Who it is forA global support team that receives tickets in Hindi, Spanish and dozens of other languages.
The problemTranslating every message before classifying it adds cost, delay and errors.
The outcomeThe same department and urgency decisions for any language, with the router reporting which checkpoint answered.
What you need first
- Python 3.10 or newer
- pip or uv
- A GPU is optional; checkpoints download on first use
- Optional extras: laya[serve], laya[mcp], laya[langchain] and others
Steps
- Install Laya and build your questions as in the Quickstart
- Pass non-English ticket text to
router.predictwithout translating it - Check result routing model to see whether the multilingual checkpoint answered
- For long documents pass model multilingual and max_len 8192, as the README warns
- Try a quick check from the shell with
laya "My payment failed twice" --preset triage
Prompt for your agent
Using NandhaKishorM/laya's Router, classify these customer messages in their original languages by department and urgency: <messages>. Print the routing model for each, and use model multilingual with max_len 8192 for any message longer than a page.
Caveat. Beyond about 4,000 tokens of text the README reports results vary, so check long-document accuracy on your own data.
Use this in your AI agent
Ask your agent
Using the Deduped MCP, get the setup steps for NandhaKishorM/laya and "Triage customer messages in 100+ languages with one model"
Add Deduped to Claude Code once
claude mcp add --transport http deduped https://dedupedai.com/mcp
Every Monday: the 5 hottest repos and exactly how to use them. Free, one email a week, unsubscribe anytime.
Get them freeShare: X LinkedIn Hacker News Email Copy link
Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer