GitLab Duo Health Check: Dynamic Status for Hybrid Model Configurations

Platform: GitLab · Type: Frontend / UX Logic · Status: Early exploration · Issue: 561092

Context

GitLab Duo provides AI-powered features that can operate in different configurations, including hybrid setups combining self-hosted models and GitLab-managed (vendor) models.

A health check system exists to verify connectivity and availability of these components, helping administrators understand whether their AI configuration is functioning correctly.

Problem

The current health check header provides only a generic status message:

"No health problems detected" when all checks pass
A generic error message when any check fails

This becomes insufficient in hybrid configurations, where multiple systems (self-hosted and vendor models) may fail independently.

As a result, administrators cannot easily determine which part of the system is experiencing issues, making debugging and recovery slower and less intuitive.

Exploration Goal

This issue aims to improve the clarity of the health check header by making it aware of hybrid configurations and displaying more specific status messages.

The intended direction is to:

Separate results into self-hosted and vendor model sections
Detect which section is failing
Display targeted, actionable messages in the header

Proposed Direction

Instead of a single generic message, the header can reflect the actual system state:

// Example logic (conceptual)
```

if (selfHostedOk && vendorOk) {
message = "No health problems detected"
} else if (!selfHostedOk && vendorOk) {
message = "Self-hosted model connection issues detected"
} else if (selfHostedOk && !vendorOk) {
message = "GitLab AI vendor connection issues detected"
} else {
message = "Multiple model connection issues detected"
}

```

This approach keeps the UI simple while significantly improving diagnostic clarity.

Current Status

This work is currently in an early exploration phase.

The next steps involve identifying the frontend components responsible for rendering the health check header and understanding how probe results are structured and passed from the backend.

Once the data flow is clear, the dynamic messaging logic can be introduced incrementally.

Reflection

This issue highlights the importance of meaningful system feedback in complex, hybrid environments.

Even when underlying health checks are accurate, insufficient communication at the UI level can limit their usefulness.

This exploration focuses on improving that communication layer, bridging the gap between system state and user understanding.

It also represents an early-stage contribution involving system exploration, component discovery, and incremental frontend enhancement within a large-scale codebase.