omgvibe

What omgvibe sets up

omgvibe is an interactive NPX wizard that links coding agents such as CodeX, Claude Code, OpenCode, and CodeWhale to OhMyGPT's unified /v1 API. The problem it solves is the repetitive, error prone setup of API tokens, endpoints, and model catalogs for each of those tools. Rather than editing configuration files by hand, a user runs a single command and answers a few prompts. The wizard is zero install because it runs through npx omgvibe, so there is no global package to manage before using it. It supports a guided flow in English, simplified Chinese, and Japanese, with automatic locale detection so the prompts appear in the user's language. The default target is CodeX, while Claude Code, OpenCode, and CodeWhale are one selection away, which reflects where the author expects most users to start. For each chosen tool the wizard re runs the official npm install -g command so the agent binary is present and current. With 25 stars and a TypeScript codebase, the project is early but focused on a concrete daily annoyance for people who switch between AI coding assistants or who set up fresh machines often. The end result is a correctly written configuration file with the right provider, base URL, and models selected.

The guided configuration flow

The quick start is straightforward. After confirming Node.js 18 or newer and npm are installed, the user runs npx omgvibe and picks a language, then selects the CLI to configure, with CodeX as the default. The wizard asks whether to run the official npm install -g for that CLI, then prompts for the OhMyGPT API token, which is created at the OhMyGPT API token page. Next it shows endpoint latency results and asks the user to confirm the preselected fastest endpoint, and finally it asks for the default model or models before writing the configuration. The README documents what gets written for each target. For CodeX it installs or updates @openai/codex and writes ~/.codex/config.toml with the provider set to omg, the responses wire API, the base URL, and the bearer token, plus a models.json catalog that defaults to models such as gpt-5.6-sol, gpt-5.6-luna, and gpt-5.6-terra. For Claude Code it writes ~/.claude/settings.json with an environment block pointing at OhMyGPT and Opus, Sonnet, and Haiku slots. OpenCode and CodeWhale each get their own provider block and model defaults. The wizard even offers a full reset for CodeX that moves the whole ~/.codex folder to a timestamped backup and rebuilds it when errors appear.

Backups and endpoint selection

Safety is built into the write step. Before omgvibe touches any existing configuration, it creates timestamped .bak backup files next to the originals, named like config.toml.20250101-103000.bak, so a user can roll back the moment something looks wrong. That matters because overwriting API configuration by hand is easy to get wrong and annoying to undo. On the network side, OhMyGPT is served from three direct /v1 endpoints, and the wizard probes GET /v1/models on all three with up to five attempts each, then auto selects the reachable endpoint with the lowest average latency. The endpoints listed are apic1.ohmycdn.com for a Cloudflare Enterprise CDN aimed outside mainland China and the US, api.ohmygpt.com for direct US access, and cn2us02.opapi.win as a China optimized CDN for mainland China. The user still reviews and confirms the selected endpoint, so the choice is visible rather than hidden. The project also ships reference docs for each CLI under a docs folder and provides README translations in Chinese and Japanese that mirror the English file. Development uses npm install, npm run lint for type checking, and npm run build to compile TypeScript into dist, after which node dist/index.js runs the CLI locally. The MIT license covers the code.

Editorial conclusion

omgvibe is released under the MIT license, is written in TypeScript, and is launched with a single npx omgvibe command that writes backed up agent configurations.