Cross-Border E-Commerce AI Suite

Problem

Cross-border sellers repeatedly handle several content types: product images with inconsistent backgrounds, localized listings, support scripts, live-stream plans, and unpredictable model usage costs. Cross-Border E-Commerce AI Suite splits these tasks into independent Python agent scripts, reducing manual prompt stitching and upload handling.

How It Works

Each module is a script that can be used on demand:
- Product image generation: supports text-to-image and original image upload, with modes such as BG_REMOVAL, ENHANCEMENT, MAIN, and SCENE.
- Listing writing: generates multilingual product listings with titles, selling points, descriptions, and SEO optimization.
- Customer support: handles multilingual replies, sentiment analysis, priority judgment, and automatic responses for pre-sale, logistics, after-sale, and complaint scenarios.
- Live-stream scripts: creates platform-adapted scripts with a full timeline, including opening, closing, interaction, and emergency lines.
- Token management: allocates quotas, tracks consumption, and emits warnings; storage is in-memory by default, or persistent JSON via storage_path.
- Image tools: accepts path, URL, Base64, or byte streams; reusable prompt templates are in references/.

Boundaries

Requires Python 3.8+ and usually no extra dependencies; Pillow is optional for image size validation. The modules are independent, so teams can embed only image generation, listing writing, or customer support into existing workflows. If long-term token usage records are needed, pass storage_path; otherwise the manager uses memory storage.