Table of Contents

👉 Try the demo: https://github.com/LM-Kit/lm-kit-net-samples/tree/main/console_net/ai-agents/streaming/agent_streaming

Streaming Agent Responses for C# .NET Applications


🎯 Purpose of the Demo

An interactive console app that shows three streaming patterns on the same agent: pull-based IAsyncEnumerable<AgentStreamToken>, push-based IAgentStreamHandler, and a streaming-to-file recipe.

All inference runs on-device.


👥 Industry Target Audience

  • App developers building chat UIs that need responsive token streaming.
  • CLI tooling that streams long outputs incrementally.
  • Pipelines that consume agent output as a stream (e.g., feed into a UI websocket).
  • Logging / audit that wants per-token capture to disk.

🚀 Problem Solved

A chat UI that waits for full inference is a dead chat UI. Streaming tokens, reasoning, and tool calls are how production apps feel responsive. The demo pairs the pull-based IAsyncEnumerable API with the push-based callback handler so the right pattern for the right surface is one menu pick away.


💻 Application Overview

Interactive menu (no command-line arguments) with three modes:

Mode What it does
Stream IAsyncEnumerable<AgentStreamToken> with typed token coloring (Content, Thinking, ToolCall, ToolResult).
Callback Push-based pattern using the bundled DelegateStreamHandler.Console(). Reports status and duration.
Save Stream tokens simultaneously to console and a UTF-8 text file.
Quit Exit.

✨ Key Features

  • StreamingAgentExecutor with BufferSize, StreamThinking, StreamToolCalls toggles.
  • StreamAsync(agent, input) for pull-based await foreach.
  • ExecuteStreamingAsync(agent, input, IAgentStreamHandler) for push-based callbacks.
  • AgentStreamToken.Type enum for fine-grained token typing.

🧠 Model

  • qwen3.5:4b (current Qwen 3.5 dense).

🛠️ Getting Started

📋 Prerequisites

  • .NET 8.0 or later

▶️ Running the Application

git clone https://github.com/LM-Kit/lm-kit-net-samples
cd lm-kit-net-samples/console_net/ai-agents/streaming/agent_streaming
dotnet run

Pick a mode from the menu.

📚 Related Content

Share