INT AI Emergency Duty System

INT AI Emergency Duty System

The INT AI Emergency Duty System turns public information into structured, source-backed rules that duty officers can act on directly, giving major events, long holidays and key conferences a single place for lookups, situational awareness and field reporting. Every rule records its domain, severity, action points, location, time window, source and link to the original document, so answers can always be traced. Daily briefings are generated from the rule base by date, and the duty assistant answers questions through three fallback layers: the local knowledge base with citations, web retrieval, then LLM summarisation clearly marked as unverified. When nothing is found the system says so instead of guessing. Field observations are captured locally and exported to CSV for the shift lead to consolidate, closing the loop from rules to briefing, reporting, updates and review. The system deploys on a duty-room PC with no external API dependency and keeps working from its built-in rule base if the server is unreachable. It currently supports the 48th WorldSkills Competition Shanghai venue operations, with event, Mid-Autumn Festival and National Day duty templates built in.

• ✓97 structured rules across 7 domains, each with a source and link to the original text
• ✓Daily duty briefing generated from the rule base by date, with no external API dependency
• ✓Three-layer Q&A: local citations, web retrieval, then LLM summarisation marked unverified
• ✓Local field reporting with CSV export, consolidated at shift handover
• ✓All-object situational map with vector basemap for large 8K displays
• ✓One-click daily data refresh, with a knowledge-gap list driving rule iteration