Health & Diagnostics

HealthService exposes an always-available uno_health MCP tool and a uno://health resource so AI agents can inspect DevServer state even before the upstream Host is ready. The CLI also exposes uno.devserver health, with --json returning the same HealthReport payload. The bridge also exposes uno_app_select_solution so an agent can explicitly choose a Uno solution when auto-discovery is ambiguous or deferred.

Active Server Ownership

list, disco, and health all need enough runtime detail to answer one practical question: who started the DevServer that is currently running for this workspace?

To support that, active server diagnostics should include:

• ProcessId
• ParentProcessId
• IdeChannelId
• a bounded process ancestry chain displayed ancestor-first (IDE → ... → dotnet → Host) with PID + process name (up to 8 levels)

The ancestry chain is diagnostic only. It helps callers distinguish:

• a host launched by the current IDE session
• a host launched by another IDE
• a host launched indirectly by an MCP server or shell

This is especially important when uno.devserver start --ideChannel ... reuses an existing Host and updates the active IDE channel in-place instead of killing the Host.

Data Flow

DevServerMonitor.LastDiscoveryInfo
  --> DiscoveryIssueMapper.MapDiscoveryIssues()   (discovery issues)
  --> HealthService.BuildHealthReport()            (+ runtime issues)
  --> HealthReport { Status, Issues[] }
  --> DiscoverySummary { ..., ActiveServers[] }

Key Files

Discovery-Time Issues (DiscoveryIssueMapper)

Static mapper that converts DiscoveryInfo fields into ValidationIssue[]. Issues are evaluated in order; some trigger an early exit (no further checks):

Runtime Issues (HealthService)

These are added directly by HealthService.BuildHealthReport():

Non-Mapped IssueCode Values

Three codes exist in the IssueCode enum but are not currently mapped in CLI code -- reserved for Host-side or future use:

• DotNetVersionUnsupported
• AddInLoadFailed
• AddInBinaryNotFound

Notable HealthReport Fields

start --ideChannel reuse semantics

When a Host already exists for the selected solution:

• uno.devserver start without --ideChannel keeps the existing "already running" behavior
• uno.devserver start --ideChannel <id> reuses the Host and replaces the active IDE channel without restarting the process

Because the Host stays alive, health and disco must show the updated IdeChannelId and current process ancestry so launchers can verify that the running instance was reused rather than respawned.

IDE channel lifecycle logging

The Host emits the following diagnostic logs during channel operations:

The CLI also forwards the Host subprocess stdout/stderr at Debug level on success, so IDE extensions can see reuse decisions ("A DevServer is already running...") and rebind POST results.

Health Status

Fatal issue present --> Unhealthy
Warning(s) only     --> Degraded
No issues           --> Healthy

Adding New IssueCode Values

1. Add the new value to the IssueCode enum in Mcp/HealthReport.cs
2. Add mapping logic in DiscoveryIssueMapper.cs (if discovery-related) or HealthService.BuildHealthReport() (if runtime-related)
3. Add test in Given_DiscoveryIssueMapper.cs verifying the trigger condition and severity