Browser automation for your Xianyu side hustle

What it does

This is a web-based control panel that automates storefronts on Xianyu, China’s largest second-hand marketplace. It drives real browser sessions via Playwright and DrissionPage to let sellers manage multiple accounts, auto-reply to buyers, trigger shipments, and confirm deliveries from a single dashboard. Under the hood it runs FastAPI, SQLite, and Bootstrap 5, packaged for Docker with multi-arch support.

The interesting bit

The reply logic follows a rigid priority ladder: specific product replies beat product keywords, which beat general keywords, which beat a default reply, which finally falls back to an AI model. It is a rare system where old-school if-trees outrank the LLM. Each account also gets its own browser fingerprint and cookie directory to avoid tripping anti-bot measures.

Key highlights

• Multi-user, multi-account architecture with full data isolation per tenant.
• Keyword-driven auto-reply with Excel import/export and image keyword support.
• Auto-delivery engine that matches product SKUs to card-based or text-based fulfillment rules, with de-duplication guards.
• Hot-update mechanism via GitHub Actions that diffs .py and .html changes into versioned releases without touching user data.
• WebSocket and SSE feeds for real-time logs and account status monitoring.

Caveats

• The README labels the project AGPL-3.0 and “for educational use only,” which is usually a signal that it violates the platform’s Terms of Service.
• It requires Node.js solely to power PyExecJS, a dependency that feels archaic in 2024.
• The default admin credentials are admin/admin123, so you’ll want to change those immediately.

Verdict

Worth a look if you are a Xianyu power seller or a developer studying e-commerce browser automation. Skip it if you need an officially blessed API—this is scraping, not sanctioned integration.