Purpose

This pull request makes several important updates to the deployment configuration, focusing on lowering the default GPT-4.1 model capacity from 150k to 50k tokens, updating documentation and templates to reflect this change, and improving security and maintainability in deployment workflows by switching from secrets to environment variables for Azure Container Registry (ACR) access and clarifying role assignments.

Model Capacity Reduction and Documentation Updates:

• Lowered the default GPT-4.1 model capacity from 150k to 50k tokens across Bicep templates (infra/main.bicep, infra/main.json), workflow files, and all related documentation. This includes updating default values, usage instructions, and quota check scripts to ensure consistency and clarity for users.
  Deployment Workflow Security and Maintenance:

• Switched from using GitHub secrets to environment variables (vars) for referencing the ACR login server in all relevant workflow files, improving maintainability and aligning with best practices for workflow configuration.

• Updated the Docker build workflow to clarify its purpose, add manual trigger support (workflow_dispatch), and set appropriate permissions for GitHub Actions OIDC authentication.

Role Assignment Improvements:

• Replaced string-based role names with explicit Azure role definition IDs in Bicep templates for Storage and Search resources, making role assignments more robust and less error-prone.

Workflow Logic and Template Updates:

• Adjusted workflow logic to control when the Docker build job runs, ensuring it only triggers on manual dispatch with the appropriate input.
• Updated Bicep-generated JSON templates to reflect the new Bicep version and template hashes, ensuring infrastructure as code is up to date.

These changes together improve the clarity, security, and maintainability of the deployment process while aligning documentation and infrastructure defaults with the new model capacity standard.

Does this introduce a breaking change?

• Yes
• No

Golden Path Validation

• I have tested the primary workflows (the "golden path") to ensure they function correctly without errors.

Deployment Validation

• I have validated the deployment process successfully and all services are running as expected with this change.

Other Information

This pull request makes several important updates to the deployment configuration, focusing on lowering the default GPT-4.1 model capacity from 150k to 50k tokens, improving Azure Container Registry (ACR) variable handling, and updating role assignments to use explicit GUIDs. These changes affect both the infrastructure code and documentation to ensure consistency and clarity for users and maintainers.

Model Capacity & Documentation Updates:

• Lowered the default GPT-4.1 model capacity from 150k to 50k tokens in deployment templates, parameters, and documentation to reduce resource requirements and align with current recommendations.

Azure Container Registry (ACR) Variable Handling:

• Switched from using secrets.ACR_TEST_LOGIN_SERVER to vars.ACR_TEST_LOGIN_SERVER throughout workflow files for improved security and flexibility. This affects Docker build, deploy, and image tagging steps.

Workflow and Job Configuration:

• Removed unused workflow inputs and conditional logic from job-docker-build.yml and deploy-orchestrator.yml, simplifying the triggering and execution of Docker build jobs.

Role Assignment Specification:

• Updated role assignments in Bicep templates to use explicit role definition GUIDs instead of role names, improving reliability of deployments.

Infrastructure Template Updates:

• Regenerated infra/main.json with Bicep v0.44.1, reflecting the above changes and updating template hashes and dependency order for improved reproducibility.

These updates collectively streamline deployment, improve security and maintainability, and ensure the documentation matches the current configuration.