# AIDevProxy

**Repository Path**: jiehanlin/aidev-proxy

## Basic Information

- **Project Name**: AIDevProxy
- **Description**: AIProxy 是一个轻量级 ASP.NET Core LLM API 反向代理。为使用llm服务的开发提供透明的llm请求响应分析。
- **Primary Language**: Unknown
- **License**: MIT
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2026-05-24
- **Last Updated**: 2026-05-24

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# AIProxy

A lightweight ASP.NET Core reverse proxy for LLM (Large Language Model) APIs. Routes requests to multiple AI service backends, captures full request/response logs, and supports SSE (Server-Sent Events) streaming.

## Features

- **Multi-Backend Routing** — Route requests to different LLM providers (OpenAI, Azure OpenAI, DeepSeek, Qwen, etc.) based on URL path prefix.
- **Request/Response Logging** — Full request and response bodies are captured and saved as structured JSON files, organized by date.
- **SSE Streaming Support** — Transparently proxies Server-Sent Events (e.g., `text/event-stream`) from backends while collecting stream content.
- **Real-time Console Output** — Color-coded console summaries for each proxied request (`green` for success, `yellow` for 4xx, `red` for 5xx).
- **Built-in Help Page** — A styled HTML page listing all configured backends when accessing the root URL.
- **Zero External Dependencies** — Built on ASP.NET Core only; no third-party NuGet packages required.

## Prerequisites

- [.NET SDK 10.0](https://dotnet.microsoft.com/download/dotnet/10.0) or later

## Quick Start

### 1. Clone the repository

```bash
git clone <repo-url>
cd AIProxy
```

### 2. Configure backends

Edit `src/AIProxy/appsettings.json` under the `Proxy` section:

```json
{
  "Proxy": {
    "Port": 5000,
    "LogDirectory": "logs",
    "MaxBodySizeBytes": 10485760,
    "Backends": [
      {
        "Name": "openai",
        "PathPrefix": "/openai/",
        "BaseUrl": "https://api.openai.com",
        "ApiKey": "sk-your-openai-api-key",
        "Headers": {},
        "TimeoutSeconds": 120
      },
      {
        "Name": "deepseek",
        "PathPrefix": "/deepseek/",
        "BaseUrl": "https://api.deepseek.com",
        "ApiKey": "sk-your-deepseek-api-key",
        "Headers": {},
        "TimeoutSeconds": 120
      }
    ]
  }
}
```

### 3. Run the proxy

```bash
cd src/AIProxy
dotnet run
```

The proxy starts on `http://localhost:5000`. Access `http://localhost:5000/` in your browser to see the help page.

### 4. Use in your LLM client

Point your LLM client's base URL to the proxy with the corresponding path prefix:

| Backend | Client Base URL |
|---------|----------------|
| OpenAI | `http://localhost:5000/openai/v1` |
| DeepSeek | `http://localhost:5000/deepseek/v1` |
| Azure OpenAI | `http://localhost:5000/azure/` |

For example, with `curl`:

```bash
curl http://localhost:5000/openai/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-4o-mini",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'
```

## Configuration Reference

| Key | Type | Default | Description |
|-----|------|---------|-------------|
| `Proxy.Port` | `int` | `5000` | Port the proxy listens on |
| `Proxy.LogDirectory` | `string` | `"logs"` | Directory for log files (organized by date) |
| `Proxy.MaxBodySizeBytes` | `int` | `10485760` | Max bytes to capture per request/response body |
| `Proxy.Backends` | `array` | — | List of backend LLM service definitions |

### Backend Definition

| Key | Type | Description |
|-----|------|-------------|
| `Name` | `string` | Backend identifier (e.g., `"openai"`, `"deepseek"`) |
| `PathPrefix` | `string` | URL path prefix for routing (e.g., `"/openai/"`) |
| `BaseUrl` | `string` | Base URL of the upstream LLM API |
| `ApiKey` | `string` | API key sent as `Bearer` token in `Authorization` header |
| `Headers` | `dict` | Additional HTTP headers to forward to the backend |
| `TimeoutSeconds` | `int` | Request timeout in seconds (default: `120`) |

## Logs

Requests and responses are logged as JSON files under the configured `LogDirectory`:

```
logs/
└── 2026-05-24/
    ├── 143025123_POST_v1_chat_completions.json
    └── 143530456_POST_v1_completions.json
```

Each log file contains:

```json
{
  "SessionId": "a1b2c3d4e5f6",
  "Timestamp": "2026-05-24T14:30:25Z",
  "BackendName": "openai",
  "BackendUrl": "https://api.openai.com/v1/chat/completions",
  "Request": {
    "Method": "POST",
    "Path": "/v1/chat/completions",
    "Headers": { "Content-Type": "application/json" },
    "Body": "{ \"model\": \"gpt-4o-mini\", ... }",
    "BodyTruncated": false
  },
  "Response": {
    "StatusCode": 200,
    "Headers": { "Content-Type": "application/json" },
    "Body": "{ \"choices\": [...] }",
    "BodyTruncated": false,
    "IsStreaming": false,
    "StreamChunkCount": 0
  },
  "DurationMs": 1234,
  "Error": null
}
```

## Project Structure

```
src/AIProxy/
├── Models/
│   ├── ProxyConfig.cs           # Global proxy configuration model
│   ├── BackendDefinition.cs     # Backend LLM service definition
│   └── LogEntry.cs              # Request-response log entry models
├── Middleware/
│   └── ProxyRouterMiddleware.cs # Core: path matching, forwarding, error handling
├── Handlers/
│   ├── CaptureDelegatingHandler.cs  # HttpClient handler for capturing bodies
│   └── SseStreamCopier.cs           # SSE stream forwarding + collection
├── Services/
│   ├── SessionLogger.cs         # JSON log file writer
│   └── ConsoleReporter.cs       # Color-coded console output
├── Extensions/
│   └── ServiceCollectionExtensions.cs  # DI registration
├── Program.cs                   # Application entry point
└── appsettings.json             # Configuration
```

## License

This project is open source. See the license file for details.