EZGPT.APP

EZGPT.APP puts a chat window for your own language models in the browser side panel. Point it at LM Studio or Ollama on this machine, a server on your LAN, or a cloud API, and talk to the model while you keep browsing. It stays open when you click around the page, and it never sends anything anywhere except the server you configured. WORKS WITH • LM Studio and Ollama on localhost or your LAN • OpenAI, OpenRouter, Anthropic and Gemini through their OpenAI-compatible endpoints • Any server that speaks the OpenAI chat completions API (vLLM, llama.cpp, LocalAI, text-generation-webui and others) WHAT YOU GET • Side panel first. Open it with one click or Ctrl+Shift+E. It stays open while you browse, and the reply keeps streaming. • Model switcher in the top bar. Discover the models a server offers and change model per chat without editing settings. • Streaming replies with a Stop button. Partial replies are kept if you stop or close the panel. • Reasoning models handled. DeepSeek, Qwen, QwQ and gpt-oss style thinking shows in a collapsible block, whether the server sends it as a reasoning field or as think tags. • Ask about the page. Attach the selected text or the readable text of the current tab to your message. Right-click any selection and choose "Ask EZGPT.APP about…". • Images. Paste, drop or attach screenshots for vision models such as Qwen-VL, LLaVA, Pixtral or GPT-4o. • Markdown answers with code blocks and one-click copy. • Per-chat parameters: temperature, top-p, max tokens, stop sequences. • Prompt snippets you can save and insert. • Chat history with search, rename and grouping by day. Export a chat as Markdown or JSON, or back up everything as one file and restore it later. • Light, dark and system themes, four accents, three densities. • Open the same chat in a full tab when you want more room. WHAT IS NEW IN 4.3 • Cloud services one tap away. • Debate rounds are a slider, up to five, and the panel stops early once no model changes its answer. • A reply is tagged with the model that wrote it whenever the model changed mid-conversation, and changing a provider's default model in Settings takes effect at once. • On a phone, starting a chat from the drawer closes the drawer, and the group chat button works with the keyboard up. • A long code line or a wide table in a panel answer no longer pushes the transcript past the edge of the screen, and closing the chat list in a tab no longer blanks the page. WHAT IS NEW IN 4.2 • Group chat: pick two to four models, ask once. Every model answers, reads the others, revises, then votes. The answer with the most votes is the reply, with the whole debate one tap away. WHAT IS NEW IN 4.1 • A refreshed look: a lit command line for your message, a quieter transcript, larger controls that work under a thumb. • Starter prompts are cards, and tapping anywhere in the message box starts typing. • Terms of service and privacy policy linked from Settings → About. PRIVACY EZGPT.APP is developed and owned by US Support LLC. There is no account, no analytics and no telemetry. Settings, chats and images live in your browser's storage. The extension talks only to the servers you add, and only when you send a message, discover models or test a connection. Page text is read only when you click "Attach page" or use the context menu. GETTING STARTED 1. Start your model server (for LM Studio: Developer → Start server; for Ollama it is already running). 2. Click the EZGPT.APP icon. The side panel opens with LM Studio preset at http://localhost:1234/v1. Change the preset in Settings if you use something else. 3. Pick a model in the top bar and type your first message. KEYBOARD SHORTCUTS Ctrl+Shift+E open the side panel · Enter send · Shift+Enter new line · Esc stop · Alt+N new chat · Ctrl+K search chats · Ctrl+Shift+P attach page · Ctrl+, settings · Alt+1…9 switch provider Upgrading from 3.x: your providers, chats, images and snippets are migrated automatically the first time 4.x starts.