Bestregards

Generates professional email replies by injecting an AI composition interface directly into Gmail/web email clients via Chrome extension content scripts, capturing the current email thread context through DOM parsing and passing it to a backend LLM API. The extension intercepts compose actions, extracts sender metadata and message body via DOM selectors, and streams generated responses back into the compose field without requiring tab switching or manual context copying.

Allows users to specify desired tone (professional, casual, assertive, empathetic) and style parameters (length, formality level, technical depth) which are passed as system prompt modifiers to the backend LLM before generation. The extension likely maintains a preset library of tone templates and maps user selections to prompt engineering patterns that influence model behavior without requiring fine-tuning.

Implements a client-side and server-side quota system that tracks API calls per user (likely per day or per month) and gates response generation based on remaining quota. The extension likely displays quota status in the UI and enforces hard limits for free-tier users, with upgrade prompts when quota is exhausted. Backend tracks usage via user ID or API key and returns quota headers in API responses.

Generates multiple response options (typically 2-3 variants) for a single email, each with slightly different tone, length, or approach, and presents them in a UI selector for the user to choose from before inserting into the compose field. The backend likely makes multiple LLM calls with different system prompts or temperature settings to produce variation, or uses a single call with a prompt requesting multiple options.

Manages the Chrome extension's installation, update, and runtime lifecycle, including requesting and handling permissions for DOM access, storage, and API communication. The extension uses Chrome's manifest.json to declare required permissions (content scripts, storage, host permissions for email domains) and implements background scripts to handle API calls and quota management without blocking the UI thread.

Extracts email thread context (sender name, email address, subject, message body, previous replies) by parsing the Gmail/Outlook Web DOM using CSS selectors and JavaScript DOM traversal. The extension identifies email elements by their HTML structure, extracts text content, and reconstructs the conversation thread to pass to the LLM. This approach avoids relying on email provider APIs, making it more portable but fragile to UI changes.

Inserts the generated response text into the email compose field while preserving formatting (line breaks, paragraphs) and avoiding conflicts with user edits. The extension uses DOM manipulation to set the compose field's value or contentEditable content, triggers input events to notify the email client of changes, and handles edge cases like partial edits or multi-part compose fields.

Manages user API keys (OpenAI, Anthropic, or proprietary LLM provider) securely by storing them in Chrome's encrypted storage (chrome.storage.sync or local) and passing them to backend API calls for LLM inference. The extension may use a proxy backend to avoid exposing keys in the browser, or allow users to provide their own keys for direct API calls. Authentication is likely handled via user account login (email/password or OAuth) to associate keys with user identity.

Collects user feedback on generated responses (thumbs up/down, rating, or text comments) and tracks satisfaction metrics to identify low-quality generations or common failure modes. Feedback is likely stored server-side and used for analytics, model improvement, or debugging. The extension may display feedback UI elements (buttons, modals) after response insertion or in a separate feedback panel.