Summarize On Select 2.1 — major update (rebuilt from the ground up since 1.4)

MULTIPLE AI PROVIDERS & MODELS
• No longer OpenAI-only. Choose your AI: OpenAI GPT, Anthropic Claude, Google Gemini, DeepSeek, Alibaba Qwen, xAI Grok, Mistral, OpenRouter — or any OpenAI-compatible endpoint, including local/self-hosted servers.
• Built-in model picker with per-model pricing hints, so you can balance speed, quality and cost.
• Reasoning-model support: the model’s reasoning is streamed and shown in a separate, collapsible section.
• Advanced per-model settings: temperature, max output tokens, reasoning effort, and custom request fields (JSON).

SUMMARIES
• Real-time streaming output — start reading immediately.
• Markdown-formatted results (headings, lists, code blocks, links).
• Adjustable summary length: weak / medium / strong compression.
• Output language: keep the source language automatically, or force English or Russian.
• One-click copy of the summary.

THREE WAYS TO TRIGGER
• Alt+N keyboard shortcut (customizable).
• Floating “TL;DR” button on selection.
• Right-click context menu “Summarize selection”.

NEW CUSTOMIZATION (2.1)
• Drag the summary window anywhere by its header — with optional position memory.
• Floating button: choose its position (left, right, bottom-center, or next to the selection), scale it, or turn it off.
• A single accent color for the whole extension.
• Light / dark / auto theme; English and Russian interface.

PRIVACY & SECURITY
• Bring your own API key — stored locally, never synced, sent only to the provider you choose.
• Host permissions are narrowed and requested only when you configure a provider (incl. custom/local endpoints).
• No analytics, no tracking, no remote code.

DIAGNOSTICS
• New “Copy debug logs” in Settings and in the toolbar popup for quick troubleshooting.

UNDER THE HOOD
• Full Manifest V3 rewrite for Chrome & Firefox (TypeScript + esbuild), robust streaming and error handling, automated tests + CI, and automatic migration of your previous settings.