Nettopo Switch (Codex ↔ LLM Local Proxy)

中文

Overview

Nettopo Switch is a local proxy that adapts Codex-compatible POST /v1/responses (including SSE streaming) and forwards requests to various LLM providers' POST /v1/chat/completions. It provides a desktop UI to manage endpoints, API keys, ports, and model mappings across multiple providers, plus tools to manage Codex config.toml with safe backups.

This is designed to work with Codex Desktop by pointing Codex's Base URL to the local proxy, so Codex can use different LLM models without changing its workflow.

Features

• Codex Responses adapter: supports POST /v1/responses (SSE) and forwards to provider POST /v1/chat/completions
• Multi-provider support: switch between 9+ built-in LLM providers, or add custom ones
• Model mapping: maps Codex-side model names (e.g. gpt-5.4-mini) to provider-specific models
• Menu bar balance display: shows real-time available usage balance in the system tray (for supported providers like DeepSeek)
• Visual configuration: configure Base URL, API key, port, and mappings in the desktop app
• Codex config.toml management: merge-write, edit raw content, create/restore/delete/clean backup history
• Health check & logs: one-click upstream connectivity check; logs for each request
• UI i18n: zh-CN (简体中文), en-US (English), ja-JP (日本語), ko-KR (한국어), fr-FR (Français), de-DE (Deutsch), es-ES (Español)
• Cross-platform builds: macOS arm64 / Windows amd64 / Windows arm64

Supported Providers

Provider	Default Base URL	Default Model
DeepSeek	https://api.deepseek.com/v1	deepseek-v4-flash
阿里通义千问 (Alibaba)	https://dashscope.aliyuncs.com/compatible-mode/v1	qwen3.6-plus
小米 MiMo (Xiaomi)	https://api.xiaomimimo.com/v1	mimo-v2.5-pro
智谱 GLM (Zhipu)	https://open.bigmodel.cn/api/paas/v4	glm-4.7-flash
百度千帆 (Baidu)	https://qianfan.baidubce.com/v2	ernie-5.1
火山引擎豆包 (Volcano)	https://ark.cn-beijing.volces.com/api/v3	doubao-seed-2-0-lite-260215
腾讯混元 (Tencent)	https://api.hunyuan.cloud.tencent.com/v1	hunyuan-2.0-thinking-20251109
硅基流动 (SiliconFlow)	https://api.siliconflow.cn/v1	deepseek-ai/DeepSeek-V4-Flash
Kimi (Moonshot)	https://api.moonshot.cn/v1	kimi-k2.6
Custom	User-defined	User-defined

Endpoints

• GET /: service info
• GET /health: health status
• GET /v1/models: model list (for Codex UI model selection)
• POST /v1/responses: Codex entrypoint (recommended)
• POST /v1/chat/completions: compatibility endpoint

Quick Start

1. In the desktop app, select a provider and configure:

   • Base URL & API Key for the chosen provider
   • Default model

2. Start the proxy service (default listen: http://127.0.0.1:11434)

3. Verify:

curl http://127.0.0.1:11434/health

Configure Codex

In the app: Preferences → Codex config.toml

• Merge write: updates ~/.codex/config.toml while preserving other existing settings
• Backups: each write creates a non-overwriting backup under ~/.codex/backups/ which can be restored or deleted

Codex Base URL should point to:

http://127.0.0.1:11434/v1

Menu Bar Balance

On macOS, the system tray icon displays the available usage balance (e.g. 12.34 CNY) for providers that support balance checking (currently DeepSeek). The balance refreshes every 60 seconds.

FAQ

Codex shows 502 / Reconnecting

• Check “Recent Logs” in the app for upstream 4xx/5xx or forwarding failures
• If the provider does not recognize a model name, update “Model Mapping” to map Codex models to the correct provider-specific model names

How do I add a custom provider?

Use the “Custom” provider type in the app and fill in your own Base URL, API key, and model mappings.

Development & Build

Local development

npm -C frontend install
go install github.com/wailsapp/wails/v2/cmd/wails@v2.12.0
wails dev

Build (example)

export PATH=”$(go env GOPATH)/bin:$PATH”
wails build -platform darwin/arm64