What is this?
One Switch is a local AI API gateway: point your AI client at one local address, and it automatically routes to and fails over between the LLM providers/models you configure behind it.
One Switch is a local AI API gateway: point your AI client at one local address, and it automatically routes to and fails over between the LLM providers/models you configure behind it.Is this a cloud service or something I run on my own machine?It's a native Electron desktop app for macOS, Windows and Linux that runs a local proxy server on 127.0.0.1; there is no cloud component.
It's a native Electron desktop app for macOS, Windows and Linux that runs a local proxy server on 127.0.0.1; there is no cloud component.How do I get it running?Download the installer for your OS from GitHub Releases (.dmg for macOS Apple Silicon/Intel, .exe for Windows x64/ARM64, .AppImage for Linux), then add a provider with its API key in Model Management, add upstream models under it, and set their order in the Model Queue.
Download the installer for your OS from GitHub Releases (.dmg for macOS Apple Silicon/Intel, .exe for Windows x64/ARM64, .AppImage for Linux), then add a provider with its API key in Model Management, add upstream models under it, and set their order in the Model Queue.Do I need my own API keys from each provider?Yes, you supply your own API key per provider you add (OpenAI, DeepSeek, OpenRouter, Ollama, Anthropic-compatible services, etc.). Your AI client itself doesn't need a real key since One Switch injects the actual upstream key.
Yes, you supply your own API key per provider you add (OpenAI, DeepSeek, OpenRouter, Ollama, Anthropic-compatible services, etc.). Your AI client itself doesn't need a real key since One Switch injects the actual upstream key.Any install pitfalls on macOS?The macOS build is currently only ad-hoc signed and not notarized, so auto-update signature verification is disabled and the first launch may be blocked; you need to allow it via System Settings or right-click > Open in Finder.
The macOS build is currently only ad-hoc signed and not notarized, so auto-update signature verification is disabled and the first launch may be blocked; you need to allow it via System Settings or right-click > Open in Finder.What happens if the connection drops mid-stream?One Switch only switches upstreams before any response has been sent to the client. If a streaming response has already started and then disconnects, the request is terminated and logged as a failure — it will never splice output from a different model into the same reply.
One Switch only switches upstreams before any response has been sent to the client. If a streaming response has already started and then disconnects, the request is terminated and logged as a failure — it will never splice output from a different model into the same reply.Does it support Gemini?Not yet. Right now it only speaks OpenAI Chat Completions, OpenAI Responses, and Anthropic Messages; Gemini and custom protocols are listed as planned but unavailable in the current version.
Not yet. Right now it only speaks OpenAI Chat Completions, OpenAI Responses, and Anthropic Messages; Gemini and custom protocols are listed as planned but unavailable in the current version.Can I fix provider quirks without writing code?Yes — the Request Rewrite feature lets you add, remove or edit headers and JSON body fields (with regex support) per provider/model via configurable, orderable rules, with no scripting involved.
Yes — the Request Rewrite feature lets you add, remove or edit headers and JSON body fields (with regex support) per provider/model via configurable, orderable rules, with no scripting involved.Can I manually pick which model handles the next request?Yes, you can drag-reorder the queue or manually set a preferred model at any time; the switch only affects new requests and won't interrupt ones already in flight, and automatic failover still kicks in if your pick fails.
Yes, you can drag-reorder the queue or manually set a preferred model at any time; the switch only affects new requests and won't interrupt ones already in flight, and automatic failover still kicks in if your pick fails.What are similar projects?casbin-gateway, freellmapi, ThinkWatch-Lite, open-codesign, opencodex, magpie, evenfire, Kun
casbin-gateway, freellmapi, ThinkWatch-Lite, open-codesign, opencodex, magpie, evenfire, KunWhich topics is it in?llm-gateway, router, openrouter, vite, openai-compatible, electron, deepseek, sqlite, proxy, local-first, codex, gui
llm-gateway, router, openrouter, vite, openai-compatible, electron, deepseek, sqlite, proxy, local-first, codex, gui