安装使用

<br>

直接安装

从 Releases 下载安装包，安装后打开。

<br>

Install

Download the installer from Releases.

<br>

Build from Source

Prerequisites: Node.js v18+, Rust, Visual Studio Build Tools (Windows, "Desktop development" workload)

git clone https://github.com/QinMoXX/AAStation.git
cd AAStation
npm install
npm run tauri dev    # Dev mode
npm run tauri build  # Build installer

<br>

操作教程

完整链路示例

一个综合了以上所有元素的完整链路：

---

<a name="english"></a>

<br>

Tutorial

Scenario 1: Direct Connection (App → Provider)

The simplest setup: route your AI client through AAStation to a single provider.

1. Add nodes

Click the + button in the toolbar to add a provider (e.g. DeepSeek) and an application node. Fill in your API Key.

2. Connect

Drag a line from the application node's right handle to the provider node's left handle.

3. Configure the client

Go to Settings → Apps, select the client type and application node, then click "Write Config".

Your existing config is automatically backed up before writing.

4. Start the proxy

Click "Start Proxy" on the main interface. The status bar will show the proxy is running.

Done! Your AI client now uses the provider through AAStation.

<br>

Scenario 2: Model-Based Routing

Route different models to different providers using a Switcher node.

Pipeline: App → Switcher → Multiple Providers

1. Add a Switcher node between the app and multiple providers 2. Configure routing rules in the side panel, e.g.: - Model claude-sonnet → DeepSeek - Model claude-opus → Zhipu AI - Others → default provider

Match by model name, path prefix, or HTTP header.

<br>

Scenario 3: Smart Failover

Have multiple provider accounts? Use a Poller node for automatic selection and failover.

Pipeline: App → Poller → Multiple Providers

1. Add a Poller node between the app and multiple providers 2. Choose a strategy in the side panel: - Weight — Distribute requests by ratio - Health First — Prefer the fastest-responding provider - Remaining Quota — Prefer providers with the most tokens left 3. (Optional) Set daily/monthly token budgets on provider nodes

When a provider fails or runs out of quota, requests automatically fall over to the next one.

<br>

Scenario 4: Real-Time Monitoring

Dashboard

Switch to the Monitor tab to see request count, token usage, latency, and success rate in real time.

Floating Message Window

Enable "Floating Message Window" in Settings → General — a translucent overlay shows live request/response content with typewriter animation.

<br>

Full Pipeline Example

A complete pipeline combining all the elements above:

常用客户端一键配置

添加Claude Code、OpenCod、Codex CLI 应用节点后，设置界面点击"写入配置" 一键配置。

Codex CLI 注意：AAStation 不支持修改旧版 Codex 配置；新版 Codex 改用 Responses API，兼容性有限。建议优先使用 Claude Code 或 OpenCode。（详见 Codex 官方说明）。

<br>

场景二：按模型分流

想让不同模型走不同供应商？在中间加一个路由器节点。

链路结构： 应用 → 路由器 → 多个供应商

1. 在应用和供应商之间加入「路由器」节点，连接多个供应商 2. 点击路由器节点，在右侧面板添加路由规则，例如： - 模型名 claude-sonnet → 走 DeepSeek - 模型名 claude-opus → 走智谱 AI - 其他模型 → 走默认供应商

支持按模型名、路径前缀、请求头三种条件匹配。

<br>