A ChatGPT client built for prompt hoarders and context keepers
It wraps the OpenAI API in a deployable SolidJS shell so you can host your own ChatGPT with granular context control and a fuzzy-search prompt library.

What it does
This is a web client for the OpenAI API, built with Solid-Start and designed to deploy on Vercel Edge Functions. It offers a polished chat interface supporting multiple concurrent conversations, each with its own URL, title, and independent settings for model and temperature. Users can fuzzy-search preset prompts and chat history via keyboard shortcuts, “lock” messages to survive chat clears or serve as persistent role instructions, and even feed queries through URL parameters to use the page like a search engine.
The interesting bit
The project treats conversation context as a tactical resource rather than a simple message log. You can lock messages to build persistent system prompts, automatically exclude failed turns from token counts, and watch real-time input and output token costs to abort expensive requests early. The interface borrows command-palette patterns—/ for prompts, // for history—which makes navigation feel closer to a code editor than a chat app.
Key highlights
- Fuzzy-searchable prompt presets and chat history via
/and//keyboard shortcuts. - “Locked” messages that persist through chat clears and act as role definitions when continuous dialogue is disabled.
- Real-time token counters for both prompts and responses, plus per-conversation model and temperature settings.
- PWA support with export/import and shareable conversation images or Markdown.
- Deploys on Vercel Edge Functions with selectable regions, plus Netlify, Railway, or Cloudflare Workers.
Caveats
vercel.appdomains are blocked in China; custom domains are mandatory, and wide sharing risks getting those blocked too.- Only Vercel allows pinning Edge Function regions; other platforms may route through OpenAI-restricted regions, risking API account bans.
- Conversation share links are local-only and not cloud-synced; editing a message refills the input box rather than editing in place.
Verdict
Worth a look if you need a self-hosted, password-protected ChatGPT web client with granular context control. Skip it if you want a managed SaaS or zero-config deployment without API keys and DNS homework.
Frequently asked
- What is ourongxing/chatgpt-vercel?
- It wraps the OpenAI API in a deployable SolidJS shell so you can host your own ChatGPT with granular context control and a fuzzy-search prompt library.
- Is chatgpt-vercel open source?
- Yes — ourongxing/chatgpt-vercel is open source, released under the MIT license.
- What language is chatgpt-vercel written in?
- ourongxing/chatgpt-vercel is primarily written in TypeScript.
- How popular is chatgpt-vercel?
- ourongxing/chatgpt-vercel has 3.2k stars on GitHub.
- Where can I find chatgpt-vercel?
- ourongxing/chatgpt-vercel is on GitHub at https://github.com/ourongxing/chatgpt-vercel.