RECAP
A Chrome extension paired with a local Python backend that indexes the pages a user actually reads and lets them search their browsing history in plain English using local retrieval-augmented generation.
RECAP quietly indexes the content of pages a user spends time on, keeping everything on their own machine, and answers questions about browsing history with retrieval-augmented generation. Nothing about browsing is sent anywhere except the local backend, and an LLM provider is contacted only when the user asks a question.
The Chrome extension tracks pages actually read, extracts their text, and provides a new-tab dashboard, a toolbar popup and a keyboard-triggered search bar; it skips sensitive sites such as banking, email, logins and password managers by default. A local FastAPI backend chunks and indexes page content, then answers queries with hybrid retrieval that combines keyword search, vector search and a small knowledge graph before ranking results and generating an answer with sources.
The backend runs on loopback only, and connects to any OpenAI-compatible LLM endpoint, or a self-hosted server such as Ollama, vLLM or LM Studio. It installs via Docker, a Python package manager, or pip, and requires Python 3.11 through 3.13.