Metadata-Version: 2.4
Name: openai-router
Version: 0.1.2
Summary: openai-router
Requires-Python: >=3.11
Description-Content-Type: text/markdown
License-File: LICENSE.txt
Requires-Dist: fastapi>=0.120.3
Requires-Dist: gradio>=5.49.1
Requires-Dist: httpx>=0.28.1
Requires-Dist: loguru>=0.7.3
Requires-Dist: sqlmodel>=0.0.27
Requires-Dist: uvicorn>=0.38.0
Dynamic: license-file


<!-- markdownlint-disable MD033 -->
<h1 align="center">
🚀 OpenAI Router 
</h1>
 
<p align="center">
<b>轻量级、持久化、零配置的 OpenAI API 统一网关</b><br>
一键聚合 vLLM、SGLang、lmdeploy、Ollama…  
</p>
 
<p align="center">
<a href="#"><img src="https://img.shields.io/badge/license-MIT-blue?style=flat-square"></a>
<a href="https://fastapi.tiangolo.com"><img src="https://img.shields.io/badge/FastAPI-v0.115+-teal?style=flat-square"></a>
<a href="https://gradio.app"><img src="https://img.shields.io/badge/Gradio-v5+-orange?style=flat-square"></a>
<a href="#"><img src="https://img.shields.io/badge/SQLite-内置存储-lightgrey?style=flat-square"></a>
<a href="https://pypi.org/project/openai-router/"><img  src="https://img.shields.io/pypi/v/openai-router?style=flat-square&logo=pypi&label=PyPI"></a> 
</p>
 
---
- 将不同推理框架（vLLM、SGLang、lmdeploy、Ollama… ）、不同`Host`、不同`Port`的`OpenAI` API接口统一聚合到同一个`base_url`上，实现更便捷的模型调用。
## ✨ Features 
| Feature       | Description                                                               |
| ------------- | ------------------------------------------------------------------------- |
| 🌍 统一入口    | `/chat/completions`、`/embeddings`、`/images/generations`… 全部转发       |
| 🧩 多后端      | vLLM、SGLang、lmdeploy、Ollama… 任意组合                                  |
| 💾 持久化      | SQLite + SQLModel 零配置存储路由                                          |
| ⚡ 实时流      | SSE & Chunked Transfer 全双工支持                                         |
| 🎨 Web UI      | Gradio 即用的管理面板                                                     |
| 🔍 兼容 OpenAI | SDK / LangChain / AutoGen / LlamaIndex / CrewAI  …等 **一行代码都不用改** |
 
---
 
## 📦 Quick Start 
### Step-1：安装 

#### PyPI（推荐）

```bash 
uv add openai-router -U
```
或者
```bash 
pip install openai-router -U
```


 
### Step-2：启动 
```bash 
openai-router --host localhost --port 8000 
```
浏览器自动打开  
📍 UI：`http://localhost:8000`  
📍 API 文档：`http://localhost:8000/docs`
 
### Step-3：添加后端样例
在 Web UI 「添加 / 更新」填入：
- 模型名：`gpt-4`
- 后端 URL：`http://localhost:8082/v1`
- 后端 API 密钥 (可选) ：
如果提供，路由器将使用此密钥覆盖原始请求中的 Authorization 标头。如果留空，将透传原始请求的密钥。

<img src="static/ui.png" width="800">
 
---
 
## 🔧 API Usage 
### **像官方 OpenAI SDK一样调用**
```python 
from openai import OpenAI 
client = OpenAI(
      base_url="http://localhost:8000/v1",
      api_key="sk-dummy"
)
resp = client.chat.completions.create(
      model="gpt-4",
      messages=[{"role":"user","content":"hello"}],
      stream=True 
)
for chunk in resp:
      print(chunk.choices[0].delta.content or "", end="")
```
 
cURL 
```bash 
curl http://localhost:8000/v1/chat/completions \
     -H "Content-Type: application/json" \
     -d '{"model":"gpt-4","messages":[{"role":"user","content":"hi"}],"stream":true}'
```
 
---
 
## 🗂️ Endpoints 
| Method | Path                       | Description                    |
| ------ | -------------------------- | ------------------------------ |
| `GET`  | `/`                        | Gradio Admin UI                |
| `GET`  | `/docs`                    | OpenAPI Swagger                |
| `GET`  | `/v1/models`               | List available models          |
| `POST` | `/v1/responses`            | Responses API                  |
| `POST` | `/v1/chat/completions`     | Chat completion                |
| `POST` | `/v1/embeddings`           | Text embeddings                |
| `POST` | `/v1/images/generations`   | DALL·E style                   |
| `POST` | `/v1/audio/transcriptions` | Whisper                        |
| …      | …                          | All OpenAI endpoints supported |
 
---
 
## ⚙️ Configuration 
CLI Options 
```bash 
openai-router --help 
```
| Flag     | Default     | Description  |
| -------- | ----------- | ------------ |
| `--host` | `localhost` | Bind address |
| `--port` | `8000`      | Bind port    |
 

---
 
## 🏗️ Architecture 
<img src="static/arch.png" width="800">
