# models.py # ============================================================================= # Typed Pydantic models for PrecisionIrrigation-v0 OpenEnv environment. # Represents a Punjab rice-paddy irrigation simulation. # ============================================================================= from pydantic import BaseModel, Field from typing import List, Optional, Literal from openenv.core.env_server import Action, Observation, State # -- Action -------------------------------------------------------------------- class IrrigationAction(Action): """ Discrete action for the irrigation agent. Action space: Discrete(5) 0 = no_water - No irrigation today. 1 = low - Low irrigation (10% pump capacity, 50 L). 2 = medium - Medium irrigation (30% pump capacity, 150 L). 3 = high - High irrigation (50% pump capacity, 250 L). 4 = drain - Drain excess water (lowers soil moisture by 0.15). """ action: int = Field( ..., ge=0, le=4, description=( "0=no_water | 1=low_50L | 2=medium_150L | 3=high_250L | 4=drain" ), ) # -- Observation sub-models ---------------------------------------------------- class DayForecast(BaseModel): """Single day weather forecast from mock IMD API.""" day_offset: int # +1, +2, or +3 days ahead rain_mm: float # predicted rainfall in mm temp_celsius: float # predicted temperature degC humidity_pct: float # relative humidity (0-100) power_cut_risk: float # probability of power cut today (0-1) class IrrigationObservation(Observation): """ Full observation returned from the Punjab irrigation environment. Dict-structured for Gymnasium Dict observation space compatibility. """ episode_id: Optional[str] = None # -- Core agronomic state ------------------------------------- day: int # current simulation day (1-indexed) total_days: int # episode length soil_moisture: float # 0.0-1.0 (fraction of field capacity) crop_growth: float # 0.0-1.0 (yield potential realized) water_reserve: float # litres remaining in tube-well tank power_status: int # 1 = grid ON, 0 = power CUT # -- Weather -------------------------------------------------- current_rain_mm: float current_temp_celsius: float weather_forecast: List[DayForecast] # next 3 days season: str # pre_monsoon/monsoon/post_monsoon/rabi # -- Episode tracking ----------------------------------------- cumulative_water_used: float # total litres irrigated so far stress_days: int # days where crop experienced moisture stress last_action_result: str # human-readable outcome of last action water_efficiency_index: float # yield/water ratio vs baseline (>1.2 = good) # -- LLM agent context ---------------------------------------- echoed_message: str # full natural-language state description # -- State (internal metadata) ------------------------------------------------- class IrrigationState(State): """Internal episode metadata, returned by /state endpoint.""" task_name: str = "" day: int = 0 total_days: int = 30 seed: int = 42 scenario: str = "monsoon" # monsoon/pre_monsoon/drought cumulative_reward: float = 0.0 crop_growth: float = 0.0 soil_moisture: float = 0.0 water_reserve: float = 0.0 cumulative_water_used: float = 0.0