WeChat Pay just dropped version 2.0 of its AI integration toolkit, giving developers a smarter way to hook large language models into its ecosystem. The biggest upgrade? Every sequence diagram and state diagram in the documentation is now rendered in Mermaid format, so even models without multimodal capabilities can read them directly. The result is a 50% reduction in token consumption compared to raw HTML.
To avoid the garbled translations that machine direct-translation often produces, the toolkit now supports nine languages, including Chinese, Japanese, and Korean — and it's tuned to understand native concepts like "WeChat Pay cashier" (微信支付收银台). The knowledge base has expanded from version 1.0's focus on payments and coupons to cover WeChat Pay's entire product line. Two new AI expert roles are available: a technical expert and a financial-grade R&D expert, capable of answering troubleshooting questions and reviewing code quality.
On the developer experience front, the new CLI dynamic troubleshooting feature lets developers query payment and refund order status right from the terminal using natural language — no more breaking your flow to log into the merchant backend. The Skill also automatically compares the official website and local knowledge base versions on startup, keeping everything synced without manual intervention.