What it does

This repo provides a working Chinese dialogue system built on Rasa 1.10.18, with a Makefile-driven workflow for training, running, and testing via command line or HTTP. It targets Python 3.6+ and includes a requirements.txt for dependency management. The author also maintains a WeChat group for Rasa enthusiasts—contact: coffee199029.

The interesting bit

The real payload isn’t the code; it’s the ten-part Jianshu blog series (“踩坑记” / “pitfall notes”) documenting the author’s journey through Rasa’s evolving Chinese support. The README candidly admits the project stalled at 1.10.18 because Rasa 2.0’s changes were too disruptive to chase, and notes that upstream Rasa has since improved BERT component support and Chinese handling natively.

Key highlights

• Pre-Rasa-2.0 Chinese pipeline preserved on master; older 1.1.x code moved to its own branch
• make train / make run / make shell / make run-x for standard Rasa workflows
• HTTP endpoint at localhost:5005/webhooks/rest/webhook for POST testing
• Extensive Chinese-language documentation via linked blog posts
• Topics cover intent classification, slot filling, policy configuration, and even small-talk (闲聊) integration

Caveats

• Explicitly outdated: author states they are waiting for Rasa 2.0 to stabilize before updating
• The external links section is a grab bag of unrelated AI tools (body visualizers, deepfake apps, etc.) with no clear connection to the project
• No candidate images available for visual reference

Verdict

Worth a look if you’re maintaining a legacy Rasa 1.x Chinese bot or want to read war stories from someone who fought BERT integration before Rasa made it official. Skip it if you need a current, production-ready Rasa 2.x+ solution.