Grab Text

A Go CLI tool that downloads YouTube subtitles, converts VTT to plain text, and summarizes the content using a local GGUF language model. Uses CGo bindings to call llama.cpp directly for on-device inference — no cloud APIs, no network requests for AI. Replaces a prior shell script + Python + opencode pipeline with a single compiled binary.

Tech Stack

GoCGollama.cppCUDAGGUFyt-dlp

Key Features

  • Direct CGo bindings to system llama.cpp — loads and runs GGUF models on GPU
  • VTT subtitle parser ported from Python to Go with full test coverage
  • Uses Gemma 4 E2B model (~5GB) for fast local text summarization
  • Single 3.4MB binary replaces a 3-file shell/Python/AI pipeline