Spaces:
Running
Running
File size: 13,658 Bytes
3eae4cc | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 | """
Gymnasium adapter for GovWorkflowEnv.
Key contract:
observation_space : Box(OBS_DIM,) float32
action_space : Discrete(N_ACTIONS)
action_masks() : np.ndarray[bool, N_ACTIONS]
"""
from __future__ import annotations
from typing import Any, Callable, Optional
import gymnasium as gym
import numpy as np
from gymnasium import spaces
from app.env import GovWorkflowEnv
from app.models import ActionModel, ActionType, ObservationModel, PriorityMode, ServiceType
from rl.action_mask import ActionMaskComputer
from rl.feature_builder import ACTION_DECODE_TABLE, N_ACTIONS, OBS_DIM, FeatureBuilder
class GovWorkflowGymEnv(gym.Env):
metadata = {"render_modes": []}
def __init__(
self,
task_id: str = "district_backlog_easy",
seed: int = 42,
hard_action_mask: bool = False,
max_non_advance_streak: int = 3,
):
super().__init__()
self.task_id = task_id
self._seed = seed
self._task_sampler: Optional[Callable[[], str]] = None
self._global_step_counter: Optional[list[int]] = None
self._hard_action_mask: bool = bool(hard_action_mask)
self._max_non_advance_streak = max(0, int(max_non_advance_streak))
self._non_advance_streak = 0
self._core_env = GovWorkflowEnv()
self._fb = FeatureBuilder()
self._amc = ActionMaskComputer()
self.observation_space = spaces.Box(
low=0.0,
high=1.0,
shape=(OBS_DIM,),
dtype=np.float32,
)
self.action_space = spaces.Discrete(N_ACTIONS)
self._current_obs: Optional[ObservationModel] = None
self._current_pm: str = "balanced"
self._last_at: str = "advance_time"
@property
def core_env(self) -> GovWorkflowEnv:
return self._core_env
def set_hard_action_mask(self, enabled: bool) -> None:
self._hard_action_mask = bool(enabled)
def set_task_sampler(
self,
task_sampler: Optional[Callable[[], str]],
global_step_counter: Optional[list[int]] = None,
) -> None:
self._task_sampler = task_sampler
self._global_step_counter = global_step_counter
def reset(
self,
seed: Optional[int] = None,
options: Optional[dict] = None,
) -> tuple[np.ndarray, dict]:
super().reset(seed=seed)
if self._task_sampler is not None:
self.task_id = self._task_sampler()
use_seed = seed if seed is not None else self._seed
task_opts = {"task_id": self.task_id}
if options:
task_opts.update(options)
obs_model, info = self._core_env.reset(seed=use_seed, options=task_opts)
self._current_obs = obs_model
self._current_pm = "balanced"
self._last_at = "advance_time"
self._non_advance_streak = 0
info_dict = info.model_dump() if hasattr(info, "model_dump") else info
if not isinstance(info_dict, dict):
try:
info_dict = dict(info_dict)
except (TypeError, ValueError):
info_dict = {}
info_dict["fairness_gap"] = self._obs_fairness_gap(obs_model)
return self._to_array(obs_model), info_dict
def step(self, action: int) -> tuple[np.ndarray, float, bool, bool, dict]:
requested_action_idx = int(action)
action_idx = requested_action_idx
if self._hard_action_mask and self._current_obs is not None:
action_idx = self._sanitize_action_idx(requested_action_idx, self.action_masks())
action_model = self._decode_action(action_idx)
obs_model, reward, terminated, truncated, info = self._core_env.step(action_model)
if self._global_step_counter is not None:
self._global_step_counter[0] += 1
self._current_obs = obs_model
self._last_at = action_model.action_type.value
if getattr(action_model, "priority_mode", None) is not None:
self._current_pm = action_model.priority_mode.value
if action_model.action_type == ActionType.ADVANCE_TIME:
self._non_advance_streak = 0
else:
self._non_advance_streak += 1
info_dict = info.model_dump() if hasattr(info, "model_dump") else info
if not isinstance(info_dict, dict):
try:
info_dict = dict(info_dict)
except (TypeError, ValueError):
info_dict = {}
info_dict["fairness_gap"] = self._obs_fairness_gap(obs_model)
info_dict["requested_action_idx"] = requested_action_idx
info_dict["executed_action_idx"] = action_idx
info_dict["action_mask_applied"] = bool(action_idx != requested_action_idx)
return self._to_array(obs_model), float(reward), terminated, truncated, info_dict
def action_masks(self) -> np.ndarray:
if self._current_obs is None:
return np.ones(N_ACTIONS, dtype=bool)
mask = self._amc.compute(self._current_obs, self._current_pm)
if self._max_non_advance_streak > 0 and self._non_advance_streak >= self._max_non_advance_streak:
forced = np.zeros(N_ACTIONS, dtype=bool)
forced[18] = True
return forced
return mask
def render(self) -> None:
return None
def _to_array(self, obs: ObservationModel) -> np.ndarray:
return self._fb.build(obs, self._current_pm, self._last_at)
def _queue_snapshot_iter(self) -> list[Any]:
if self._current_obs is None:
return []
raw = getattr(self._current_obs, "queue_snapshots", [])
if isinstance(raw, dict):
return list(raw.values())
if isinstance(raw, list):
return list(raw)
try:
return list(raw)
except Exception:
return []
def _queue_service(self, snap: Any) -> Optional[ServiceType]:
value = getattr(snap, "service_type", None) or getattr(snap, "service", None)
if value is None:
return None
if isinstance(value, ServiceType):
return value
try:
return ServiceType(str(value))
except Exception:
return None
def _queue_active_cases(self, snap: Any) -> int:
return int(getattr(snap, "total_pending", getattr(snap, "active_cases", 0)) or 0)
def _queue_urgent_cases(self, snap: Any) -> int:
return int(getattr(snap, "urgent_pending", getattr(snap, "urgent_cases", 0)) or 0)
def _obs_fairness_gap(self, obs: ObservationModel) -> float:
"""
Canonical fairness signal for RL info payload.
Current ObservationModel exposes fairness as `fairness_index`, while
episode-level grading uses `fairness_gap` from EpisodeStateModel.
Keep backward-compatible fallback to avoid runtime breaks.
"""
return float(getattr(obs, "fairness_gap", getattr(obs, "fairness_index", 0.0)) or 0.0)
def _build_action_model(self, action_type: ActionType, **kwargs: Any) -> ActionModel:
service = kwargs.get("service")
target_service = kwargs.get("target_service")
officer_delta = int(kwargs.get("officer_delta", 1) or 1)
priority_mode = kwargs.get("priority_mode")
candidates: list[dict[str, Any]] = []
if action_type == ActionType.ADVANCE_TIME:
candidates.append({"action_type": action_type})
elif action_type == ActionType.SET_PRIORITY_MODE:
candidates.append({"action_type": action_type, "priority_mode": priority_mode})
elif action_type == ActionType.ASSIGN_CAPACITY and service is not None:
candidates.extend(
[
{"action_type": action_type, "service": service, "officer_delta": officer_delta},
{"action_type": action_type, "service_target": service, "officer_delta": officer_delta},
{"action_type": action_type, "capacity_assignment": {service.value: officer_delta}},
]
)
elif action_type == ActionType.REQUEST_MISSING_DOCUMENTS and service is not None:
candidates.extend(
[
{"action_type": action_type, "service": service},
{"action_type": action_type, "service_target": service},
]
)
elif action_type == ActionType.ESCALATE_SERVICE and service is not None:
candidates.extend(
[
{"action_type": action_type, "service": service},
{"action_type": action_type, "service_target": service},
{"action_type": action_type, "escalation_target": service},
]
)
elif action_type == ActionType.REALLOCATE_OFFICERS and service is not None and target_service is not None:
candidates.extend(
[
{
"action_type": action_type,
"service": service,
"target_service": target_service,
"officer_delta": officer_delta,
},
{
"action_type": action_type,
"reallocation_delta": {
service.value: -officer_delta,
target_service.value: officer_delta,
},
},
]
)
for candidate in candidates:
try:
return ActionModel(**candidate)
except Exception:
continue
return ActionModel(action_type=ActionType.ADVANCE_TIME)
def _decode_action(self, action_idx: int) -> ActionModel:
if action_idx not in ACTION_DECODE_TABLE:
return ActionModel(action_type=ActionType.ADVANCE_TIME)
action_type_str, service_str, priority_mode_str, delta = ACTION_DECODE_TABLE[action_idx]
action_type = ActionType(action_type_str)
if action_type == ActionType.SET_PRIORITY_MODE and priority_mode_str is not None:
return self._build_action_model(
action_type,
priority_mode=PriorityMode(priority_mode_str),
)
if action_type == ActionType.ASSIGN_CAPACITY:
if service_str == "__most_loaded__":
target = self._find_most_loaded_service()
elif service_str == "__most_urgent__":
target = self._find_most_urgent_service()
else:
target = ServiceType(service_str) if service_str and not service_str.startswith("__") else None
if target is None:
return ActionModel(action_type=ActionType.ADVANCE_TIME)
return self._build_action_model(
action_type,
service=target,
officer_delta=max(int(delta or 1), 1),
)
if action_type == ActionType.REQUEST_MISSING_DOCUMENTS:
target = ServiceType(service_str) if service_str and not service_str.startswith("__") else self._find_most_loaded_service()
if target is None:
return ActionModel(action_type=ActionType.ADVANCE_TIME)
return self._build_action_model(action_type, service=target)
if action_type == ActionType.ESCALATE_SERVICE:
target = ServiceType(service_str) if service_str and not service_str.startswith("__") else self._find_most_urgent_service()
if target is None:
return ActionModel(action_type=ActionType.ADVANCE_TIME)
return self._build_action_model(action_type, service=target)
if action_type == ActionType.REALLOCATE_OFFICERS:
source = ServiceType(service_str)
target = self._find_reallocation_target(source)
if target is None:
return ActionModel(action_type=ActionType.ADVANCE_TIME)
return self._build_action_model(
action_type,
service=source,
target_service=target,
officer_delta=1,
)
return ActionModel(action_type=ActionType.ADVANCE_TIME)
def _find_most_loaded_service(self) -> Optional[ServiceType]:
snaps = self._queue_snapshot_iter()
if not snaps:
return None
best = max(snaps, key=self._queue_active_cases)
return self._queue_service(best)
def _find_most_urgent_service(self) -> Optional[ServiceType]:
snaps = [snap for snap in self._queue_snapshot_iter() if self._queue_urgent_cases(snap) > 0]
if not snaps:
return None
best = max(snaps, key=lambda snap: (self._queue_urgent_cases(snap), self._queue_active_cases(snap)))
return self._queue_service(best)
def _find_reallocation_target(self, source: ServiceType) -> Optional[ServiceType]:
snaps = [snap for snap in self._queue_snapshot_iter() if self._queue_service(snap) != source]
if not snaps:
return None
best = max(snaps, key=self._queue_active_cases)
if self._queue_active_cases(best) <= 0:
return None
return self._queue_service(best)
def _sanitize_action_idx(self, action_idx: int, masks: np.ndarray) -> int:
if 0 <= action_idx < N_ACTIONS and bool(masks[action_idx]):
return action_idx
if 0 <= 18 < N_ACTIONS and bool(masks[18]):
return 18
valid = np.flatnonzero(masks)
if valid.size == 0:
return 18
return int(valid[0])
|