在 Cherry Studio 中使用 oairoute
- 桌面端的多模型客户端，支持自定义 API 端点
- 配置一次即可用 OAIRoute 的所有模型
- 支持多 Key 切换与对话历史本地保存
第一步：获取 API Key
在 OAIRoute 控制台 创建 Key，复制备用。
Create a key in the OAIRoute console and copy it for later.
第二步：添加服务商
打开 Cherry Studio，进入 设置 → 模型服务，点底部的 添加，填写：
自定义，建议填 OAIRoute，便于和其他服务商区分。
选 OpenAI。OAIRoute 的文本接口与 OpenAI 结构完全兼容。
填 https://api.oairoute.com/v1。注意结尾的 /v1 不能省。
填上一步复制的 Key。
Open Cherry Studio, go to Settings → Model services, click Add at the bottom, and fill in:
Any name you like; OAIRoute is a good choice so it's easy to tell apart from other providers.
Choose OpenAI. OAIRoute's text endpoints are fully compatible with the OpenAI schema.
Enter https://api.oairoute.com/v1. Don't drop the trailing /v1.
Paste the key you copied in the previous step.
第三步：添加模型
在同一页面点 管理，手动添加模型 ID。常用的几个：
| 模型 ID | 用途 |
|---|---|
gpt-4o-mini | 日常对话，成本低 |
claude-sonnet-5-5 | 长文写作与代码 |
gemini-2.5-pro | 超长上下文 |
deepseek-chat | 中文场景 |
模型 ID 必须与 OAIRoute 返回的完全一致。可以调用 GET /v1/models 拿到完整清单，不要凭印象填。
On the same page click Manage and add model IDs manually. Some common ones:
| Model ID | Use |
|---|---|
gpt-4o-mini | Everyday chat, low cost |
claude-sonnet-5-5 | Long-form writing and code |
gemini-2.5-pro | Very long context |
deepseek-chat | Chinese-language scenarios |
Model IDs must match exactly what OAIRoute returns. Call GET /v1/models for the full list; don't type from memory.
第四步：验证
回到主界面，新建对话，在顶部的模型下拉里选中刚添加的模型，发一条消息。能正常回复即配置成功。
Back in the main window, start a new conversation, select the model you just added from the dropdown at the top, and send a message. A normal reply means it's configured correctly.
图像生成
图像接口走 /v1/images/generations，但域名取决于模型：
- GPT-Image 等非火山模型走
https://api.oairoute.com/v1，与文本共用同一个服务商条目即可 - Seedream 与 Seedance 系列走
https://ark.oairoute.com/v1，需要单独再建一个服务商条目，并到 ark 站 获取 Key
两站的 Key 不通用。只用 GPT-Image 的话不必建第二个条目。
Image requests go to /v1/images/generations, but the domain depends on the model:
- Non-Volcengine models such as GPT-Image use
https://api.oairoute.com/v1and can share the same provider entry as text - The Seedream and Seedance series use
https://ark.oairoute.com/v1; create a second provider entry and get a key from the ark site
Keys aren't shared between the two sites. If you only use GPT-Image, you don't need a second entry.
常见问题
回复为空或一直转圈
多数是 API 地址少了 /v1，或者多了结尾斜杠。正确写法是 https://api.oairoute.com/v1。
提示 401
Key 复制时带了空格，或者用了已经删除的 Key。重新生成一个再试。
提示 404 model not found
模型 ID 写错了。用模型列表接口核对，注意大小写和连字符。
对话很慢
换用更小的模型，或在设置里把「上下文消息数」调小。上下文越长，首 token 延迟越明显。
Empty replies, or a spinner that never stops
Usually the API URL is missing /v1 or has an extra trailing slash. The correct form is https://api.oairoute.com/v1.
401 error
The key was copied with a trailing space, or it's a deleted key. Generate a new one and retry.
404 model not found
The model ID is wrong. Check it against the models list endpoint, paying attention to case and hyphens.
Slow conversations
Switch to a smaller model, or lower "context messages" in settings. The longer the context, the more noticeable the first-token delay.