sgrep

sgrep is an open-source command-line search tool from Context.dev on GitHub. Describe what you're looking for in a plain English sentence and it finds the matching passages in the code or document folder you point it at. It scores results by both keywords (BM25) and meaning, with meaning weighted at 75%; it can also re-rank results piped in from ripgrep, and it can output JSON for other programs. It uses static embeddings, so no large model runs at search time, and the README says an ordinary computer's CPU can search locally with no cloud API call per query. It needs no account, hosted index or model server, and the README says "Source and queries stay on your machine."

Who it's for:programmers who dig through large codebases or piles of documents, and developers who want their AI coding assistant to spend fewer tokens. Readers who don't write code won't have a use for it.

Before you start: install Rust 1.90 or later and ripgrep, then build it from the repository with cargo. A prebuilt binary exists only for macOS on Apple Silicon; support for other systems isn't spelled out. The first search downloads a model of about 32 to 34 MB plus any code parsers it needs, and after that it can work offline. The local index keeps copies of code passages, so clear the cache if you've searched sensitive code. The code and models are MIT licensed.

Based on the project's public documentation. Not tested by AI Inspiration World.