🔄 GPT-4 End of Life & GPT-4o Model Upgrade

Important: GPT-4 and GPT-4-32k will reach end of life on June 6, 2025.

💡 What’s Changing

With Release 22, we are upgrading all GPT-4 models to GPT-4o (AZURE_GPT_4o_2024_1120) across both our:

• Chat solution

• Recording solution

This new model brings:

• Swiss-based inference only

• Expanded context window (128K tokens)

• Higher throughput

• More verbose reports (e.g., ending with “Do you want me to refine this?”)

🚀 Upgrade Path

• If you deploy Release 22:

  • The GPT-4o upgrade is applied automatically for both Chat and Recording.

• If you do not deploy Release 22 by June 6:

  • You must manually switch your GPT configuration to AZURE_GPT_4o_2024_1120

  • This applies to both Chat and Recording

⚠️ Not upgrading by June 6 may lead to broken or unexpected behavior in Unique’s chat and reporting.

📞Need help? Contact Customer Success to configure GPT-4o manually or test the setup early.

📊 Analytics

🚀 What’s Changing

Starting with Release 22, we’re rolling out the analytics improvements that were previously available in Release 2025.16 as a beta feature. These enhancements are designed to provide greater stability when exporting analytics data and are now being made available to more users.

Key Update:

• In the Chat Interactions Analytics, the order of exported data in the CSV file has changed.

  • Previously: Data was sorted by the Email column.

  • Now: Data is sorted by the Date column.

🔧 Feature Availability

• The feature flags for these improvements are now enabled by default. To configure this in your tenant, set this flags to be true

  • ENABLE_DOWNLOAD_UPLOAD_FILES_AS_STREAM

  • ENABLE_UPLOAD_CHAT_INTERACTIONS_AS_STREAM

  • ENABLE_ANALYTICS_CHAT_DETAILED_INTERACTIONS_PRISMA_CURSOR

• In the next release, these improvements will become Generally Available (GA), and the feature flags will be removed—making the functionality available to all users without additional configuration.

Changes on Application environment

Path changes of hooks

The folder structure inside the docker images has been adjusted. These changes need to be replicated on the helm hooks.

hooks:
  migration:
    command: |
      cd /home/node/app/prisma; npx prisma migrate deploy

hooks:
  migration:
    command: |
      cd /home/node/app/services/configuration-backend/prisma; npx prisma migrate deploy;

hooks:
  migration:
    command: |
      cd /node/dist/apps/node-chat; npx prisma migrate deploy; cd /node; RUNNING_MODE=DATA_MIGRATION  node /node/dist/apps/node-chat/main.js up;

hooks:
  migration:
    command: |
          cd /home/node/app/services/node-chat/prisma; npx prisma migrate deploy; cd /home/node/app/services/node-chat/dist; RUNNING_MODE=DATA_MIGRATION node /home/node/app/services/node-chat/dist/main.js;

hooks:
  migration:
    command: |
      cd /node/dist/apps/node-ingestion; npx prisma migrate deploy

hooks:
  migration:
    command: |
      cd /home/node/app/services/node-ingestion/prisma; npx prisma migrate deploy;

hooks:
  migration:
    command: |
      cd /node/dist/apps/node-scope-management; npx prisma migrate deploy; cd /node; RUNNING_MODE=DATA_MIGRATION  node /node/dist/apps/node-scope-management/main.js up;

hooks:
  migration:
    command: |
      cd /home/node/app/services/node-scope-management/prisma; npx prisma migrate deploy; cd /home/node/app/services/node-scope-management/dist; RUNNING_MODE=DATA_MIGRATION  node /home/node/app/services/node-scope-management/dist/main.js;

New Helm Charts available

This change is voluntary. There are new Unique Helm Charts available for backend-service, ai-service and web-app which improves the security by adding defining a stricter securityContext rule. If using them you can get rid of the custom overrides in the values file of the apps.

Check the changes on Artifact Hub: https://artifacthub.io/packages/search?org=unique

Upgrading Kong Ingress Controller (KIC)

This change is voluntary and reduces technical debt. It does not modify or alter functionality.

Steps

1. Upgrade the Kong Ingress Helm Chart to latest or at least 0.19.

2. Upgrade the Kong Ingress Controller to the latest version or at least 3.4.4.

3. Remove the manual extraObjects#uniquens-chat-redirect and extraRoutes#root-redirect deployed to the chat backend (node-chat, backend-service-chat).

4. Enable the disabled default route for the chat backend (node-chat, backend-service-chat):

   …
   routes:
     …
     paths:
       default:
         enabled: true

5. Deploy the changes

6. Verify the redirect by calling your Unique AI without /chat and observing the redirect.

Example

This commit within hello-azure shows the necessary steps.