charvithaaa/Coldchain-sentinel

ColdChain Sentinel is an OpenEnv environment for **medical cold-chain diagnosis and intervention** across a district network of a warehouse, a hub and clinics in under-resourced communities. The AI agent must detect and respond to cold-chain problems while protecting temperature-sensitive stock and keeping clinical services running. # ColdChain Sentinel ColdChain Sentinel is an OpenEnv environment for **medical cold-chain diagnosis and intervention** across a district network of a warehouse, a hub and clinics in under-resourced communities. The AI agent must detect and respond to cold-chain problems while protecting temperature-sensitive stock and keeping clinical services running. ## Why this environment exists Medical cold-chain management is a critical real-world problem. Human operators often need to: - check unusual temperature alerts - determine whether the problem is real or a false alarm - turn on backup systems - send repair staff - move stock between sites (warehouses, hubs, and clinics) - protect key clinics when transport resources are limited This environment turns that real-world challenge into a repeatable OpenEnv test. ## Environment overview This environment simulates: - 1 warehouse - 1 hub - 6 clinics - 2 vehicles - 1 technician - 3 product classes: - `routine_vaccine` - `priority_pediatric_vaccine` - `biologic` The agent interacts through the standard OpenEnv API: - `reset()` - `step(action)` - `state()` ## Task set ### 1. `single_clinic_fridge_drift` (easy) A clinic fridge shows an abnormal temperature alert. The issue might be either a real fridge failure or a false sensor alarm. ### 2. `transport_delay_and_partial_outage` (medium) A shipment is delayed. At the same time, one clinic has power issues and another clinic has rising demand. ### 3. `district_compound_failure` (hard) Several clinics show alerts at the same time. Some are real failures and some are false alarms. The agent must decide what to do with limited vehicles, staff, and supplies while making sure priority clinics still get what they need. ## Observation The observation includes: - current task and step number - inventory at each site - expected demand at each site - priority sites - active alerts - site status: - power - backup - sensor temperature - sensor confidence - estimated risk that te …