ReClip

A self-hosted, open-source media downloader supporting 1,000+ platforms via yt-dlp. Forked from averygan/reclip and improved with Android deployment, security hardening, performance optimization, and automated testing — all authored by Google Jules and reviewed through GitHub Copilot.

The Problem

The original ReClip was a clean, minimal media downloader — Python + Flask backend, vanilla HTML/CSS/JS frontend, roughly 150 lines of backend code. It worked well on macOS and Linux desktop environments, but had several gaps:

• No mobile/Android deployment story
• No production-grade WSGI server (ran Flask's dev server)
• A command injection vulnerability in yt-dlp argument construction
• No test coverage for the download API
• Performance bottlenecks in string sanitization routines
• High cyclomatic complexity in the core download function

The Solution

Forked the repository and used Google Jules as an AI coding agent to systematically address each gap across 9 pull requests. GitHub Copilot provided automated code review, and Jules iterated on feedback before merge.

Android / Termux Deployment

• Replaced Flask's development server with Waitress (WSGI) for production-grade concurrency with configurable thread count via WAITRESS_THREADS
• Integrated aria2c as an external multi-threaded downloader for yt-dlp
• Created install.sh for one-click Termux setup — installs Python, FFmpeg, Git, Aria2, Termux-API, and Cloudflared automatically
• Created start-background.sh for persistent background execution via nohup with PID tracking and log output
• Optional Cloudflare Quick Tunnel exposure via ENABLE_CLOUDFLARED_TUNNEL=1 with explicit security warnings
• Automatic termux-wake-lock acquisition to prevent Android from suspending background processes
• Full deployment documentation in DEPLOYMENT.md

Security Hardening

• Fixed argument injection vulnerability in yt-dlp command construction — user-supplied URLs starting with hyphens were interpreted as CLI flags
• Added -- separator before URL arguments in both get_info and run_download to force positional argument parsing
• Cloudflare Tunnel exposure changed from automatic to opt-in with security guidance documentation

Performance Optimization

• Replaced character-by-character generator expression with pre-compiled re.sub() for filename sanitization
• Short titles: 32% faster, long titles: 90% faster (100K operations)
• Pre-compiled RE_UNSAFE regex pattern: 47% reduction in execution time (6.71s → 3.53s over 1M iterations)

Code Quality

• Moved inline imports to top-level per PEP 8 with HAS_WAITRESS flag for optional dependency handling
• Extracted _build_download_command and _finalize_download helpers to reduce cyclomatic complexity in run_download
• Added pytest test suite for /api/download covering missing URL, empty URL, missing JSON body, and valid request paths
• Fixed crash on POST requests with missing JSON body (request.json or {})

Highlights

Results

• Fully functional self-hosted media downloader running on Android via Termux
• Accessible at localhost:8899 or via public Cloudflare Tunnel URL from any device
• Supports YouTube, TikTok, Instagram, Twitter/X, Reddit, Facebook, Vimeo, Twitch, SoundCloud, and 1,000+ other platforms
• MP4 video and MP3 audio extraction with quality/resolution selection
• Bulk downloads with automatic URL deduplication
• ~990 lines added, ~340 lines deleted across the fork

What I Learned

Google Jules paired with GitHub Copilot code review creates a surprisingly effective contribution pipeline. Jules handled everything from deployment scripting to security patching to performance profiling — all from task descriptions sent via phone. The AI-to-AI feedback loop (Jules writes, Copilot reviews, Jules fixes) caught real issues: suppressed JSON output, hardcoded port values, unreliable environment detection, and public tunnel exposure risks.

The entire contribution — 9 PRs, 8 merged — was done from a phone on a train with no IDE. AI coding agents have fundamentally changed what "I don't have time to contribute" means.

---

Live Demo · GitHub · Upstream