Constructor ChatBackendReport

Namespace

Qavren.Edge.Chat

Assembly

Qavren.Edge.Chat.Onnx.dll

ChatBackendReport(string, string, string, string?, IReadOnlyList<string>, bool, EdgeGuidanceProbeResult, string)

What the shipped native build and this model actually turned out to be.

public ChatBackendReport(string GenAiVersion, string OrtVersion, string RuntimeIdentifier, string? Abi, IReadOnlyList<string> Providers, bool ChatTemplateSupported, EdgeGuidanceProbeResult Guidance, string PromptFormatter)

Parameters

GenAiVersion string

The Microsoft.ML.OnnxRuntimeGenAI package version, taken from the assembly metadata the csproj emits out of the central package pin. Not read from the GenAI assembly: it reports AssemblyVersion 0.0.0.0, FileVersion 0.0.0.0, carries no AssemblyInformationalVersionAttribute, and its managed surface has no version API (measured - plan adjustment 1).

OrtVersion string

The ONNX Runtime assembly version loaded in this process.

RuntimeIdentifier string

The running RID.

Abi string

The running Android ABI, or null off Android.

Providers IReadOnlyList<string>

The execution providers named in the resolved configuration.

ChatTemplateSupported bool

Whether minja parsed the model's own Jinja template. False is not a guess: TokenizerImpl::LoadChatTemplate returns OK with only a warning when it cannot, so the failure would otherwise surface mid-sentence on a user's first message.

Guidance EdgeGuidanceProbeResult

The guidance probe's verdict. NotEnforced means not proven enforced and not proven absent.

PromptFormatter string

Which formatter this model's prompts go through: the model's own chat template, the consumer's ChatPromptFormatter, or the role-prefixed fallback whose output quality is materially worse.