2026.18 Infrastructure Changes
3 min read
Changes on Application environment
Change | Environment Variable Name | Application Default Value (if env variable unset) | Example | Required | Applications | Short Description |
|---|---|---|---|---|---|---|
Added |
|
|
| false |
| Sets the placeholder folder name shown when a user has access to a file, but not to the folder that contains it. When it is set to |
Added |
|
|
| false |
| When enabled, the system can include folders that the user does not directly have folder access to, but that contain files they are allowed to access. Those restricted folder names are shown using |
Added |
|
|
| false |
| When enabled, the chat's Select Files panel gains an Uploaded files tab which lists files uploaded in the current chat. Users can select or deselect individual files (or all at once) to control which uploaded files are included as context for their next message. Requires |
Added |
|
|
| no |
| When enabled, will start the version pooling behaviour on chat to detect version changes and show the modal if there is a missmatch. |
Added |
|
|
| no |
| When enabled, user would be able to preview files in sidebar |
Added |
|
|
| no |
| When enabled, users interacts and sees new sidebar navigation |
Added |
|
|
| no |
| Enables the new sharing functionality for Agentic Table |
Changed |
| (set per environment overlay) |
| yes |
| Routes RfpAgent (Agentic Table) to the dedicated assistants-agentic-table service at |
Added |
|
|
| no |
| Enables the new sharing modal UI in Space Management |
Added |
|
|
| no |
| Set to true or company IDs to enable LiteLLM usage for all or select companies . |
New service: assistants-agentic-table
With release 2026.18 there is a new service introduced named assistants-agentic-table. It’s a Python AI service (dedicated Helm chart / ArgoCD app) and must be deployed wherever the Agentic Table feature is enabled. The service will handle and process the Agentic Table requests - the functionality has been extracted from the ai-service-assistants-core service into it’s own service in order to better distribute the load.
In order to switch traffic from ai-service-assistants-core to ai-service-assistants-agentic-table , align backend-service-chat INTERNAL_MODULES_CONFIG environment variable so the RfpAgent targets this service eg. {"name": "RfpAgent", "apiURL": "http://assistants-core.<namespace>.svc.cluster.local:8081/core/rfp_agent"}.
New version Modal
With this release, the Chat web app can optionally detect when users are still on an older build after you’ve deployed a newer one. When that option is turned on, people who keep a tab open across a rollout get a short message asking them to reload, so they don’t hit odd errors from mixing old UI with new server behavior.
That option is controlled on the Chat deployment by the environment variable FEATURE_FLAG_DEPLOYMENT_VERSION_CHECK (set it to "true" to enable).
What you need to do in your environment
Chat’s Helm defaults still use a deployment style that replaces the app in one step. For this “new version” prompt to work reliably—and to avoid a rough cutover for users—you should run web-app-chat with a rolling update (new pods come up alongside old ones, then traffic shifts gradually) instead of recreate-style updates. In practice that means: in your Helm values overlay, set the deployment strategy to RollingUpdate.
Images should continue to be built with the usual release version passed in at build time so the browser and the server agree on “which build this is.”
LITELLM
Set LITELLM_ENABLED to true or example-company-id-1,example-company-id-2 to enable LiteLLM for all or specific organizations on an environment.
It’s important to set this variable for organizations that currently have LiteLLM enabled so as to not cause disruptions, as default value when not set is false. Failure to set variable for environments currently using LiteLLM will lead to a unavailability OF LiteLLM Models.