# Release 2026.24 Infrastructure Changes

## Changes on Application environment

| **Change** | **Environment Variable Name** | **Application Default Value**<br>**(if env variable unset)** | **Example** | **Required** | **Applications** | **Short Description** |
|------------|-------------------------------|--------------------------------------------------------------|--------------|--------------|-------------------|------------------------|
| Added      | `CHAT_BACKEND_API_URL`       | _(set per environment overlay)_                               | `http://node-chat.chat.svc.cluster.local:8093` | `true`       | - `backend-service-configuration`           | The route to reach the `backend-service-chat` from `backend-service-configuration` |
| Added      | `SCOPE_MANAGEMENT_API_URL`   |                                                              | `http://node-scope-management.chat.svc.cluster.local:8094` | `true`       | - `backend-service-configuration`           | Needed for the feature flag service to fetch list of users for user targeting |
| Removed    | `FEATURE_FLAG_REFERENCE_STYLE_2_UN_16900` | -                                                    | -            | -            | - `backend-service-configuration`           | -                      |
| Removed    | `FEATURE_FLAG_REFERENCE_STYLE_REVAMP_UN_16900` | -                                                   | -            | -            | - `backend-service-configuration`           | -                      |
| Added      | `CONFIGURATION_BACKEND_URL`  | _(none — raises ValueError if unset)_                      | `http://configuration-backend.<namespace>.svc.cluster.local:8095` | `true`       | - `ai-service-agentic-ingestion` | In-cluster URL of `backend-service-configuration` used for per-request remote feature flag evaluation. When unset, the service raises a `ValueError` and returns HTTP 500 on every extraction request. |
| Removed    | `FEATURE_FLAG_ENABLE_PDF_CONTENT_EXTRACTION` | -                                                 | -            | -            | - `ai-service-agentic-ingestion` | Moved to `backend-service-configuration` 
| Added      | `FEATURE_FLAG_ENABLE_PDF_CONTENT_EXTRACTION_UN_16985` | `true` | `true` / `false` | `false`      | - `backend-service-configuration` | Controls PDF content extraction in `ai-service-agentic-ingestion` per client. Defaults to `true`. |
| Added      | `FEATURE_FLAG_AGENTIC_TABLE_COPY_ACTION_UN_20824` | `false` | `true` | `false`      | - `backend-service-configuration` | Enables copy functionality on Agentic Table cells that copies the answer and strips citation references and supporting notes. |
| Added      | `FEATURE_FLAG_ENABLE_HTML_INLINE_IMAGE_EXTRACTION_UN_20936` | `false` | `true` | `false`      | - `backend-service-configuration` | Enables extraction of text from inline base64 images during HTML ingestion in `backend-service-ingestion-worker`, per client. For the UI to write the config, `FEATURE_FLAG_ENABLE_MULTI_FILE_IMAGE_CONTENT_EXTRACTION_UN_20936` must also be enabled. (UN-20936) |
| Added      | `FEATURE_FLAG_ENABLE_MULTI_FILE_IMAGE_CONTENT_EXTRACTION_UN_20936` | `false` | `true` | `false`      | - `backend-service-configuration` | Enables the multi-file Image Content Extraction UI in `web-app-knowledge-upload` — a single toggle that applies image content extraction to both PDF and HTML files, per company. Backend HTML extraction itself is gated by `FEATURE_FLAG_ENABLE_HTML_INLINE_IMAGE_EXTRACTION_UN_20936`. (UN-20936) |
| Added      | `PROMETHEUS_MULTIPROC_DIR`   | _(unset — metrics collection not aggregated)_          | `/tmp/prom_multiproc` | `false`      | - `ai-service-assistants-core` | Enables Prometheus multiprocess metrics collection. When set, the entrypoint wipes and recreates the directory before gunicorn starts, and the gunicorn `child_exit` hook flushes per-worker metric files on worker exit. When unset, the feature is a no-op. |
| Added      | `FEATURE_FLAG_ENABLE_USER_MODEL_SELECTION_UN_20804` | `false` | `true` | `false`      | - `backend-service-configuration` | Enables user model switching in chat: the admin "Allow users to change the model" space configuration and the in-chat model dropdown that lets users pick from the admin-curated list of models. |
| Removed    | `FEATURE_FLAG_ENABLE_NEW_SPACE_ACCESS_MODAL_UN_17544` | - | - | - | - `backend-service-configuration` | The new space access modal is now GA and enabled for all clients. The old access modal code has been removed.