name: cascade-containment
version: "1.0.0"
description: >
  An RL benchmark for epidemic containment policy under uncertainty.
  A city health authority must allocate limited resources across districts
  to contain a spreading outbreak — with delayed data, resource scarcity,
  and cascading hospital stress. Generalises to wildfire deployment,
  cyberattack isolation, and misinformation containment.

author: SST-Team
license: MIT

server:
  module:     server.app
  app:        app
  port:       7860
  dockerfile: Dockerfile

action:
  type: object
  class: ContainmentAction
  fields:
    action_type:
      type: string
      description: "One of: 'test', 'restrict', 'allocate'"
      enum: [test, restrict, allocate]
    district_id:
      type: integer
      description: "Target district index (0-indexed)"
      minimum: 0

observation:
  type: object
  class: CityObservation
  fields:
    districts:
      type: array
      description: "Per-district state visible to agent"
      items:
        type: object
        fields:
          district_id:
            type: integer
          reported_infection_rate:
            type: number
            minimum: 0.0
            maximum: 1.0
          growth_rate_hint:
            type: number
            minimum: 0.0
            maximum: 1.0
          hospital_capacity_remaining:
            type: number
            minimum: 0.0
            maximum: 1.0
          population_density:
            type: number
            minimum: 0.0
            maximum: 1.0
          tested_recently:
            type: boolean
          restriction_active:
            type: boolean
    available_resources:
      type: integer
      description: "Resource units remaining this turn"
    current_step:
      type: integer
    max_steps:
      type: integer
    done:
      type: boolean
    reward:
      type: number
      nullable: true
    data_lag_days:
      type: integer
      description: "Reporting lag in days (0 = real-time, 3 = hard task)"
    message:
      type: string
      nullable: true

tasks:
  - name: easy
    description: "2 districts, 1 outbreak seeded in D1, real-time data"
    max_steps: 10
    num_districts: 2

  - name: medium
    description: "4 districts, 2 simultaneous outbreaks, limited resources"
    max_steps: 15
    num_districts: 4

  - name: hard
    description: "6 districts, 3-day data lag, scarce resources"
    max_steps: 15
    num_districts: 6

tags:
  - reinforcement-learning
  - resource-allocation
  - sequential-decision-making
  - epidemic-containment
  - cascade-dynamics
  - partial-observability
  - openenv