Globant Enterprise AI Deployment Guide - 2026-01

Below are the end‑to‑end procedures to deploy Globant Enterprise AI for Version 2026-01 and its Hot Fix.

Check also version updates for the main platform components.

Hot Fix 1 (March 12)

No manual steps required.

Version 2026-01 (March 1)

Breaking Changes

• Station-Frontend & saia-chat-frontend now listen on port 8080

  Starting with this version, you must perform a container port migration. The updated frontend images move the container port from 80 to 8080.
  Update the service deployment definition accordingly to avoid deployment failures.

• Single tools are grouped into Integrations

  Public Tools from the installation (Remote Modules) are converted into Integrations that group one or more Tools.
  • Changes:
    - Common parameters are moved to the Integration level.
    - Parameters are removed from the Tool level.
    - Authentication parameters are replaced by a Credential.
    - Credentials are linked to the Integration.
  • During migration:
    - Sensitive parameters from Tools are backed up.
    - They are restored into Credentials.
    - They are automatically linked to the new Integrations.
    Private Tools in each Project are grouped into an automatically created Integration with a 1-1 relationship.

• Flows Module

  The SAIA integration URL, previously configured in Integration settings, now uses an environment variable to address a security issue.

Required manual steps

• Step 1: Pre-deployment
• Step 2: Deployment
• Step 3: Post-deployment

Step 1: Pre-deployment

1. Disable Auto-Update for all Remote Modules involving Agents, Tools, and Integrations.
   • Run the following SQL before deployment:

     UPDATE Module SET ModuleAutoUpdate = '0' WHERE (ModuleId = 'saia.lab.agents.tools.gdrive.v1');
     UPDATE Module SET ModuleAutoUpdate = '0' WHERE (ModuleId = 'saia.teamstudio');
        

     After updating the modules in step 1 of the post-deployment phase, the Auto-Update property is automatically reactivated.
2. Perform a database backup.
3. Add the Flows base URL environment variable to the API and Console containers.

Add the following environment variable:
GEAI_SAIA_SERVER_URL

Description:
GEAI_SAIA_SERVER_URL = SAIA API endpoint

Sample:
https://api.saia.ai

Flows Container - OAuth Variables

To obtain OAuth configuration values:

1. Go to Console.
2. Navigate to Security → GAM Configuration.
3. Open Applications.
4. Select the application used for login (Lab & Station).

In the OAuth Authentication tab you can find the required values.

Sample: FLUENT_LAB_BASE_URL="https://botbuilder.fluentlab.ai/fluentlab"

OAuth variables:

Step 2: Deployment

Execute the deployment:

1. Update container images.
2. Run database migrations.

Important considerations

• Launch the API before updating any containers so schema and data migrations can run.
• Migrations run asynchronously and may take time.
• Start them before performing container updates.
• This process can take approximately 1 hour because it includes data migration.
• A maintenance window of 1-2 hours is required.
• Perform a full database backup before deploying the new version.

Step 3: Post-deployment

The execution order is crucial for a successful migration.

Integrations and Tools

1. Verify Lab module version. The Lab module must be in Revision 34. Do NOT reinstall the module. Nothing should be done here.
2. Update The Lab Public Component modules. This groups Public Tools into Integrations and moves parameters from Tools to Integrations.
3. Update The Lab Tools: Google Drive Tools modules (if previously installed). This groups Public Tools into Integrations and moves parameters from Tools to Integrations.
4. Reinstall The Lab module. This processes backed-up parameters and links them to the new Public Integrations. The reinstall process of The Lab module can take 10-15 minutes. The Web UI will timeout after clicking on Reinstall, but the process will continue running in the background.

To verify completion:

• Check the console logs.
• Search for TeamStudio entries.
• When no new logs appear, the process has finished.

Reprocess requests to recalculate dashboard metrics

A new dashboard dimension called Provider has been added.

To update historical data:

• Reprocess the data.
• Otherwise, the Provider field will appear empty in historical dashboards.
• Follow the instructions in Reprocess Requests and load summary table

Configuration updates

1. A new profile disables analytics containers in Flows (fluentlab). To disable them, add the analytics-disabled profile. To activate it, include it in the SPRING_PROFILES_ACTIVE environment variable.
2. Ensure system modules are up to date. Navigate to Console → System → Modules. Verify that the Security Configuration module is updated.

Configure Web Search Tool

EngineId

Add the following system parameter for Web Search Google EngineId:

$(toolplugin_d39eadfb-35cc-4852-9203-0fe6c1650bab_engineId)

API Key

Add the following environment variable to infrastructure (Container: API Y Worker):

CREDENTIAL_d39eadfb-35cc-4852-9203-0fe6c1650bab_APIKEY: ${{ secrets.WEB_SEARCH_API_GOOGLE_KEY}}

Keycloak configuration update

1. Open the Keycloak Admin Console.
2. Select the FluentlabTenant realm.
3. Go to Identity Provider:
   • Select Globant Enterprise AI identity provider.
   • Enable Store tokens.
   • Enable Stored tokens readable.
4. Go to Realm roles:
   • Select default-roles-fluentlabtenant.
   • Select Assign role → Client role.
   • Search for client broker.
   • Select the role read-token.
5. Go to Client scopes:
   • Select roles.
   • Open the Mappers tab.
   • Select Add mapper → By configuration.
   • Select User Client Role.
   • Configure the mapper:
     • Name: Broker Client Roles
     • Client Id: broker
     • Multivalued: On
     • Token Claim Name: resource_access.${client_id}.roles
     • Claim JSON Type: String
     • Add to ID token: Off
     • Add to access token: On
     • Add to lightweight access token: Off
     • Add to userinfo: Off
     • Add to token introspection: On

Optional configurations

Azure AI Foundry Clients

Create environment variables for the newly available GPT models in Azure AI Foundry.

Models in the eastus2 region:

• azure_ai_foundry/gpt-5.2
• azure_ai_foundry/gpt-5.1-codex
• azure_ai_foundry/gpt-5.1-codex-mini

Required environment variables in container saia-litellm:

AZURE_AI_FOUNDRY_OPENAI_API_BASE_2
AZURE_AI_FOUNDRY_API_KEY_2

In the GEAI Console, set the new models status to Active.

AWS Bedrock Clients

• In the GEAI Console, install the Rev. 51 of the remote module.
• Set the following models to Active:
  • awsbedrock/us.anthropic.claude-sonnet-4-6
  • awsbedrock/us.anthropic.claude-opus-4-6-v1
• Test the models by creating an agent in The Lab.

Vertex AI Clients

• In the GCP Console, go to Model Garden and enable the following models:
  • Anthropic-> Claude Sonnet 4.6 (region global)
  • Anthropic-> Claude Opus 4.6 (region us-east5)
  • GLM 5 API Service
• In the GEAI Console, install Rev. 128 of the remote module.
• Set the following models to Active:
  • vertex_ai/claude-sonnet-4-6
  • vertex_ai/claude-opus-4-6
  • vertex_ai/gemini-3.1-pro-preview
  • vertex_ai/zai-org-glm-5-maas
• Test the models by creating an agent in The Lab.