rgfile

A command-line client for GigaFile.nu.

English | 简体中文 | 日本語

Install

curl -fsSL https://raw.githubusercontent.com/Maymall/gigafile-rust-cli/main/install.sh | sh   # Linux / macOS
cargo install rgfile                                                                          # Rust 1.85+
brew install Maymall/tap/rgfile                                                               # Homebrew

Windows: irm https://raw.githubusercontent.com/Maymall/gigafile-rust-cli/main/install.ps1 | iex

Prebuilt archives and a Debian package are on the releases page. Release-installed binaries upgrade themselves with rgfile self-update.

Usage

rgfile ul file.bin                   # upload; prints the URL, delete key, expiry
rgfile ul file.bin --lifetime 7      # keep for 7 days (3–100)

rgfile dl <url>                      # download; interrupted transfers resume
rgfile dl <url> --threads 8          # segmented download over several connections
rgfile dl <url> --select 1,3-5       # pick files from a multi-file page

rgfile info <url>                    # inspect a page without downloading
rgfile delete <url>                  # take an upload down, using its delete key
rgfile parts list                    # leftover partial downloads
rgfile parts clean --older-than 7    # drop stale ones; active downloads are never touched

rgfile config init                   # interactive configuration
rgfile history list                  # local history (opt-in)
rgfile completions zsh               # shell completions

Every command takes --json. rgfile <command> --help has the rest.

Configuration

Optional TOML at ~/.config/rgfile/config.toml (macOS: ~/Library/Application Support/rgfile/, Windows: %APPDATA%\rgfile\). CLI flags override it.

[download]
dir = "/home/alice/Downloads"
threads = 8                    # connections per file, 1–16

[upload]
lifetime = 7                   # days: 3/5/7/14/30/60/100
threads = 4                    # read-ahead chunk window, 1–16

[history]
enabled = true                 # off by default
store_delete_keys = false      # plaintext, opt-in

Behavior

• Downloads resume from where they stopped, even when the page masks the file name; completion is atomic and size-checked. Names come from Content-Disposition, so UTF-8 survives.
• Segmented downloads keep a few connections active and back off when the server pushes back, instead of hammering it.
• Ctrl-C prints how much reached disk and how to resume. Delete keys and download passwords never appear in logs.
• Uploads stream with per-chunk retry; chunk completion stays ordered because the server drops out-of-order chunks (verified against the live service).
• rgfile does not bypass GigaFile restrictions, guess passwords, or scrape links.

Exit codes

Code	Meaning
0	Success
2	Invalid arguments
10	Not a GigaFile URL
11	Network failure, retries exhausted
12	Unexpected HTTP status
13	Page could not be parsed
14	Not found or expired
15 / 16	Download key required / rejected
17	Size mismatch
18	Filesystem error, or the target already exists (--force overwrites)
19	Upload rejected
20	Verification failed
21	Target locked by another rgfile process
22	Delete rejected
130	Interrupted; the kept .part resumes on re-run

Changelog: docs/CHANGELOG.md

License

MIT — see LICENSE.

The GigaFile.nu protocol flow was originally worked out with reference to Sraq-Zit/gfile and its fork fireattack/gfile. Thanks!