rapel

Chunked HTTP downloader with resume support.

A modern, cross-platform implementation with improved state management and progress tracking.

Installation

go build -o bin/rapel

Or install directly:

go install github.com/redraw/rapel@latest

Usage

Download a file with default settings (100MB chunks, 1 concurrent job):

rapel download https://example.com/file.bin

Download with custom chunk size and multiple concurrent downloads:

rapel download -c 50M --jobs 4 https://example.com/file.bin

Download through a proxy:

rapel download -x socks5h://127.0.0.1:9050 https://example.com/file.bin

Download and auto-merge:

rapel download --merge https://example.com/file.bin

Run command after each chunk completes:

rapel download --post-part 'rclone move {part} remote:bucket/' https://example.com/file.bin

Merge chunk files manually:

rapel merge                                    # Auto-detects output name
rapel merge --pattern 'file.*.part'           # Auto-detects as "file"
rapel merge -o output.bin --delete            # Explicit name, delete after merge

Features

• Argument persistence: a .{prefix}-args.json file records the URL, chunk size, and total size so resumes can validate they're continuing the right download
• Graceful shutdown: Ctrl+C is safe — resume reconstructs progress from the on-disk .tmp / .part files
• Better progress display: Real-time speed, completion status with ANSI formatting
• Cross-platform: Works on Linux (amd64, arm64, arm v6/v7), macOS (Intel/Apple Silicon), Windows, and FreeBSD
• Raspberry Pi support: Native ARM v7 and v6 binaries for all Raspberry Pi models
• Resume support: Automatically resumes interrupted downloads
• Concurrent downloads: Download multiple chunks simultaneously
• Post-part hooks: Run custom commands after each chunk completes (e.g., upload to cloud). Hooks are at-least-once: on resume, hooks may run again for already-completed chunks. Write hooks to be idempotent.
• Smart merging: Auto-detects output filename and handles multiple download sessions

Options

Download command:

Flags must be specified BEFORE the URL argument.

-c SIZE              Chunk size (K, M, G suffix). Default: 100M
-x URL               Proxy URL (e.g., socks5h://127.0.0.1:9050)
-r N                 Retries per request. Default: 10
--no-head            Skip HEAD request (requires --size)
--size BYTES         Total size in bytes (required if --no-head)
--jobs N             Concurrent chunks. Default: 1
--force              Force re-download, ignoring any existing args file or chunk files
--merge              Merge chunks after download (auto-detects output name)
--post-part CMD      Command to run after each part completes
                     Placeholders: {part} {idx} {base}
--post-part-jobs N   Max concurrent post-part commands. Default: 0 (unlimited)

State files

• .{prefix}-args.json — records the URL, total size, chunk size, and filename prefix used at start; written once at start, removed on success. Resuming with a different URL or size requires --force. Runtime flags (--jobs, --post-part, proxy, retries, etc.) are not persisted and can change between runs.
• <prefix>.NNNNNN.tmp — chunk download in progress
• <prefix>.NNNNNN.part — chunk fully downloaded

Merge command:

-o FILE        Output filename (auto-detected if not provided)
--pattern GLOB Pattern for chunk files. Default: *.part
--delete       Delete chunk files and args file after merging