Our own engineering — Harmonica
Connecting Azure model deployments to an existing workflow
Bringing organization-specific Azure model deployments into chat and model-backed workflows.
Context
An Azure model deployment has an endpoint, deployment name, credentials, and operating requirements of its own. An existing application needs to associate those settings with the correct users and requests.
Approach
We integrated Azure OpenAI access through Microsoft Foundry into Harmonica’s chat and model-backed workflows. Organization-specific settings travel with the application’s request context.
Our role
The integration connects the organization’s configured endpoint, deployment, and API credentials to model requests. The application retains the boundaries around the organization that owns those settings.
Streamed responses, tool behavior, configuration errors, and credential separation are handled in the integration. This is the work needed to fit model access into an existing product workflow.
- Azure model deployment access
- Organization-specific endpoint and credentials
- Application integration and failure handling
What this demonstrates
The implementation connects an Azure model deployment to existing application workflows while keeping the organization’s settings associated with its requests.
Explore the engineering
Microsoft Foundry integration
Bring experienced engineers into the systems and workflows where the challenge lives.
Explore Microsoft FoundryMore work
Let’s put your next priority into production.
Tell us about the workflow, integration, or delivery challenge your team needs to move forward.
Discuss your next priority