Class LiteRtResponse

Namespace

LiteRtLmSharp

Assembly

LiteRtLmSharp.dll

A model response: plain Text or one or more ToolCalls, plus — on reasoning models — a separate Thinking trace (see Channels). RawJson always holds the unparsed response (escape hatch).

public sealed class LiteRtResponse

Inheritance

object

LiteRtResponse

Inherited Members

object.Equals(object)

object.Equals(object, object)

object.GetHashCode()

object.GetType()

object.ReferenceEquals(object, object)

object.ToString()

Properties

Channels

Out-of-band channels the model emitted alongside the primary Text, keyed by channel name (the reasoning build emits a "thinking"/"thought" channel). Empty for models without channels, or when reasoning mode is off. See Thinking for the common case.

public IReadOnlyDictionary<string, string> Channels { get; }

Property Value

IReadOnlyDictionary<string, string>

IsToolCall

True when the model requested one or more tool calls.

public bool IsToolCall { get; }

Property Value

bool

RawJson

The raw response JSON from the engine.

public string RawJson { get; }

Property Value

string

Text

The assistant text, or null when the model emitted tool calls.

public string? Text { get; }

Property Value

string

Thinking

The model's reasoning ("thinking") trace, separate from the final Text answer, or null when there is none. Convenience over Channels (the concatenation of all channel content). Populated only when the conversation was created with EnableThinking = true on a reasoning model.

public string? Thinking { get; }

Property Value

string

ToolCalls

Tool calls the model wants executed (empty for a plain text answer).

public IReadOnlyList<LiteRtToolCall> ToolCalls { get; }

Property Value

IReadOnlyList<LiteRtToolCall>

Methods

ToMessage(bool)

Captures this reply as a LiteRtMessage (role Model) so it can be recorded in a history and later restored via History. Preserves the answer Text and, when includeToolCalls is true, the ToolCalls. The reasoning (Channels/Thinking) is intentionally dropped — replaying a thinking trace would re-consume the context window (the same reason FilterThinkingFromKvCache exists).

public LiteRtMessage ToMessage(bool includeToolCalls = true)

Parameters

includeToolCalls bool

Returns

LiteRtMessage