Avatar Backend Diagnostics
Local diagnostics inspect the selected model, capability profile, route, and presentation result without granting Runtime privileges. This is Avatar-internal tooling, not a separate third-party App diagnostic API.
Current Result Shape
The internal factory returns AvatarDebugSession, not DiagnosticSessionResult:
| Field | Meaning |
|---|---|
debugSessionId | Local diagnostic session ID |
probeId and probeKind | Identity and kind of this evaluation |
avatarInstanceId | Nullable instance reference |
avatarPackageRef | Nullable package reference |
backendCapabilityProfileRef | Nullable capability-profile reference |
backendKind | live2d or vrm |
observedAt | Observation timestamp |
evidence | Single AvatarDebugEvidence object, not an array |
evidence contains evidenceId, evidenceKind, status, source, nullable reasonCode, and bounded refs. status is passed, failed, unsupported, or invalid. Read the single evidence object directly rather than calling evidence.map. The backend discriminator field is named backendKind, not backend.
Interpreting a Result
An evaluation can check package descriptors, backend loading, profiles, route support, generated motion, expressions, lipsync, or carrier results. It proves only the object and conditions actually inspected. Structural or local diagnostic success does not establish a live App session, arbitrary model support, or acceptance by another backend.
Retain the returned reasonCode and refs on failure, then repair the implicated model or profile. Do not reinterpret unsupported as successful idle playback. The source references below define the actual return types and factory implementation.