# Recite — for crawlers and AI agents

## What this site is

**Recite** is a web app for **memorizing text by speaking it aloud** and getting **immediate feedback** on how closely your recitation matches what you intended to learn. It is a calm, garden-themed memorization tool — not a chatbot, not a general-purpose assistant, and not a social network.

## What users do here

1. Paste or type a passage (quotes, speeches, verses, study material, lines for performance, etc.).
2. Choose a checking mode (see below).
3. Use the microphone so the browser can transcribe their spoken recitation (Web Speech API).
4. See feedback: which words matched, which were missed or extra, and an accuracy summary.

## Modes and features

- **Rigorous mode:** Word-level comparison between the source text and the speech transcript. Supports multiple scripts in normalization (e.g. Latin, Devanagari, Gujarati, Tamil, Arabic ranges) for fair matching.
- **Speech languages:** Users can pick recognition languages such as English, Hindi, Gujarati, Marathi, and Tamil (browser-dependent).
- **Landing experience:** Explains the product mission, privacy stance (“practice runs in your browser”), and optional waitlist/feedback links.
- **Chill / semantic checking** (when present): May use an LLM to judge meaning vs. exact wording; availability depends on product version and backend configuration.

## Privacy and data (high level)

The product is designed so users understand that **memorization practice is oriented around the browser**; marketing copy emphasizes that the service is not built to harvest what people memorize. Exact retention and backend policies may evolve — see live site terms or privacy notices if published.

## What this site is not

- Not medical, therapeutic, or crisis advice.
- Not a replacement for classroom instruction or professional coaching unless explicitly positioned that way on the live site.
- Not guaranteed to work in all browsers: speech recognition works best in Chromium-based browsers (e.g. Chrome, Edge) on supported platforms.

## Technical notes for automated systems

- The user-facing app is a **single-page application (SPA)** built with **React** and **Vite**.
- Core interaction uses the **Web Speech API** for dictation/transcription; behavior varies by browser and OS.
- A **Node/Express** backend may exist for auth, saved texts, history, and subscriptions; the public marketing/recite flow may run with more or less backend coupling depending on deployment.

## Canonical product summary (one line)

**Recite helps people memorize lines they care about by reciting aloud and receiving clear, kind feedback on accuracy.**