Ready-to-paste prompts for handing to an AI coding agent (Claude, ChatGPT, Cursor, whatever you use) so it builds something real on top of the actual LuisSearch API -- not made-up endpoints, the same ones documented there.
luisearch.pages.dev is the actual website you're on right now -- what a person visits in a browser. The trycloudflare.com address in the prompts below is the real API backend the website itself calls -- that's the one your code (or an AI agent's code) should hit directly for data. It looks like a random/temporary address because it genuinely is one (a Cloudflare Tunnel), but it's live and it's the real endpoint right now.
A full search-engine-style site, from scratch.
Build me a real, working website with a search page powered by the real LuisSearch API (no API key needed for basic use, it's free and public).
Note: luisearch.pages.dev is the website a person browses; the trycloudflare.com address below is the actual real API backend that website itself calls. It looks like a random/temporary domain because it's a Cloudflare Tunnel, but it is genuinely live -- call it directly.
Endpoints to use:
- GET https://bizrate-translated-protocols-mainland.trycloudflare.com/api/search?q=<query> -> {"results":[{"id","url","title","snippet"}, ...]}
- GET https://bizrate-translated-protocols-mainland.trycloudflare.com/api/stats -> {"pages","terms"} (real live index size)
- GET https://bizrate-translated-protocols-mainland.trycloudflare.com/api/image-search?q=<query> -> {"results":[{"src","alt","page_url","host"}, ...]}
- GET https://bizrate-translated-protocols-mainland.trycloudflare.com/api/audio-search?q=<query> -> {"results":[{"title","url","duration","description"}, ...]}
- GET https://bizrate-translated-protocols-mainland.trycloudflare.com/api/video-search?q=<query> -> {"results":[{"title","channel","platform","duration","url","thumbnail"}, ...]}
Requirements:
- A clean, modern single-page site: a search box, results list (title links to the real url, shows the snippet), and tabs or a toggle for image/audio/video results.
- Show the real "pages indexed" / "unique terms" count from /api/stats somewhere on the page.
- Handle empty results and network errors gracefully (don't just show a blank page).
- Debounce input if you add live-search-as-you-type; otherwise a plain submit button is fine.
- No fake/mock data anywhere -- every result shown must come from a real fetch() call to these endpoints.
A native or cross-platform app (mobile/desktop) using the same API.
Build me a real app (ask me what platform if you need to -- otherwise default to whatever's easiest to run and test here) that searches the real LuisSearch API.
Note: luisearch.pages.dev is the website a person browses; the trycloudflare.com address below is the actual real API backend that website itself calls. It looks like a random/temporary domain because it's a Cloudflare Tunnel, but it is genuinely live -- call it directly.
Base URL: https://bizrate-translated-protocols-mainland.trycloudflare.com
Real endpoints (JSON, no auth needed for light use):
- GET /api/search?q=<query> -> {"results":[{"id","url","title","snippet"}, ...]}
- GET /api/stats -> {"pages","terms"}
- GET /api/image-search?q=<query> -> {"results":[{"src","alt","page_url","host"}, ...]}
- GET /api/audio-search?q=<query> -> {"results":[{"title","url","duration","description"}, ...]}
- GET /api/video-search?q=<query> -> {"results":[{"title","channel","platform","duration","url","thumbnail"}, ...]}
Requirements:
- A real search screen: input field, submit, scrollable results list.
- Tapping/clicking a result opens the real URL (in-app browser or system browser, whichever is idiomatic for the platform).
- Loading state while the request is in flight, and a real error state if the request fails (not a silent failure).
- Keep it to one screen unless I ask for more -- don't over-build.
- Every piece of data shown must come from an actual network call to the endpoints above -- no placeholder/lorem-ipsum results.
Drop-in integration into a site you already have.
Add a real search feature to my existing site using the LuisSearch API -- don't scaffold a new project, integrate into what's already here.
Note: luisearch.pages.dev is the website a person browses; the trycloudflare.com address below is the actual real API backend that website itself calls. It looks like a random/temporary domain because it's a Cloudflare Tunnel, but it is genuinely live -- call it directly.
GET https://bizrate-translated-protocols-mainland.trycloudflare.com/api/search?q=<query>
Response: {"results": [{"id": number, "url": string, "title": string, "snippet": string}, ...]}
Requirements:
- Match my site's existing visual style -- don't introduce a different design language.
- A search input (reuse an existing one if my site already has a nav/search bar) that calls the endpoint above on submit.
- Render results as a list: title (linking to the real url), and the snippet below it.
- Handle the empty-results and network-error cases visibly, not silently.
- Don't add any backend/server code for this -- it's a direct client-side fetch() to the public API.
A real terminal command that searches LuisSearch.
Build me a real, installable command-line tool (pick whatever language is idiomatic for how I'd actually run this, ask me if unsure) that searches the real LuisSearch API from the terminal.
Note: luisearch.pages.dev is the website a person browses; the trycloudflare.com address below is the actual real API backend that website itself calls. It looks like a random/temporary domain because it's a Cloudflare Tunnel, but it is genuinely live -- call it directly.
Usage should look like: lsearch <query>
Endpoint: GET https://bizrate-translated-protocols-mainland.trycloudflare.com/api/search?q=<query>
Response: {"results": [{"id", "url", "title", "snippet"}, ...]}
Requirements:
- Print the top 5-10 results to stdout: title, url, and a truncated snippet, nicely formatted (color if the terminal supports it).
- Handle no-results and network-error cases with a clear message, not a stack trace.
- Keep dependencies minimal -- standard library HTTP + JSON if the language has it built in, no heavyweight framework for something this small.
- Don't fabricate example output -- actually run it against the real endpoint and show me real results.
A slash command that searches the real index.
Add (or build, if I don't have one yet) a Discord bot slash command, /search <query>, that calls the real LuisSearch API and replies with real results as an embed.
Note: luisearch.pages.dev is the website a person browses; the trycloudflare.com address below is the actual real API backend that website itself calls. It looks like a random/temporary domain because it's a Cloudflare Tunnel, but it is genuinely live -- call it directly.
Endpoint: GET https://bizrate-translated-protocols-mainland.trycloudflare.com/api/search?q=<query>
Response: {"results": [{"id", "url", "title", "snippet"}, ...]}
Requirements:
- /search takes one required string option, "query".
- Reply with an embed listing the top 3-5 results: title as a field name, snippet + url as the field value.
- Handle the no-results case with a plain reply, not an empty embed.
- If the request fails or times out, reply with a real error message, don't fail silently.
- Use whatever Discord library/language the rest of the bot already uses if one exists; otherwise pick something reasonable and tell me what you picked.