What it does

This repo is a hand-curated index of Chinese-language articles, talks, and whitepapers about production chatbot systems. It is organized by company—Alibaba (Xiaomi), JD (JIMI), Ctrip, 58.com, Bilibili, Kuaishou, and others—and covers both pre-LLM era architectures and newer “large-model era” agentic-AI practices. Most entries link to WeChat articles, Aliyun Drive files, or conference recordings rather than code.

The interesting bit

The value is geographic and linguistic: it surfaces engineering write-ups from Chinese tech giants that rarely get English-language coverage. You can trace how Alibaba’s Xiaomì evolved from retrieval-plus-generation models to emotion-aware replies, or how Ctrip runs a ten-thousand-seat call center with active-active disaster recovery.

Key highlights

• Covers both classic NLP pipelines (intent prediction, knowledge graphs, multi-turn dialogue management) and LLM-era shifts (RAG, Copilot-to-Agent transitions, AI SRE).
• Heavy representation from e-commerce and travel verticals—Alibaba, JD, Ctrip, Qunar—with recent additions from ride-sharing (Hello), ed-tech (51Talk), and property management (Vanke).
• Includes non-technical material: product UX teardowns, whitepapers, and “how we built it in five days” war stories (e.g., Alibaba’s pandemic-response bot).
• Last updated August 2025; actively maintained for roughly two years.

Caveats

• Almost all sources are behind WeChat or Chinese cloud-drive links; expect paywalls, extraction codes, and link rot.
• No code, no benchmarks, no reproducible artifacts—this is purely a reading list.
• English speakers will need translation tools; the repo itself is Chinese-only.

Verdict

Worth bookmarking if you are researching how Chinese platforms scale customer-service AI, or if you need case-study ammunition for your own bot roadmap. Skip it if you want runnable models or English-language documentation.