlabeled
- Fixed issue where datasets were not being retrieved and created based on the selected datetime during dataset creation for relabeling
- Modified the condition for model status when listing deployable model versions - models in 'training' and 'training_stopped' states will no longer be visible
2.4.1 (Mar 17, 2025)
Security
- LGERootCA certificate expiration date (2025-04-07) certificate replacement
Bugs
- Stream scheduling feature bug fix
- Jobscheduler remains active in the backend even after deactivation
- Even after deleting the stream, the jobscheduler remains active in the backend
- UI
- When adding a new label on the relabeling screen without specifying a color, an operational error occurs
- In the Dataset-Data tab (when the source is S3), improved the issue where the 'File Path & Name' column overflows when the S3 URI is long
- In the Dataset, after selecting a Dataset and clicking on the Statistics tab, the lower panel does not disappear even after deselecting the Dataset
- In the Inference results window, when the score is 0.0, it is displayed as "-", and it has been modified to output as 0.0
2.4.0 (Feb 12, 2025)
Features
- Modify the selection of the learning DataSet so that DataSets generated with different solution settings can also be used as training data
- Solution metadata (inference_result_datatype, train_datatype, support_labeling) values must be compatible with the solution data
- Provide a model build environment to compare Champion and Challenger models
- Enhance Edge Inference Monitoring
- Modify the Edge App inference input file name to be visible in the Edge Conductor (Input File Column in Edges Detail, Inf. Result Tab)
- Add functionality to view inference logs (Log Column in Edges Detail, Inf. Result Tab)
- Add an inference failure tab - allowing users to view the inference failure list and logs (Edges Detail, Inf. Failure Tab)
- Modify the Stream Inference Warning Score scenario: Disable the Warning Notification feature when does not input values during stream creation
- Modify the Remember Me feature (change functionality so that the remember me checkbox remains checked even after logging in by pressing the logout button; resolve the issue of being unable to log in when accessing from a new window)
- Remove the action that resets the ID saved with remember me when logging out by pressing the logout button
- Improve the use of session cookies to determine login validity
- Change the execution conditions for the Daily Usage Summary Service: Modify so that the service does not run when the host information of the Summary Server defined in the Config is 0.0.0.0
- Save AI Conductor Train Fail Reason in the Backend DB
- Modify the learning dataset URL (dataset_uri) when requesting training from AI Conductor
- UI
- Change the DataSet Detail Tab UI: For S3 DataType, show Data, Statistics; for Edges/Local types, show Data, Label, Statistics
- Add a popup during background tasks (Celeray Task) when data is not visible
- Provide a popup during the process of converting large inference data to a database
- Provide a popup during the creation of a DataSet with large data (Edge App Inference Data, Local Importing Data)
- Shorten the Stream deletion failure message (if the stream name exceeds 10 characters, show the first and last 5 characters, and shorten the failure reason)
- Change the initial sorting criteria column on the Dataset page from data source to last updated
- Adjust the column width for the Data tab and Label tab of the Dataset, apply text truncation for long text, and add tooltips
Performance & Internal & Stabilization
- Add LifeCycle functionality to the inference_data table in the database
- If the retention_period value is set in the config, delete inference data that has exceeded the set period
- Maintain the existing 730-day policy based on the Privacy Policy
- Remove file and data-related content from the previous deletion of expired inference tables
- Improve memory usage when storing large amounts of inference data in the database
- Enhance performance when creating datasets (add Celery Worker Task)
- UI - Provide popup messages for dataset creation tasks, a waiting UI for creation, and a UI for creation failure
- Improve notification processing speed
- Modify to prioritize reading and outputting the most recent data instead of processing the entire list simultaneously on the web
- Improve the Notification DB query in the backend
- Distribute the load of system services that run simultaneously at 0:00 UCT
- Add Backend Healthz API
- Fix connection closed issues between the backend and Redis
- Add the socket_keepalive=True attribute when creating the Redis Connection Pool
- Add reconnection code in the Task Manager when monitoring Redis, in case the connection is lost
Bugs
- Fix the issue where the web page does not dynamically change after a user with permissions logs in (issue with SSE not functioning)
- Fix the bug where the Edge Status value is incorrectly displayed when there are rapid changes in the Edge App connection status.
- Fix the error that occurs when attempting to stop model training through the Action on the Stream detail page, due to a URL composition issue when stopping training on the Stream Detail page
- UI
- Change the deletion confirmation message on each page (Edges, Dataset, Streams, Account) to a pluralized message 'item(s)'
- Apply CSS to display line breaks in the description area of the solution detail page shown during Stream creation when newline characters are included
- Fix the issue where the names of Edges, DataSet, and Streams cannot be changed in case sensitivity from the existing names
- Fix the issue where, in cases where only the pipeline log exists in the Inference Log, the log is incorrectly displayed in the process log tab, and the modal closes when clicking on the pipeline log tab
2.3.1 (Dec 17, 2024)
Features
- Modified to allow inference results to be uploaded from the Edge App even when inference data file is missing
- Add UI/UX so that users can check the download status of the Inference Docker in the Solution
- UI
- Added multi-selection range feature using the Shift button on the web
- Improve the multi-selection deletion scenario - apply bulk API, display results of multiple processing in a popup
- UI development for the case where the File Path does not exist in the Inference Result. display a notification indicating that image and table non-disclosure cases are not provided
- In the Streams menu, for the solution Cardview UI area - if the card's title or overview is long, truncate it and display a tooltip
- Request for additional improvements related to the Activate Schedule settings - display the operation time for the next three scheduled actions
- Reflect the length limit specification in the input box UI - display an explicit message when the limit is exceeded
Performance & Internal & Stabilization
- Improvement of response speed for inference data uploaded from the Edge App-modified to be processed through Celery(Worker)
- Improved speed performance through MySQL Query Tuning when users save after labeling datasets
- Enhancing Data Retrieval Performance through Lazy Loading
- Gallery View (Edge Inference Result Tab, Dataset Data Tab): Apply IntersectionObserver, replace existing loading image with spinner icon
- Table View (Dataset Data Tab)
- Added Edge App Debug Log: added logs for model-related REST API params & responses
Bugs
- Fixed issue where special characters were not functioning during name duplication checks when creating resources (Edges, DataSet, Stream)
- Fixed issue where Inference Docker Image remained when Solution was deleted
- Changed AI Conductor URL Schema check routine (fixed issue where Cluster IP URL http://{namespace}.{service} within EKS was not recognized)
- Fix the issue regarding dataset creation names
- Fixed the issue where another person cannot create with the same name during dataset generation
- Fixed the issue of being unable to regenerate a dataset with the same name after an unexpected termination
- Modify to integrate into a single API when creating a dataset-Reflect the redesign of the dataset registration scenario and logic
- Fix the issue where the label information set by the user is reset while relabeling data in the Dataset menu
- UI
- Fixed bug where clicking the training button twice consecutively caused duplicate calls
- Fixed issue where only part of the data was displayed in the Inference Gallery View of the Edges menu
- Fixed issue where Invalid Access occurred when entering the Label Data function from the 'Dataset Details' screen in the DataSet menu
- Fixed the UI alignment error in the Edge menu Detail Meta Info
- Modified the guide text when creating a dataset with Local File Upload in the Dataset menu
2.3.0 (Sep 27, 2024)
Features
- Add a feature to receive messages via webhook
- Support the ability to receive notification information using Slack webhooks and Teams workflow tool
- Add notification item grouping and selective message receiving settings
- Modify solution & stream UX
- Apply Card View format when displaying the solution list
- Add an option for users to select resources during training
- Improve Notification UX features
- Improve notification panel GUI
- Add Notification Setting feature: item grouping and allow users to configure notifications to receive only the ones they want
- Add a feature that allows the user to stop the model in training
- Support setup paging feature to allow users to dynamically configure settings after installation
- Add a note feature to the Edge App to help users easily identify the app
- Improve the UX for adding LDAP accounts
- Remove remote edge app reboot menu
- UI
- Change the data displayed on the Edge details page
- In the Edges table, the Edge Status and Inference Status columns are integrated into one column value as Status
Performance & Internal & Stabilization
- Support lazy loading feature for displaying tabular inference data
- Support debugpy mode in backend for tracing debugging for only develop & debugging
- Fix the intermittent issue where checked items disappear from the screen during labeling
- Fix the intermittent bug where data cannot be saved after deleting data in Edit Dataset
- Improve the issue where frequent SSE calls cause the screen to intermittently blackout or freeze
Bugs
- Fix the issue where the popup does not appear when an error occurs during model deletion
- Fix the issue where items do not appear in in-use edges under certain conditions
- Separate and manage Celery queues for each backend to fix the issue of request task entanglement when multiple conductors use a single Redis instance"
- Fix the issue where deleted Edge Apps are still visible in the stream schedule
- Fix the issue where 'Unlock' is displayed in account settings when the account is locked after 5 incorrect password attempts
- Fix the issue where the admin account cannot change the password of a regular user
2.2.1 (Jul 16, 2024)
Bugs
- Fixed an issue where Timezone could not be processed during the Stream Schedule function
- Fixed an issue where DB updates could not be performed occasionally due to a mysql connection error (MySQL server has gone away)
- Fix sql query statement where in used edge counting is calculated incorrectly
2.2.0 (Jun 28, 2024)
Features
- Apply stream automatic training & deployment function usging scheduling
- Apply edge app error report rest api
- Implementation of reject edge function during edge register
- Account page UI improvements : add lasted login datetime info, and lock status
- Notification
- apply notification lifecycle (7 days)
- add notification at training fails
- add detail message regarding inference failure
- remove notification at disconnecing edge app
- Real-time update data in web page (stream menu): Transfer data asynchronously from backend to frontend
- Change in admin account lockout cycle (90 days -> 30 days)
- Apply single pipeline model versing policy (when performing a single pipeline update stream, major version up and existing version is maintained)
- Do not delete models that are being trained
- UI
- Modified invalid ID/PW error message for security
- Display stream deletion failure pop-up - Add train and deploy button tooltips
- Basic sorting after adding created_at column to Edges
- Delete the title area of the main page
- When creating a dataset, duplicates exist in the AI Solution list.
- When creating a new stream, the Solution title of Cardview is displayed in a maximum of 2 lines.
- Modified to print edge name when printing notifications
- When registering Dataset, Solution Dropdown output is modified to 10.
- etc
Performance & Internal
- Asynchronous distributed processing Celery application development in the backend
- Improved Stream, Edge, Dataset, and Model deletion scenarios: Manage delete flag rather than delete from DB
- Add log to check inference docker pull, tag, push status
- Add edge connection/disconnection log
- Fixed issue where special characters could not be set when setting db password ($)
- Delete proxy private docker registry
- Modified to not issue login token every time ai conductor api is called, but to reissue it only when it expires
Bugs
- Fixed a traning error that occurred after update stream (fixed an issue where dataset_uri was not updated)
- Stream/edge/dataset name duplicate check logic modified: leading space in name is not allowed
- Fixed issue with socket connection from deleted Edge
- Fixed display issue of deleted Edge deployment status in Stream and Model Status.
- Improved deletion logic for failed train requests to AIC
- Fix removing dataset file fail
- UI
- “Do not show again” not working bug
- Improved bugs related to Table Selection when update occurs after SSE connection
- Dataset Edit - Note information not output bug logic improvement
- Drag and drop file upload - Add conditional statement to allow only zip and gzip extensions
- Resolving Inference Data ModalVersion output error
2.1.0 (Apr 24, 2024)
Features
- New implementation of alarm notification function
- real-time update data in web page (edge menu): Transfer data asynchronously from backend to frontend
- Implementing log file viewer function from train artifact
- Collecting inference failure data from edge-app
- UI
- display inference status more details (-, No Stream, Ready, Inferencing) edge-app
- display input/output path in edge details edge-app - display serial number of edge in details
- display inference time to detailed data album view
Bugs
- Fix failure with dataset trying training Image solution - write the path as a relative path in metadata (M)
- Fix failure with sample data - modified to list-up more than 1000 from s3 buckets (M)
- Stream Update
- resolved a pop-up not being displayed error when clicking on a previously selected instance
- Add logic to change dataset-url when updating instance (M)
- Modified the method of converting the schema passed from UI to dict when updating an instance (M)
Internal
- Support arm64 arch docker image in inference docker private registry (M)
- Add workspace info to redis key to follow up multiple edge conductors in iox
2.0.6 (Apr 9, 2024)
Features
- Apply account permission changes
- In the inference result, only data compatible with the solution_id of the current model is displayed.
- Apply data copy on s3 at training with sample data
2.0.4, 2.0.5 (Apr 5, 2024)
Bugs
- File upload sync issue when creating a local upload dataset of a large zip file
- Fixed the issue, Deploy Modal - Model Version Sorting
2.0.3 (Apr 4, 2024)
Bugs
- Improved Model status issue in Model item
2.0.1 (Apr 2, 2024)
Features
- Application of personal information impact assessment items
- User-Access-Event. Data-Access-Event log tracing
- Delete old data
- UI
- Add creator for Edges, Dataset, Stream
- Improved Dataset local data upload guide text
- Add solution information to inference data
Improved
- Improved db session performance
Bugs
- Training Duration issue improved
2.0.0-rc2 (Mar 15, 2024)
Bug fixed
- Fixed an issue where Magna could not use the private docker registry port within the edge conductor.
2.0.0-rc1 (Mar 14, 2024)
Internal
- Flatten alembic scripts
- terminate redis listener & reduce cpu usuage
2.0.0-beta3.8 (Mar 12, 2024)
Bug fixed
- Fixed an issue where inference Docker URL was stored incorrectly.
2.0.0-beta3.7 (Mar 12, 2024)
Features
- Apply ldap for account
- Apply account management policies
- Account locked after 90 days of non-login
- Locked if password is incorrect more than 5 times
- Automatic logout if you are away for 90 minutes
- Prevent simultaneous session connections (automatic logout when connected from another host)
- Apply train using dataset with aws s3 path
- Apply train using sample data defined in metadata of solution
- Apply usability collection (send to aiadvisor manager)
- UX/UI improvements
- apply UI to view deployment status for each model
- add model learning time display function
- add ai conductor link
Improved
- Apply solution instance docker to be downloaded in the background
Internal
- Structural improvement for horizontal scaling at EKS
- Add model_deployment and app_update information into edge/me api (request on edge app)
- Remove nginx version information exposure according to security policy
- Code static analysis - code x ray fixed