Changelog
Navigator Release Notes
-
Release Date: 1/29/2026
Environment: Production
New Features
Canvas Templates and Projects Update
Technical Details:
We've updated the UI for creating, searching, and navigating featured templates within Navigator.
User Impact:
- Users will now see additional information associated with each featured template that describes the flow in more detail. They will also see the option for creating an individual canvas or adding the template to an existing project folder. Users will also now have the option to create a project from a template. Currently there is only one featured project available - KG RAG Flow - but we plan to add more featured projects in the future!
Improvements
Report a Bug From Toolbar
User Impact:
- Inside of Navigator we've improved the experience for reporting a bug. Accessing the top toolbar - Help - Report a Bug will now display a "Show what will be sent" option that users can click to see exactly what is being shared with the webAI team when reporting a bug.
Deployment Details Display Improvements
User Impact:
- Users will now see a "Connection Details" button when clicking into the deployment details within Navigator. This pop up modal will show the Port for which a deployment is running as well as the API Element Key. You can now easily copy these details to integrate with third-party platforms. Each element will have its own status indicator. Clicking each element name will now open a side panel displaying that element's details providing more insight into each element within a deployment. The Cluster for which a deployment is running on is now hyperlinked providing a smoother user experience when navigating the deployment and cluster details page.
Bug Fixes
Various Bug Fixes for Preview Settings Panel
Resolution: There were several bugs that impacted adjusting element settings in the Preview tab.
User Impact:
- Users should now have a better experience when updating and searching the element settings panel within the Preview tab.
-
Release Date: 1/12/2026
Environment: Production
New Features
Reasoning Toggle
Technical Details:
A reasoning toggle is now displayed in the Preview tab when interacting with a model. If a model with reasoning capabilities is running in Navigator and the feature is enabled, the reasoning steps are parsed and highlighted to the user in the interface.
User Impact:
- Users will now see a reasoning toggle within the Preview tab on a Canvas containing the LLM and API elements. Once reasoning has finished, the total duration is shown as the title and the user can scroll the dialogue to review the reasoning steps that took place.
Device Metrics Tooltips
Technical Details:
Within a Deployment Details window - a user can now hover over the devices connected to each Element and get a tooltip that shows that device's metrics.
User Impact:
- This will help users understand each device's usage quicker while monitoring a deployment.
Improvements
Worker vs Controller Mode
Improvements made to Worker and Controller modes.
Updated Available Models List
Added and removed several models from models list.
User Impact:
- Users should see an updated list of models with newly added models and the removal of outdated models.
Offline Elements Shown In Element Drawer
We've improved the behavior in Offline Mode to only show downloaded elements making them available in the Element Drawer.
Increased Speeds On Project Deletion
User Impact:
- Users should now experience a more seamless interaction with Navigator when a Project folder is deleted, updating the UI to display the correct list of available Canvases and Projects.
Toast Message for Multiple Canvases Running
We will now present a toast message notifying users that they currently have a Canvas running when multiple Canvases are prompted to run.
User Impact:
- This will assist users in understanding which Canvases they have currently running and if they would like to proceed in running another Canvas simultaneously.
Resizable Log Viewer
The log viewer is now resizable similar to devtools.
User Impact:
- Users now have the ability to resize the Log Viewer by clicking and dragging the border of that viewer similar to devtools.
Reset Logs After Hard Reset
If the Log Viewer was open before a hard reset of Navigator we will now close and reset the logs to avoid any unwanted errors.
User Impact:
- Users will now see the logs reset after a hard reset of Navigator.
Bug Fixes
Remove Chat Preview Floating Modal
Resolution: There was an issue with the deprecated floating chat preview modal appearing from the LLM Chatbot template. This will open the Preview tab as originally intended.
User Impact:
- Users will no longer see the ability to open the floating chat preview modal. All chat interactions with a running canvas in Navigator will take place in the Preview tab.
Scrolling Broken on Multi-Modal Chat Prompts w/ Images
Resolution: Fixed an issue with multiple images being added to a chat in the Preview tab which was preventing scrolling.
User Impact:
- Users can now use multi-modal models as expected, adding in multiple images in the same conversation without any scrolling issues.
Multi-Modal Filtering Parameters On LLM Element
Resolution: We've updated the filtering parameters on the select model settings within the LLM element. The multi-modal filter should now use OR rather than AND in conjunction with other filters.
User Impact:
- Users can now narrow down their search more clearly using the multi-modal filter in conjunction with other filtering parameters.
Update Element Settings in Preview Tab Not Updating In Canvas Tab
Resolution: When updates are made to an Element's settings in the Preview tab, those updates should now be reflected in the Canvas Tab.
User Impact:
- Users should now see a more seamless experience when modifying an Element's settings in either the Canvas tab or the Preview tab.
Closing Organization Edit Window Name Retention
Resolution: When editing the name of an Organization, if you type in a new name and close the modal the previously entered name no longer is retained the next time you edit.
User Impact:
- Users should no longer see previously entered names while editing an Organization without saving.
-
Release Date: 12/17/2025
Environment: Production
New Features
Export LLM Feedback
Technical Details:
The ability to export LLM Feedback is now available! For each Application that has LLM Feedback provided there is now an export button.
Export Formats:
- CSV
- JSONL
User Impact:
- This allows users to get additional insights on user feedback they are receiving from generated responses within Companion. LLM Feedback exports provide users with the raw data for them to analyze and make insightful discoveries on how their custom LLMs are performing.
Offline Mode
Technical Details:
Users can now access projects, canvases, and previously downloaded elements without internet connection. There are toast messages inside of Navigator that will indicate that the application is offline, however users can still run and build flows with previously downloaded elements.
User Impact:
- This allows for work to continue inside of Navigator in areas with no or low connectivity with Elements that are locally hosted. See our Offline Mode article for more information!
KG RAG Text-Only Document Ingestion
Technical Details:
There is now a toggle inside of the Document Ingestion element to enable text-only ingestion from your provided PDFs. This will ignore images, graphs, and diagrams during the ingestion process and only extract text.
User Impact:
- Users will now have more flexibility when building out their knowledge graphs for inference allowing them to provide extensive documentation during ingestion but build a knowledge base solely on the text within that documentation.
Submit Feedback
Technical Details:
Under the Help Menu in Navigator there is now a Report a Bug… option to send any issues to our team including the Navigator logs.
User Impact:
- This provides an easy way to report bugs within Navigator along with any coinciding logs that may have accumulated while building flows, running deployments, or general user behavior.
Improvements
Added Article(s)
Accessing Deployments from Other Devices - this article helps step through how to view deployments created by other members in your organization.
KG RAG Toast Messages - Stopped Flow
User Impact:
- If a KG RAG ingestion flow stops but a toast message is present to create inference a user can still click and proceed on those toast message buttons.
Load Speed Improvements for Canvases
User Impact:
- Users should now see fast loading speeds when clicking between tabs, specifically between the Projects tab and other tabs. Canvas and Project lists loading without delay.
Controller Mode Cleanup On Worker Disconnect
User Impact:
- Users should now see a quicker, more accurate display of connected workers inside of their Controller instance within Navigator when a Worker is disconnected.
Current Running Canvases Warning
User Impact:
- As a user, if I have a Canvas running and I attempt to run another Canvas, we will now show a toast message that asks if you would like to stop the currently running Canvas.
Disable Node Removal On Remote Clusters
User Impact:
- A user can now only remove nodes from clusters that they own.
Version Selection After Canvas Selection Change
User Impact:
- We now reset the version selection if a user changes the Canvas selection during deployment creation.
Multiple Canvas Capacity
User Impact:
- Users can now run multiple canvases at once based on device capacity and element selection specifications.
Bug Fixes
Worker Node Not Connecting To Controller - Mode Switching
Resolution: Mechanisms put in place to ensure a full reboot and mode switching would connect successfully for each machine.
User Impact:
- Users should no longer see issues with switching modes in Navigator for Controller vs Worker and should be able to connect successfully when switching between the two.
Update Learn More Links - Worker and Controller Mode
Resolution: Updated the hyperlinks inside of Navigator on the selection screen for Worker and Controller mode.
Empty Element Settings Panel On Element Version Change
Resolution: Changing element versions should no longer provide an empty Element Settings panel.
User Impact:
- Users should be able to switch between element versions seamlessly without losing access to Element Settings.
Shimmer Icon - Model Selection
Resolution: Reimplemented the shimmer icon to show in model selection for models that are compatible with webFrame.
Various Preview Issues - Multiple Flows Running
Resolution: Fixed issues with the chat preview window when multiple flows were running at the same time.
User Impact:
- Users should not see issues with chatting between multiple flows in Navigator when switching between running canvases.
Wrapped Text "In Progress" Button
Resolution: In smaller windows we are no longer wrapping the text of the "In Progress" button.
Multi-Modal Upload Image Inconsistency
Resolution: Corrected functionality inside of chat preview within Navigator for multi-modal supported LLMs.
User Impact:
- Users should not see issues with uploading images within a chat preview when allowed by a supported LLM.
Canvas Already Running - Toast Message Incorrect
Resolution: Changed logic for when this toast message should appear as it was not displaying in the correct scenarios.
Blank Screen On Collection Naming Step
Resolution: Updated logic for clicking between collection naming step and element settings.
User Impact:
- Users should no longer get into a blank screen state in Navigator when clicking between Collection naming step and Element Settings.
Back Button Screen Landing
Resolution: Updated the landing pages when clicking the back button. Will now land on previous page.
User Impact:
- This decreases confusion while clicking through Navigator and returning to the previous page you were on.
Deselect Worker While In Preview Mode
Resolution: While in Preview Mode logic was updated to the Element Settings to deselect offline device workers.
User Impact:
- Users can now utilize the device selection seamlessly inside of the Preview tab and Element Settings window.
Profile Worker Message Still Shows When Worker Offline
Resolution: Offline workers should no longer show on a profile when that worker is offline.
Back Button On Collection Window
Resolution: Cleaned up the user interaction with the back button on the collection window within a KG RAG flow.
Scroll Functionality - Preview Element Settings Panel
Resolution: Fixed issues that prevented users from scrolling in the Preview window to adjust element settings.
Order of Logs Not Chronological
Resolution: Fixed issue where logs were not showing in chronological order when viewing in the Log Viewer.
-
Release Date: 11/24/2025
Environment: Production
New Features
KG RAG Document Ingestion and KG RAG Inference
Technical Details:
There are now two new templates to help a user get started using Knowledge Graph RAG flows. New articles have been created for each of these templates:
Three new elements have been added to the element drawer for these templates:
- Document Ingestion
- Knowledge Graph Builder
- Knowledge Graph Retrieval
User Impact:
- Allows users to build an AI-ready document repository by converting PDFs into vector data and creating a knowledge graph. This document ingestion pipeline is the first step in setting up a KG RAG system, using webAI's Navigator to process documents and prepare them for semantic retrieval.
Applications Tab
Technical Details:
Introducing Applications! You can now assign a deployment to an Application so that it can be discoverable by Companion. For more information see our Applications Overview.
User Impact:
- Creating Applications allows for custom trained LLM's to be accessible within the Companion app. Users can now interact with the LLM deployments they built through Companion by creating a Deployment and assigning that Deployment to an Application.
LLM Feedback (Beta)
Technical Details:
Within an Application Details page you will see a tab titled "LLM Feedback". This feature is currently in beta. This allows for users to give feedback on the responses an LLM has provided through within the Companion application.
User Impacts:
- As users begin to interact with their remotely deployed LLM's through an Application they can provide a thumbs up, a thumbs down, or flag the responses within Companion. The feedback users provide on those responses are listed within this tab allowing developers to see what feedback users have on the provided responses.
Improvements
Improved Chat Preview Experience
There were some issues with spacing in the chat preview within Navigator between element settings on the right panel. These spacing and various have been improved.
Progress Bars
There are now progress bars for the KG RAG Builder and Document Ingestion elements.
User Impact:
- This will help avoid confusion of when a flow has completed the ingestion process.
Log Viewer Automatically Opens - KG RAG Ingestion
During the ingestion process for KG RAG - the Log Viewer will now open automatically.
User Impact:
- This will give the user a quick insight and indication (along with the progress bars) that the ingestion flow is running.
Collection Selection For Knowledge Graph Elements
On the Knowledge Graph elements there are now dropdowns and selection options for choosing a collection.
User Impact:
- This will help avoid a lot of confusion around which collection is being input and output into knowledge graph elements. This creates a much smoother experience for building a Knowledge Graph artifact and implementing it into a retrieval flow.
Toast Message On KG RAG Ingestion Completion
Once a KG RAG ingestion process has completed there is now a toast message to automatically create an inference flow as well as add it to a project or create it as a single canvas.
User Impact:
- This assists a user's journey to quickly get started interacting with their Knowledge Graph output and begin to test out the results of their ingestion.
Controllers Now Show Connected Workers
When clicking on the user initials in the bottom left corner of Navigator, if you have remote workers connected to your controller instance, these workers will now display in the pop up window.
User Impact:
- This provides a quick insight into how many workers are connected to a controller instance of Navigator.
Bug Fixes
Back Button When Viewing Run History
Resolution: When viewing the Run History while you are on the Preview tab the back button should now take you back to the Preview tab rather than the Canvas.
User Impact:
- Improves user navigation while reviewing the run history and navigating back to the previous page they were on.
Authentication Issues While Offline
Resolution: While offline, users should now be able to login through the login screen without getting stuck
User Impact:
- Improves user offline experience if user does get logged out while offline.
Recent Projects Navigation Improvements
Resolution: When clicking on a Project from the Recent Projects folder you should now be directed to the correct Canvas/Project.
User Impact:
- Avoids user confusion on incorrectly displaying the Project that was clicked within the Recent Projects folder.
Chat History Displaying On New Canvas Run
Resolution: A Chat history should be cleared in Navigator on each new run within an LLM and API flow.
User Impact:
- Prevents older conversation history from being present on new Canvas runs.
Cleanse Canvases of Deleted Workers/Devices
Resolution: When a worker node or device is deleted we now remove that node/device from any connected elements within a canvas.
User Impact:
- This should provide a consistent up to date canvas during development for what devices and nodes are available while building and deploying workflows.
-
Release Date: 11/6/2025
Environment: Production
New Features
New Model Select for LLM Element Settings
Within an LLM Element Settings there is a new display of the selected model, along with filtering and search options.
Technical Details:
-
New default state for a selected model with details on model capabilities when hovering over model name
- Filtering options for webFrame, API Keys, No API Keys, Multi-Modal, and Optimization.
- Search functionality across all available models.
- Icons updated for model details regarding the filter options.
- The model select will also be disabled while a flow is running.
User Impact:
- Users will now see this new view on any flow that utilizes the LLM Element
New Preview and Apps Toggle
We've introduced a new way to interact with flows running locally in Navigator! You will now see a Preview, Canvas, and Apps toggle in the top left of the Canvas.
Additionally inside of the Preview tab, you will see a panel on the right side allowing you to edit the Element Settings directly from the Preview.
Technical Details:
- After a flow has completed running locally, it will automatically open the Preview tab for you to start interacting with your flow if it contains an LLM Element.
- You can also easily modify these Elements with the Element settings panel while in the Preview tab.
- This will also automatically create an Application which can be viewed in the Apps tab.
User Impact:
- Creating a dedicated tab for interacting with your LLM flows locally in Navigator provides a more seamless experience for indicating when a flow has completed running as this Preview tab will automatically open.
- Rather than navigating back to the canvas to make updates to an Element's settings, this can be achieved directly inside of the Preview tab.
Controller Mode vs Worker Mode
We've added a new article that details Controller Mode vs Worker Mode and some of the updates that were made to the views in Worker Mode and how to switch between each of these modes.
Improvements
Removed Add Node Button From Devices Page
Technical Details:
- Functionality stays the same, users can still add different devices manually or through the network as needed using the Add Worker button.
User Impact:
- This should help avoid confusion between the terms "node" and "worker" and consolidate these two buttons into one.
Long Device Names Overflowing On Deployment Configuration
If a device had a long name, this device was overflowing off the table during a deployment configuration.
Resolution: Reformatted the device names to not overflow from the table
User Impact:
- If using long device names, these devices should be displayed in full during the configuration steps of a deployment.
About Navigator Window Not Fullscreenable
When opening the "About Navigator" screen when Navigator was fullscreen it would take up the entire screen and not allow for this screen to be closed.
Resolution: Now allows for the "About Navigator" window to be closed while Navigator is being in fullscreen.
User Impact:
- Users shouldn't need to quit the entire application now to get rid of this "About Navigator" screen if they have Navigator open in fullscreen.
Replaced Device ID with Passcode Copy
We updated the copy when adding a worker to align with the copy of a device in worker mode. These should each use the terminology of "Passcode" now avoiding confusion.
New Document Hyperlink For Blank Canvas
We added a new document for our Blank Canvas template. Building Flows Inside a Canvas is a good starting point for users to get information on the Projects tab in Navigator with helpful next steps in building a flow within Navigator.
Run History Back Button
The back button when viewing the run history was sending users back to the canvas when it should send users back to where they were. This update should correctly send users back to their previous state.
Bug Fixes
Scrolling Issue - Multiple Users In Application
Resolution: Fixed formatting in Applications tab when scrolling an application with many users.
User Impact:
- Users can now scroll more fluidly in the Application Details page when several users are present on an application.
Previous System Prompt Set on LLM Element Settings
Resolution: We enabled system prompting for all models where previously we didn't show the system prompt for unsupported models.
User Impact:
- Now a user should see the system prompt associated with each LLM Element regardless of the model selected.
Dragging Canvases Into Project Folders
Resolution: Fixed bug allowing for loose canvases to be dragged and added into specific project folders.
User Impact:
- Users should not see any issues with dragging and dropping canvases into project folders.
Filter Out Used Deployments For Applications
Resolution: A fix was implemented to filter out used deployments, preventing them from appearing in the deployment dropdown during reassignment.
User Impact:
- Should help avoid confusion when selecting deployments from the deployments dropdown in the Edit Application modal.
-
-
Release Date: 10/9/2025
Environment: Production
New Features
Importing and Exporting a Canvas or Project
In Navigator you now have the ability to import and export canvases and projects. See the Importing and Exporting a Canvas or Project guide for more information.
Technical Details:
- Exported files are in JSON format of a Canvases and Project details
User Impact:
- Users can now share and upload Canvases and Project configurations built in Navigator across their teams.
- Allow for more collaboration and a file based version control allowing for more cross collaboration for Canvases and Projects specifically.
Improvements
ESC Key to Close Modals
User Impact:
- Users can now use the ESC key to exit out of any pop up modals.
Model Deployment From AI Models Tab In Navigator/Redesigned Buttons
There are now redesigned buttons inside of the AI models tab within Navigator to deploy a model automatically to Companion with just one click!
User Impact:
- If a user clicks `Try With` from the AI Models tab in Navigator, they now have the option to `Deploy to Companion` which automatically:
- 1) Creates a Canvas with the LLM + API element connected
- 2) Saves a version of this Canvas
- 3) Creates a Deployment
- 4) Deploys the created version
Controller/Worker Mode Switching
There are now options to switch between Controller Mode and Worker Mode within Navigator.
User Impact:
- A user can now click their profile icon in the bottom left corner of Navigator and select switch to Worker Mode. While in Worker Mode a user could then also switch back to Controller Mode.
UX Improvements for Filling Out Device IDs and Passcodes
Implemented auto-focus and auto-select on digit input boxes to reduce manual clicks and keystrokes, enhancing UX.
User Impact:
- This should reduce the friction of adding Device IDs and Passcodes by allowing users to finish these steps with less clicks.
Bug Fixes
Canvas Version Creation During Deployment
When creating a new Deployment the option to create a new Canvas version was providing a successful toast message but no new Canvas version was created.
Resolution: Updated logic for Deployment and Canvas version creation to work as expected aligning with the successful toast message.
User Impact:
- Stops confusion on Deployment creation without a Canvas version.
"Add" API Key Button Cutoff After Long Name Input
When adding an API key for the LLM element, if a long name was input the "Add" button would get cut off in the right panel of Navigator UI.
Resolution: Adjusted CSS to prevent shifting of buttons
User Impact:
- Users can now add longer API key names while still having the ability to click the "Add" button within an Element's panel settings.
Deployment Multi-Select State Persisting On New Pages
When selecting multiple deployments inside of a Cluster, those selected deployments continued to be selected when taking actions in other tabs within Navigator.
Resolution: Reset the state of the selected deployments when navigating and taking actions in other tabs.
User Impact:
- Stops confusion and incorrect action when redeploying or deleting specific deployments.
Organization Feature Disappears After Machine Sleeps
Resolution: Refresh expired access tokens for Organizations feature
User Impact:
- Users should now see the Organizations feature present and the org is correctly identified on the hosted cluster after machine sleeps
LLM + API Element Displays Incorrect Status with Distributed Flows
Resolution: Aggregated webFrame node statuses
User Impact:
- When users are utilizing multiple Nodes with an LLM + API setup they should not see the status turn green until all involved Nodes are actually ready and the flow can run.
-
Release Date: 9/4/2025
Environment: Production
New Features
Qdrant Vector DB Support
Build a working POC of using qdrant as an enterprise scalable Vector DB for large document repositories
- See how adding the qdrant DB to Runtime works
- Test out an Ingestion API for Navigator that would allow to provide data to that DB
- Test out querying the DB from Elements while running a Canvas/Deployment
- Investigate how the Sharding feature of qdrant works and how this relates to our Scaling solution
Technical Details:
- Download and install qdrant in runtime
- Adds in the Qdrant vector db as a managed service managed by runtime, following the same pattern of how RT already manages the webframe service.
- qdrant can be set as target for document ingestion or retrieval
User Impact:
- Available to users with the Qdrant build of Navigator
- Support for a faster, more enterprise scale Vector DB to be used to create RAG applications with Navigator
Offline Handling
Ability to perform a limited set of actions in Navigator while offline, and communicating to the user what will and won't happen while they are not connected to the internet.
Technical Details:
- Error handling and specific notifications based on internet availability
User Impact:
- Available to all users in current RC
- Always: A status indicator above their user icon indicating that Navigator may have reduced functionality since it does not have Internet connection
- Warning when the user logs out letting them know they will not be able to log back in
- If a run fails due to Internet connectivity problems, a descriptive toast is displayed
- If the user has no models downloaded, the AI models page doesn't show up
- If they do have models downloaded, then the downloaded models show for the user
Improvements
Add webFrame Support for OpenAI OSS 20B and 120 Model
OpenAI open source models are now supported in webframe
User Impact:
- Users can now run, distribute, and optimize OpenAI's open source models on their own hardware.
Set computer name during onboarding and change from User settings
Ability to give your computer a friendly name and change it via settings
User Impact:
- When setting up computers, you can give them a friendly name instead of the set device name
Disk Space Requirements Error Message
If a user does not have enough disk space to run a model they are presented with an error showing how much space they need to run the model. The error shows the TOTAL disk space required to run the model
User Impact:
- Easier for users to understand how much space is required to run their model and if they should add or change clusters/models
Support file type and folder setting hints
Tells users if their folders have unsupported file types, and which file types are supported.
User Impact:
- Stop users from attempting runs with unsupported files and let them know which files are supported by that element
Bug Fixes
LLM Chat Template Documentation Link Updated
Link to template documentation was out of date
Resolution: Points to new URL
User Impact:
- Updated documentation for LLM Chat Template
Redeploy previous deployments failing
Attempting to redeploy old deployments was failing in the last Release Candidate
Resolution: Improvements to underlying deployments logic
User Impact:
- Redeploying old deployments works as expected
Deployment Options dropdown fix
When making a new application, after you select the cluster of interest, the deployment options dropdown doesn't pick up changes to the actual deployment status.
Resolution: New application deployment options dropdown
User Impact:
- Whenever a user changes their deployment statuses, the changes are immediately reflected in the new application form
Canvas Run button should be disabled depending on the status
Whenever a user presses the Run button, the status changes to In Progress and to Active. When it's in these states, clicking the button is still available even though clicking it doesn't do anything.
Resolution: The button is not clickable when the state is In Progress or Active
User Impact:
- Stops user confusion or errors caused by users not knowing the state of their run, or clicking the button additional times which may cause flow errors or unintended runs.
-
Release Information
- Version Number: v2.323.0
- Release Date: 8/18/25
- Environment: Production
Summary
This release continues the focus on error handling and user feedback, adding additional layers of inspection for Navigator as well as more descriptive actions for users to take.
Improvements
More detailed webframe and canvas run errors
What Changed:
- Checks Hugging Face api key validity
- Checks Hugging Face api key model access
- Settings components now get mounted in the background so they can perform validation and raise errors up to new CanvasProvider, which allows checking for errors when attempting to run the flow
Screenshots/Visuals (if applicable):
Webframe & Canvas errors.mov
User Impact:
- Clearer feedback to users why their run is failing, directing users to the element and setting that is failing. This change is specifically focused on API key validation and model access.
Add Element filepath validation
What Changed:
- If the user tries to run a flow with a dataset whose filepath no longer exists (i.e. the dataset has been deleted), they are presented with an error.
- No run. Prior to running flow. No removing the file path but showing the error.
- Error toast won't show unless user hits run
- Filepath is checked
- When the file is set
- At regular intervals
- When Navigator is focused
Screenshots/Visuals (if applicable):
Filepath error.png
User Impact:
- When users run a flow, if there is an error with the filepath they have specified they will be told that is the problem and prompted to update the filepath setting.
User alerted when they don't have compute resources required to run a selected model
What Changed:
- User is informed via banner that their current cluster does not have enough RAM to run their current model. An alert is shown in the settings panel as well.
Screenshots/Visuals (if applicable):
- Requirements not met - webframe.png
- Requirements not met - webframe settings panel.png
- Requirements not met - non-webframe.png
User Impact:
- Users know they are not capable of successfully running a flow before they run it
User alerted when they have not accepted a license agreement required to run a model
What Changed:
- The user is provided with an error when they have not accepted the HuggingFace license agreement for gated models.
- We provide a link to the gated model license page
- The user is notified once the license agreement has been approved by HuggingFace
Screenshots/Visuals (if applicable):
License Agreement Not Accepted.png
User Impact:
- Users know which models they need to accept an agreement to use. This stops them from having failing runs because they have not accepted the agreement, and keeps them in compliance with the models they are using.
Navigator errors and debug data shows more information about element specific errors
What Changed:
- More specific errors from elements show above the element on canvas and in the debug data
Screenshots/Visuals (if applicable):
- Element error - on canvas.png
- Element error - debug settings.png
User Impact:
- For this version, the errors are limited to telling the user that a dataset folder is empty. But going forward this is the basis for more element configuration user feedback.
Remove stop canvas tooltip appears when hovering over running canvas button
What Changed:
- The tooltip shows "Stop canvas" when hovering over the run button when a canvas is in progress.
- This tooltip should be removed completely, it's never possible to stop the flow by clicking on the button. The stop button is in the dropdown.
Screenshots/Visuals (if applicable):
Stop Canvas Tool Tip.png
User Impact:
- Removing superfluous and confusing tool tip.
Bug Fixes
Clear chat button overlapping with chats in the navigator chat modal window
Issue: "Clear Chat" button overlapped with chats in modal
Resolution: "Clear Chat" Button moved to the chat modal title bar
User Impact: Clear Chat button is out of the way and in a less intrusive place
Change Log Level stops Navigator from Loading
Issue: If you change the log level on 2.296.2-stable it prompts you to restart the Navigator Console, on restart Navigator console will not progress past step 3 of the load.
Resolution: Bug was caused by an encryption key mismatch - Log Level changes no longer require a restart, which was causing the key mismatch.
User Impact: When switching log levels Navigator will stay running and not require restart, or lock users out
Too many decimal places shown in total memory available.
Issue: Available memory would show as many as 14 decimal places
Resolution: Show max 2 decimal places
User Impact: More readable and useful memory calculations
Projects get stuck and are unable to load Buttons & dropdowns are unresponsive
Issue: lag when clicking the "Run" button in the Projects screen. After clicking the "run" button it gives no indication it has been clicked for up to 15 seconds. This results in users clicking more times, this generates multiple notifications, such as the attached run error and multiple messages when Canvas run stops.
Resolution: Set canvas status to busy sooner
User Impact: Buttons respond faster, stopping users from clicking them repeatedly or thinking Navigator is stuck
Deprecations & Removals
Github Login
What Changed: Removing the GitHub Login option
Reason: Removed for easier publishing to the App Store, will be brought back with other "login with" options later
Migration Path: Login with your webAI credentials. If you need to set a password for the email used with GitHub, use Forgot Password to create a password for that email.
Known Issues
Cannot Redeploy an existing Deployment if a node added to the cluster is offline or disconnected
Impact: Users with clusters containing more than 1 node, especially if a node is offline.
Workaround: Remove the worker node from the cluster, and re-add it if necessary.
Expected Resolution: End of next sprint (Planned for end of August)
-
Release Date: July 7, 2025
Summary
This release is focused on user feedback improvements for canvas and element settings to help users better understand what components and settings work together, and what to do if the flow is not configured properly. We also introduce Fast EWQ, an 10x improvement in the speed of EWQ. Additional security improvements are also included
Improvements
Security Improvements
Additional security has been added to data and access to critical components of the webAI platform, including encryption enhancements to the data transferred between Navigator and runtime.
User Impact:
- Better app security posture
- A fix for Navigator not having a matching API key with Runtime in certain circumstances that caused Navigator to crash.
Loading state for elements
Navigator will now show a Loading status for elements where no download progress is reported.
User Impact:
- Clearer communication to users about the state of elements as flows start.
Fast EWQ button
A Fast Optimizer for EWQ option has been added to the LLM Element's settings. Fast EWQ is a result of us developing a classifier model which checks model weight json and understands how to distribute rather than loading them and analyzing like what regular EWQ does.
User Impact:
- Enable Fast EWQ instead of Regular EWQ, which is 10x faster in analyzing the weights for quantization
- Fast EWQ option is not ready for multimodal models. It is suggested to still use Regular EWQ with non-multimodal models.
Additional User feedback on flows
If an element requires an output to function and the user does not provide one, an error message is displayed when the user hits run:
"Project has no output: Your current flow needs an output to be complete. Please add one to finish this flow. View output elements"
For example, this error would be displayed if the media loader is placed on canvas with no other element.
This fix also provides a link that will direct a user to the Output elements.
User Impact:
- Users will now get feedback from the Canvas when flows are missing output elements, and will be prompted to add one from the list linked in the notification banner.
Refresh Button in Element Drawer
A refresh button has been added to the element drawer to make it easier to see all sideloaded elements, as well as pull any new elements that may have been added to a registry.
User Impact:
- Pull side loaded or other elements into the drawer without restarting Navigator.
Updated Element API key Settings
- Invalidate all element tags when registry updates are received
- Support for api key provider hints (added in latest version of dataset generator - https://gitlab.com/iris-core/systems-senior/runtime/webai-elements/-/merge_requests/319)
User Impact:
- Only compatible API keys will be selectable within elements, helping users more easily configure elements and flows
New initial chat message shown on first chat
A new message in the chat box inside Navigator has been added that covers the entirety of the chat window until the user has sent a message.
User Impact:
- Clearer first chat experience when chatting with models in Navigator
Removal of default system prompt
Navigator no longer sends a generic system prompt when no prompt is set in the LLM element.
Navigator previously sent a generic system prompt by default if no prompt was set in the element's settings; now it sends no system prompt unless explicitly defined.
This change does not affect the user interface; only the JSON payload to the /completions endpoint is altered.
User Impact:
- No change to the UI, but responses in companion will be better and more predictable
Bug Fixes
Unclear error message when an org invite code has already been accepted
Previously, our error codes were unclear when accepting an invite to an Organization. This was caused by a user accepting an event that they already accepted.
Resolution: Updated error message telling the user the code has already been used
User Impact:
- Clearer understanding that a user has joined the org successfully
Handle Memory cap settings as GB not bytes
Previously, a slider for adjusting the memory for Large Language Models (LLMs) used bytes as its unit, which led to display issues.
Resolution: The slider has been updated to use gigabytes (GB), making it much easier for users to understand and adjust the amount of available RAM for your models.
User Impact:
- The slider now properly handles cases where the total usable memory might initially be zero.
- It automatically adjusts the value if you set it above the maximum limit, ensuring you can't accidentally assign an impossible amount of memory.
Stop setting max value of slider to total memory available for non-memory cap related sliders
In the slider setting component, we have some logic which is only relevant to when the setting is being used to set memory cap and was setting the maximum to the total available memory. This logic was being incorrectly applied to all setting types, which resulted in updating the setting value to the max, which was defaulting to 100.
Resolution: Updated slider logic
User Impact:
- All sliders now accurately reflect the intended values for their respective setting
Redact all known oauth tokens from logs export and console
We've improved security by redacting all known OAuth tokens from logs and the console. Previously, these tokens were sometimes stored in plain text, which was a security risk.
Resolution: Tokens are removed from log exports and the renderer-state.json file
User Impact:
- Better security and your sensitive information is better protected.
Companion Release Notes
-
Release Date: 1/12/2026
Environment: Production
Improvements
Various bug fixes and stability improvements
-
Release Date: 12/17/2025
Environment: Production
Improvements/Updates
Switched Local Model to Gemma
User Impact:
- Users will now see the local models for Companion using Gemma models.
Bug Fixes
Various bug fixes and stability improvements.
-
Release Date: 11/24/2025
Environment: Production
New Features
Response Feedback
Technical Details:
There are now button icons to provide feedback on responses inside of Companion! See our LLM Feedback article for more information.
Three new button icons have been added:
- 1) Thumbs Up
- 2) Thumbs Down
- 3) Flag
User Impact:
- Users will now have the option to provide feedback regarding the responses an LLM provides. These feedback button icons will only show up for actively deployed LLM's in Navigator that have an associated Application.
Improvements
Show Offline Chats Toggle
All offline remote chat applications will be nested under a "Show Offline Chats" toggle.
User Impact:
- This provides a cleaner, less noisy interface for users to chat with only active remote Applications or the locally downloaded models.
-
We're excited to announce the release of Companion, webAI's private AI chat interface that allows you to interact directly with models running on your device or within your organization's compute environment. Companion is a standalone application that works seamlessly with Navigator, automatically displaying any models you've deployed from Navigator. With features like file generation, web search, reasoning mode, and speech-to-text, Companion provides a comprehensive and private AI chat experience. For detailed setup instructions and usage guides, see Getting Started with Companion and Using Companion.