eurouter-integration

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

EUrouter Integration Guide

EUrouter集成指南

You are helping a developer integrate EUrouter into their application. EUrouter is a drop-in replacement for the OpenAI API, purpose-built for EU customers requiring GDPR compliance. It routes AI requests to 10 providers through a single OpenAI-compatible API at

https://api.eurouter.ai

Supported providers: OpenAI, Anthropic, Mistral, Scaleway (FR), GreenPT, Microsoft Foundry, Nebius, IONOS (DE), OVHcloud (FR), AWS Bedrock.

When helping developers, always load the relevant reference files from the

references/

directory in this skill folder based on what they need. Do not load all references at once.

您正在协助开发者将EUrouter集成到他们的应用程序中。EUrouter是OpenAI API的直接替代工具，专为需要GDPR合规的欧盟客户量身打造。它通过

https://api.eurouter.ai

这个统一的OpenAI兼容API，将AI请求路由至10家提供商。

支持的提供商： OpenAI、Anthropic、Mistral、Scaleway（法国）、GreenPT、Microsoft Foundry、Nebius、IONOS（德国）、OVHcloud（法国）、AWS Bedrock。

在协助开发者时，请根据他们的需求从当前技能文件夹的

references/

目录中加载相关参考文件，不要一次性加载所有参考文件。

Authentication

身份验证

EUrouter uses API keys in the format

eur_<publicId>.<secret>

Send the key via either header:

```
Authorization: Bearer eur_...
```
```
x-api-key: eur_...
```

For organization-scoped requests, add:

x-org-id: <uuid>

Keys are created via

POST /api/v1/keys

and the full key is shown only once at creation. It cannot be retrieved later.

EUrouter使用格式为

eur_<publicId>.<secret>

的API密钥。

可通过以下任一请求头发送密钥：

```
Authorization: Bearer eur_...
```
```
x-api-key: eur_...
```

若要发送组织范围的请求，请添加：

x-org-id: <uuid>

密钥通过

POST /api/v1/keys

创建，完整密钥仅在创建时显示一次，后续无法找回。

Drop-in Migration from OpenAI

从OpenAI无缝迁移

EUrouter is designed as a drop-in replacement. Only two things change:

Python (openai SDK):

python

from openai import OpenAI

EUrouter被设计为直接替代工具，仅需修改两处内容：

Python（openai SDK）：

python

from openai import OpenAI

Before (OpenAI direct)

之前（直接调用OpenAI）

client = OpenAI(api_key="sk-...")

After (EUrouter) — change base_url and api_key

之后（使用EUrouter）—— 修改base_url和api_key

client = OpenAI( base_url="https://api.eurouter.ai/api/v1", api_key="eur_YOUR_KEY_HERE" )

response = client.chat.completions.create( model="gpt-4o", messages=[{"role": "user", "content": "Hello!"}] )


**Node.js (openai SDK):**
```typescript
import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.eurouter.ai/api/v1",
  apiKey: "eur_YOUR_KEY_HERE",
});

const response = await client.chat.completions.create({
  model: "gpt-4o",
  messages: [{ role: "user", content: "Hello!" }],
});

Important: All existing OpenAI model names work as-is (e.g.,

gpt-4o

claude-3-5-sonnet

mistral-large

). No model name changes needed when migrating.

client = OpenAI( base_url="https://api.eurouter.ai/api/v1", api_key="eur_YOUR_KEY_HERE" )

response = client.chat.completions.create( model="gpt-4o", messages=[{"role": "user", "content": "Hello!"}] )


**Node.js（openai SDK）：**
```typescript
import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.eurouter.ai/api/v1",
  apiKey: "eur_YOUR_KEY_HERE",
});

const response = await client.chat.completions.create({
  model: "gpt-4o",
  messages: [{ role: "user", content: "Hello!" }],
});

重要提示： 所有现有OpenAI模型名称可直接使用（例如

gpt-4o

、

claude-3-5-sonnet

、

mistral-large

），迁移时无需修改模型名称。

Core Endpoints

核心端点

All endpoints are prefixed with

/api

. Base URL:

https://api.eurouter.ai

Endpoint	Method	Auth	Description
`/api/v1/chat/completions`	POST	Yes	Chat completions (streaming, tools, vision, audio, reasoning)
`/api/v1/completions`	POST	Yes	Legacy text completions
`/api/v1/embeddings`	POST	Yes	Text embeddings
`/api/v1/responses`	POST	Yes	OpenAI Responses API compatible
`/api/v1/models`	GET	No	List/filter model catalog
`/api/v1/providers`	GET	No	List available providers
`/api/v1/credits`	GET	Yes	Check credit balance (EUR)
`/api/v1/keys`	CRUD	Yes	Manage API keys
`/api/v1/routing-rules`	CRUD	Yes	Manage named routing configs
`/health`	GET	No	Health check

For full endpoint schemas, load

references/endpoints.md

所有端点均以

/api

为前缀，基础URL：

https://api.eurouter.ai

端点	请求方法	需要身份验证	描述
`/api/v1/chat/completions`	POST	是	聊天补全（流式传输、工具调用、视觉、音频、推理）
`/api/v1/completions`	POST	是	传统文本补全
`/api/v1/embeddings`	POST	是	文本嵌入
`/api/v1/responses`	POST	是	兼容OpenAI Responses API
`/api/v1/models`	GET	否	列出/筛选模型目录
`/api/v1/providers`	GET	否	列出可用提供商
`/api/v1/credits`	GET	是	查看信用余额（欧元）
`/api/v1/keys`	CRUD	是	管理API密钥
`/api/v1/routing-rules`	CRUD	是	管理命名路由配置
`/health`	GET	否	健康检查

如需完整的端点架构，请加载

references/endpoints.md

。

EU/GDPR Compliance

欧盟/GDPR合规

This is EUrouter's key differentiator. Add a

provider

object to any generation request to enforce compliance:

json

{
  "model": "gpt-4o",
  "messages": [{"role": "user", "content": "Hello"}],
  "provider": {
    "data_residency": "eu",
    "eu_owned": true,
    "data_collection": "deny",
    "max_retention_days": 0
  }
}

Field	Type	Description
`data_residency`	`string`	Region code — `"eu"` , `"de"` , `"fr"` , etc. Only endpoints in that region
`eu_owned`	`boolean`	Only EU-headquartered providers (Scaleway, IONOS, OVHcloud, GreenPT, Mistral)
`data_collection`	`"allow" \| "deny"`	`"deny"` excludes providers that use data for training
`max_retention_days`	`number`	Max data retention. `0` = zero-data-retention only

Ready-to-use compliance profiles:

json

// Strict GDPR: EU owned, no training data, zero retention, EU hosted
{ "eu_owned": true, "data_collection": "deny", "max_retention_days": 0, "data_residency": "eu" }

// EU hosted, cheapest price
{ "data_residency": "eu", "sort": "price" }

// Specific EU providers only
{ "only": ["scaleway", "ionos", "ovhcloud"] }

For all provider preference fields, load

references/provider-preferences.md

这是EUrouter的核心差异化优势。在任何生成请求中添加

provider

对象即可强制合规：

json

{
  "model": "gpt-4o",
  "messages": [{"role": "user", "content": "Hello"}],
  "provider": {
    "data_residency": "eu",
    "eu_owned": true,
    "data_collection": "deny",
    "max_retention_days": 0
  }
}

字段	类型	描述
`data_residency`	`string`	地区代码 — `"eu"` 、 `"de"` 、 `"fr"` 等，仅使用该地区的端点
`eu_owned`	`boolean`	仅选择欧盟总部的提供商（Scaleway、IONOS、OVHcloud、GreenPT、Mistral）
`data_collection`	`"allow" \| "deny"`	`"deny"` 会排除将数据用于训练的提供商
`max_retention_days`	`number`	最大数据保留天数， `0` 表示仅选择零数据保留的提供商

即用型合规配置文件：

json

// 严格GDPR合规：欧盟所有、不用于训练数据、零保留、欧盟托管
{ "eu_owned": true, "data_collection": "deny", "max_retention_days": 0, "data_residency": "eu" }

// 欧盟托管、价格最优
{ "data_residency": "eu", "sort": "price" }

// 仅指定欧盟提供商
{ "only": ["scaleway", "ionos", "ovhcloud"] }

如需查看所有提供商偏好字段，请加载

references/provider-preferences.md

。

Model Fallback

模型降级回退

Pass a

models

array to try multiple models in order. If all providers for the first model fail, EUrouter tries the next:

json

{
  "models": ["claude-3-5-sonnet", "gpt-4o", "mistral-large"],
  "messages": [{"role": "user", "content": "Hello"}],
  "provider": { "data_residency": "eu" }
}

Note: Model fallback is NOT supported for embeddings (vector dimensions differ between models). Only provider fallback (same model, different provider endpoint) works for embeddings.

传入

models

数组可按顺序尝试多个模型。如果第一个模型的所有提供商都请求失败，EUrouter会尝试下一个模型：

json

{
  "models": ["claude-3-5-sonnet", "gpt-4o", "mistral-large"],
  "messages": [{"role": "user", "content": "Hello"}],
  "provider": { "data_residency": "eu" }
}

注意： 模型降级回退不适用于嵌入（不同模型的向量维度不同），嵌入仅支持提供商降级回退（同一模型，不同提供商端点）。

Routing Rules

路由规则

Create reusable routing configurations to avoid repeating provider preferences on every request:

json

// Create a rule once:
// POST /api/v1/routing-rules
{
  "name": "gdpr-strict",
  "provider": {
    "eu_owned": true,
    "data_collection": "deny",
    "max_retention_days": 0,
    "data_residency": "eu"
  }
}

// Then reference it by name:
{
  "model": "gpt-4o",
  "rule_name": "gdpr-strict",
  "messages": [{"role": "user", "content": "Hello"}]
}

You can also reference by UUID with

rule_id

. Per-request

provider

preferences merge with the rule — request-level values take precedence.

Dry-run testing: Use

POST /api/v1/routing-rules/dry-run

to preview which providers satisfy a routing config without making a real request.

For full routing rules API, load

references/routing-rules.md

创建可复用的路由配置，避免在每个请求中重复设置提供商偏好：

json

// 一次性创建规则：
// POST /api/v1/routing-rules
{
  "name": "gdpr-strict",
  "provider": {
    "eu_owned": true,
    "data_collection": "deny",
    "max_retention_days": 0,
    "data_residency": "eu"
  }
}

// 之后通过名称引用：
{
  "model": "gpt-4o",
  "rule_name": "gdpr-strict",
  "messages": [{"role": "user", "content": "Hello"}]
}

您也可以通过UUID使用

rule_id

引用。请求级别的

provider

偏好会与规则合并，且请求级别的值优先级更高。

试运行测试： 使用

POST /api/v1/routing-rules/dry-run

可预览哪些提供商符合路由配置，无需发起真实请求。

如需完整的路由规则API，请加载

references/routing-rules.md

。

Model Discovery

模型发现

GET /api/v1/models
GET /api/v1/models?provider=scaleway
GET /api/v1/models?supported_parameters=tools,vision
GET /api/v1/models?category=embedding

Prices are returned as strings in the model's

pricing

object with a

currency

field. Credits are in EUR.

GET /api/v1/models
GET /api/v1/models?provider=scaleway
GET /api/v1/models?supported_parameters=tools,vision
GET /api/v1/models?category=embedding

价格会在模型的

pricing

对象中以字符串形式返回，并包含

currency

字段，信用额度以欧元为单位。

Response Headers

响应头

Every generation endpoint returns routing transparency headers:

Header	Description
`x-provider-id`	Internal provider UUID
`x-provider-slug`	Provider name (e.g., `openai` , `scaleway` )
`x-routing-strategy`	Strategy used for routing
`x-model-used`	Actual model slug used
`x-model-variant`	Variant suffix if applicable ( `:free` , `:nitro` )
`x-fallback-count`	Number of provider fallbacks (only present if > 0)
`x-model-fallback-count`	Number of model fallbacks (only present if > 0)

每个生成端点都会返回路由透明性头信息：

响应头	描述
`x-provider-id`	提供商内部UUID
`x-provider-slug`	提供商名称（例如 `openai` 、 `scaleway` ）
`x-routing-strategy`	所使用的路由策略
`x-model-used`	实际使用的模型标识
`x-model-variant`	模型变体后缀（如有，例如 `:free` 、 `:nitro` ）
`x-fallback-count`	提供商降级回退次数（仅当次数>0时返回）
`x-model-fallback-count`	模型降级回退次数（仅当次数>0时返回）

Streaming

流式传输

Set

"stream": true

for SSE streaming. Format is OpenAI-compatible:

data: {"id":"chatcmpl-...","object":"chat.completion.chunk","choices":[{"delta":{"content":"Hello"},"index":0,"finish_reason":null}]}

data: [DONE]

For usage stats in the final chunk, add

"stream_options": { "include_usage": true }

设置

"stream": true

即可启用SSE流式传输，格式与OpenAI兼容：

data: {"id":"chatcmpl-...","object":"chat.completion.chunk","choices":[{"delta":{"content":"Hello"},"index":0,"finish_reason":null}]}

data: [DONE]

如需在最终分块中返回使用统计信息，请添加

"stream_options": { "include_usage": true }

。

Error Format

错误格式

All errors follow:

json

{
  "error": {
    "code": "NOT_FOUND",
    "message": "Model not found"
  },
  "requestId": "uuid"
}

Key error codes:

UNAUTHORIZED

(401),

FORBIDDEN

(403),

NOT_FOUND

(404),

BAD_REQUEST

(400),

INSUFFICIENT_QUOTA

(402),

SERVICE_UNAVAILABLE

(503).

所有错误均遵循以下格式：

json

{
  "error": {
    "code": "NOT_FOUND",
    "message": "Model not found"
  },
  "requestId": "uuid"
}

主要错误代码：

UNAUTHORIZED

（401）、

FORBIDDEN

（403）、

NOT_FOUND

（404）、

BAD_REQUEST

（400）、

INSUFFICIENT_QUOTA

（402）、

SERVICE_UNAVAILABLE

（503）。

Reference Files

参考文件

Load these based on the developer's specific need:

Need	File
Full endpoint request/response schemas	`references/endpoints.md`
All provider routing preference fields	`references/provider-preferences.md`
Routing rules CRUD + dry-run API	`references/routing-rules.md`
Python code examples (openai SDK)	`references/code-examples-python.md`
Node.js/TypeScript code examples	`references/code-examples-node.md`
curl examples	`references/code-examples-curl.md`

根据开发者的具体需求加载以下文件：

需求	文件
完整的端点请求/响应架构	`references/endpoints.md`
所有提供商路由偏好字段	`references/provider-preferences.md`
路由规则CRUD及试运行API	`references/routing-rules.md`
Python代码示例（openai SDK）	`references/code-examples-python.md`
Node.js/TypeScript代码示例	`references/code-examples-node.md`
curl示例	`references/code-examples-curl.md`