[Feature Request] 渠道测试支持 Audio Transcription 端点类型（gpt-4o-transcribe 等模型）

Read This First (Do Not Remove This Section)

• Docs: https://docs.newapi.ai/
• Usage questions first: https://deepwiki.com/QuantumNous/new-api
• Warning: issues with this template removed, section headings deleted, or content cleared may be closed directly. Repeated abusive submissions may result in a block.

Your current newapi version

v1.0.0-rc.10-101aix.5 (also tested on latest upstream)

Submission Checks

• No duplicate issue: I have searched existing Issues and confirmed there are no duplicates.
• Pre-submission reading: I have read the docs at https://docs.newapi.ai/ and the project README.
• Template intact: I have not removed any guidance or section headings from this template.
• Maintainer cost: I understand maintainers have limited time and issues not following the template may be ignored or closed directly.

Feature Description

问题

当前渠道测试（controller/channel-test.go → buildTestRequest）不支持 Audio Transcription 类模型（如 gpt-4o-transcribe、gpt-4o-transcribe-diarize、gpt-4o-mini-transcribe）。

测试这些模型时，会默认构建一个 Chat Completion 请求发送到 /v1/chat/completions，导致上游返回 500 server_error：

bad response status code 500, message: The server had an error while processing your request.
Sorry about that!, body: { "error": { "message": "The server had an error while processing
your request. Sorry about that!", "type": "server_error", "param": null, "code": null } }

实际上模型和渠道配置都没问题——用 /v1/audio/transcriptions 发送音频文件可以正常返回结果。

根因分析

1. constant/endpoint_type.go 中没有定义 Audio Transcription 端点类型
2. normalizeChannelTestEndpoint() 无法自动识别 transcribe 模型名
3. buildTestRequest() 的 switch 没有 audio transcription 分支，无法构建 multipart/form-data 音频请求

建议方案

方案 A（推荐）：新增 EndpointTypeAudioTranscription 端点类型

// constant/endpoint_type.go
EndpointTypeAudioTranscription EndpointType = "audio-transcription"

// controller/channel-test.go — normalizeChannelTestEndpoint()
// 自动检测 transcribe 模型名
if strings.Contains(modelName, "transcribe") || strings.Contains(modelName, "whisper") {
    return string(constant.EndpointTypeAudioTranscription)
}

// controller/channel-test.go — buildTestRequest()
case constant.EndpointTypeAudioTranscription:
    // 生成最小 WAV（16kHz mono, ~0.5s 静音）
    // 构建 multipart/form-data 请求体
    return &AudioTranscriptionTestRequest{Model: model, File: minimalWAV}

• 前端测试弹窗的「端点类型」下拉菜单中增加 audio-transcription 选项
• 在 normalizeChannelTestEndpoint 中对 transcribe/whisper 等模型名自动推断

方案 B（最小改动）：仅自动检测 + 跳过

在 normalizeChannelTestEndpoint 中识别 transcribe 模型，直接返回提示"该模型类型不支持渠道测试"，避免误报 500。

影响的模型

Model	Endpoint
gpt-4o-transcribe	/v1/audio/transcriptions
gpt-4o-transcribe-diarize	/v1/audio/transcriptions
gpt-4o-mini-transcribe	/v1/audio/transcriptions
whisper-1	/v1/audio/transcriptions

Use Case

运营人员在渠道管理界面点击「测试」按钮时，期望能验证 transcribe 模型的可用性，目前只能通过手动 curl 发送音频文件来测试，无法在 UI 中操作。这与 #4281 (Anthropic 端点检测) 和 #3298 (流式测试硬编码) 属于同一类问题——渠道测试未覆盖所有端点类型。

Read constant/endpoint_type.go and controller/channel-test.go, starting at normalizeChannelTestEndpoint and buildTestRequest. Compare the existing channel-test request paths, then implement the chosen approach so transcribe and whisper models are recognized and either tested through /v1/audio/transcriptions or reported as unsupported. Confirm the endpoint-type option matches if the UI change is included.