makecindy / makecindy/cindy

手机端远程控制时 deepseek-v4-flash 模型名未补全 provider 前缀导致报错

Open
#2,245 2 comments 0 reactions 0 assignees View on GitHub
bug
Dominant language
TypeScript
Stars
2.7k
Forks
395
Avg merge
21h 48m
Merged PRs (30d)
776

Description

**提交人**: 匿名
**客户端版本**: 0.1.38

---

## 现象

手机端远程控制电脑上的 Codex,选择 `deepseek-v4-flash` 模型时,请求报错:

```
{"error":{"message":"/responses: Invalid model name passed in model=deepseek-v4-flash. Call `/v1/models` to view available models for your key.","type":"None","param":"None","code":"400"}}
```

电脑端本地使用同一个模型一切正常。

## 复现步骤

1. 电脑端 Cindy 正常使用 deepseek-v4-flash 模型(电脑端本地正常)
2. 手机端远程连接控制这台电脑,选择同样的 deepseek-v4-flash 模型
3. 发送任意消息,请求失败并报上述错误

## 实际行为

手机端远程控制时,把裸名 `deepseek-v4-flash` 直接传给网关,网关不认识这个名字,返回 400。

## 根因

网关(laxarouter)上该模型的完整名称是 `deepseek/deepseek-v4-flash`,带 `deepseek/` provider 前缀。

- 电脑端本地运行时,Cindy 会自动把模型名补全为 `deepseek/deepseek-v4-flash`,请求成功。
- 手机端远程控制时,没有做这层前缀补全,直接把裸名透传给网关。

用网关 key 直接请求 `/v1/responses` 验证:传 `deepseek-v4-flash` 复现完全相同的 400 报错;传 `deepseek/deepseek-v4-flash` 正常返回。`/v1/chat/completions` 同理。

## 期望行为

手机端远程控制时应与电脑端本地行为一致,对需要 provider 前缀的模型名做自动补全,而不是把裸名直接透传给网关。
---
**OS**: darwin arm64 (25.2.0)
**界面语言**: zh-CN

Contributor guide

Open the contributing guide

Research direction

Start by tracing the mobile remote-control request path and model selection handling, comparing it with the desktop local path that normalizes provider names. Check requests to /v1/responses and /v1/chat/completions; done means a selected deepseek-v4-flash model is sent with the required provider prefix and both request paths no longer return the reported 400 error.

Written by the indexing model from the issue text.

Assessment

Tech stack
react-native, typescript
Domain
api, mobile
Issue type
Bug
Difficulty
3/5
Estimated time
1-2 days
Activity status
Active
Clarity
Mostly clear
Newbie friendliness
68/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.