Built a quick, sub-15ms Rust CLI/TUI to pack repos into prompts without burning 40k tokens on lockfiles and junk

Whenever I feed codebases into local models (Qwen, DeepSeek R1) or API models, the biggest annoyance is prompt pollution:

- Lockfiles (`Cargo.lock`, `package-lock.json`) burning 30,000+ tokens for zero reason.

- SVGs, binary files, or build artifacts slipping into the context.

- Existing packers taking 3-4 seconds just to generate the prompt.

I built a small tool called repOx to fix this for my own workflow.

GitHub: https://github.com/WVDYC/repOx

Features:

1. Speed: Written in Rust, takes ~14ms to dump a 3k-file repo.
2. Lazygit-style TUI (`repox -i`): Opens a fast terminal UI where you can uncheck folders with Space, preview files, search with `/`, and watch a live token gauge before copying.
3. Clean output: Strips lockfiles and binaries by default using Git NUL-byte heuristics. Dumps straight to clipboard (`repox -c`).
4. Offline Tokenizer: Supports token budgets for Claude, GPT, Gemini, DeepSeek, and Llama contexts so you know beforehand if you're exceeding your window.

One-line install (macOS / Linux):

`curl -fsSL https://raw.githubusercontent.com/WVDYC/repOx/main/install.sh | sh`

Code is open source (MIT / Apache). Curious what you all currently use for feeding code into LLMs and if there are specific prompt templates you'd like added.