Streaming

Provider-level streaming

Every provider implements ILanguageModel.StreamAsync(LlmRequest) and emits LlmStreamChunk items.

Use this when you want raw provider streaming without the higher-level agent loop.

Agent-level streaming

Set UseStreaming = true on Agent when you want streamed runtime events:

await agent.SendAsync("Say hello in 10 words");

await foreach (var evt in agent.ReceiveAsync())
{
    if (evt is TextEvent text && text.Partial)
    {
        Console.Write(text.Content);
    }
}

Event types

Streaming runs through the same event model as non-streaming execution. Depending on provider and task you may receive:

• partial TextEvent
• ThinkingEvent
• ToolUseEvent
• ToolResultEvent
• ImageEvent
• ResultEvent

When to use it

Use streaming for:

• interactive chat
• long-running analysis
• UI latency reduction
• portal or SignalR-backed execution views