Overview
Apogee is a private AI summarizer that runs in the browser. It can summarize articles, videos, emails, PDFs, DOCX files, and pasted text while keeping model processing local rather than sending content to a cloud API.
Inspired by Mozilla's discontinued Orbit project, Apogee is designed as an offline-first browser extension with no account, subscription, API key, telemetry service, or Apogee backend required.
Key Features
- Performs tokenization, inference, retrieval, and caching on the user's device.
- Uses GPU acceleration through WebGPU on Chrome, Edge, and other Chromium browsers.
- Supports CPU inference through WebAssembly, which is the default on Firefox and an optional fallback on Chromium browsers.
- Connects directly to a locally hosted Ollama instance or llama.cpp server for users who want larger models.
- Keeps page content, transcripts, files, and generated summaries away from cloud AI APIs.
- Caches downloaded model weights so subsequent use can work offline.
How to Use
After installing the browser extension, users open an article, video, PDF, or DOCX file, or paste text into the Apogee popup. They then choose a summary format and start the summarization process.
The first in-browser run downloads the selected model. Once those model files are cached, Apogee can run offline; users who prefer larger local models can configure Ollama or llama.cpp in the extension settings.
Privacy and Processing
Apogee has no cloud backend and does not require content uploads for summarization. Its local-first architecture covers both the source material and the generated summary, making it suitable for users who want browser-based assistance without routing their reading data through a remote AI service.
WebGPU provides an accelerated path on supported Chromium browsers, while WebAssembly broadens compatibility to systems or browsers without WebGPU. The optional local-server connections preserve local processing while offering an alternative model runtime.