PolyLingo is a full-stack web application consisting of a React/TypeScript frontend and a FastAPI Python backend. The backend exposes a single `/translate` endpoint that forwards user text to Google's Gemini 2.5 Flash model with a structured prompt, then parses the JSON response into translation, meaning, pronunciation, and example fields.

The problem it addresses is the limitation of plain text-to-text translators. Traditional tools like Google Translate give a literal translation but rarely explain context, pronunciation, or usage. PolyLingo tries to bundle those learning-oriented artifacts into a single API call, marketing itself as a translation-plus-comprehension tool.

The intended users are language learners and casual translators who want more than a one-word output. The UX flow is straightforward: a landing page with animated hero, a translation panel with source/target language selectors, a history panel backed by localStorage (capped at 50 entries), and a settings panel for theme and animation preferences.

Under the hood, the frontend is a Vite + React 18 app using Radix UI primitives, Tailwind CSS v4, Motion for animations, and shadcn-style components. The backend is minimal: FastAPI with CORS locked to localhost:5173, one Pydantic model, one route, and a service module that calls `google.generativeai` and strips markdown code fences from the response before parsing JSON.

Notable observations: the frontend package.json is named `@figma/my-make-file` and vite.config.ts contains a `figmaAssetResolver` plugin plus a `MARKER-MAKE-KIT-INVOKED` comment in App.tsx, indicating the UI was scaffolded via Figma Make. The backend JSON parsing is naive (a single `json.loads` with no error handling), and there is no test suite, no license file, and no rate limiting on the API endpoint.