I’m refreshing a small proofreading site and want to replace the current home-grown spell-checker with Gemini AI. The stack is deliberately simple—static HTML/CSS on the front end, one lean PHP file at the back—and I’d like to keep it that way. Here’s the flow I’m aiming for: a visitor pastes up to 2,000 words, hits “Check,” and almost immediately sees (1) a fully corrected version of the text, (2) a concise list of spelling fixes and light grammar tweaks, plus (3) a short explanation next to each change. Results should stream in as soon as the API responds so the page never feels stuck. Key points • Connect the existing PHP back-end to Gemini’s API—no heavyweight frameworks. • Build or polish the front-end form so it’s clean and responsive on desktop and mobile. • Return a change log that clearly shows original vs. corrected words. • Add a lightweight logging table or flat-file log that records date, time, and word count for each request so we can watch daily usage. • Provide a brief README covering where to drop API keys, rate-limit settings, and any deploy/rebuild commands. If you’ve wired up Gemini—or OpenAI, Claude, etc.—into a custom interface before, a link or short description will help me gauge fit. Deliverables 1. Updated PHP script(s) that call Gemini and stream responses. 2. Refreshed HTML/CSS/JS front end with input box, streaming output, and change log display. 3. Usage logging mechanism. 4. Setup & maintenance documentation. The lighter the footprint, the better. Let me know your rough approach and any similar work you’ve shipped.