Spaces:
Sleeping
Sleeping
File size: 12,328 Bytes
9f7b0e1 1bb7020 9f7b0e1 1bb7020 9f7b0e1 1bb7020 9f7b0e1 1bb7020 9f7b0e1 1bb7020 9f7b0e1 | 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 | import asyncio
import math
from typing import Union
from .models import (
AgentModel,
DeathEvent,
MessageEvent,
LeadershipVoteEvent,
LeaderElectedEvent,
WaterCollectedEvent,
FireExtinguishedEvent,
FireSpreadEvent,
SimulationState,
TickResponse,
ChatEntry,
)
from . import groq_client
from . import movement
FIRE_GROWTH_RATE = 1.0 # radius growth per tick
FIRE_INTENSITY_GROWTH = 0.9 # intensity per tick
BASE_EXTINGUISH_RATE = 15.0 # baseline intensity reduction per agent
MIN_EXTINGUISH_RATE = 8.0
MAX_EXTINGUISH_RATE = 28.0
TICK_INTERVAL_SECONDS = 3
WATER_PICKUP_RANGE = 40
EXTINGUISH_RANGE = 45
FIRE_SAFE_BUFFER = 10
class SimulationEngine:
def __init__(self, state: SimulationState) -> None:
self.state = state
def _normalize_message(self, message: str | None) -> str:
if not message:
return "Staying focused and moving."
cleaned = " ".join(str(message).strip().split())
if not cleaned:
return "Staying focused and moving."
return cleaned[:220]
def _move_toward(self, agent: AgentModel, target_x: float, target_y: float, stop_distance: float = 0) -> None:
dx = target_x - agent.x
dy = target_y - agent.y
dist = math.sqrt(dx**2 + dy**2) or 1
if dist <= stop_distance:
return
step = min(movement.MAX_AGENT_SPEED, dist - stop_distance)
agent.x += int((dx / dist) * step)
agent.y += int((dy / dist) * step)
agent.x = max(0, min(agent.x, self.state.map_width))
agent.y = max(0, min(agent.y, self.state.map_height))
async def tick(self) -> TickResponse:
"""
Main simulation loop:
1. Get decisions from all living agents
2. Handle coalition leadership voting
3. Execute agent actions (search water, collect, extinguish, escape, etc.)
4. Grow fire
5. Extinguish fire if agents with water are present
6. Kill agents in fire (but protect coalition members)
7. Check win condition
"""
if self.state.status != "running":
raise ValueError(f"Cannot tick a simulation with status '{self.state.status}'.")
fire = self.state.fire
assert fire is not None, "Fire must be placed before ticking."
events = []
bounds = (self.state.map_width, self.state.map_height)
living_agents = [a for a in self.state.agents if a.alive]
recent_radio = [
f"{a.display_name}: {a.last_message}"
for a in living_agents
if a.last_message
][-8:]
# 1. Get decisions from all living agents
decisions = await asyncio.gather(
*[groq_client.generate_fire_decision(agent, fire, self.state.water_sources, living_agents, bounds, recent_radio)
for agent in living_agents],
return_exceptions=True
)
decision_map = {}
for agent, decision in zip(living_agents, decisions):
if isinstance(decision, Exception):
decision = groq_client._fallback_escape(agent, fire)
decision_map[agent.model_name] = decision
# 2. Leadership voting phase (if coalition leader not elected)
if not self.state.coalition_leader:
vote_events = await self._voting_phase(living_agents, decision_map)
events.extend(vote_events)
# 3. Execute actions
action_events = await self._execute_actions(living_agents, decision_map, fire)
events.extend(action_events)
# 4. Grow fire
fire.radius += FIRE_GROWTH_RATE
fire.intensity += FIRE_INTENSITY_GROWTH
if fire.intensity > 100.0:
fire.intensity = 100.0
if fire.intensity > 0:
events.append(FireSpreadEvent(new_radius=fire.radius, new_intensity=fire.intensity))
# 5. Extinguish fire if agents with water are present
extinguish_events = self._check_extinguish(living_agents, fire)
events.extend(extinguish_events)
# 6. Kill agents in fire
death_events = self._kill_agents_in_fire(living_agents, fire)
events.extend(death_events)
# 7. Check win condition
self.state.round += 1
living_count = len([a for a in self.state.agents if a.alive])
if fire.intensity <= 0:
# Fire extinguished!
self.state.status = "finished"
top_score = max((a.extinguish_score for a in self.state.agents), default=0)
top_agents = [a.model_name for a in self.state.agents if a.extinguish_score == top_score and top_score > 0]
if top_agents:
self.state.winner_model = f"Top extinguisher: {', '.join(top_agents)} ({top_score:.1f} impact)"
else:
self.state.winner_model = "Fire extinguished"
elif living_count <= 1:
# Only one agent left
self.state.status = "finished"
winner = next((a.model_name for a in self.state.agents if a.alive), None)
self.state.winner_model = winner or "No survivors"
return TickResponse(
simulation_id=self.state.simulation_id,
round=self.state.round,
events=events,
chat=[],
state=self.state
)
async def _voting_phase(self, agents, decision_map):
"""
Agents vote for a coalition leader.
Get votes from LLM based on current situation.
"""
events = []
# Gather votes
votes = {} # candidate -> vote count
for agent in agents:
decision = decision_map.get(agent.model_name, {})
vote_for = decision.get("vote_for")
if vote_for:
votes[vote_for] = votes.get(vote_for, 0) + 1
events.append(LeadershipVoteEvent(voter=agent.model_name, candidate=vote_for))
# Elect leader if there are votes
if votes:
leader_name = max(votes, key=votes.get)
leader_agent = next((a for a in agents if a.model_name == leader_name), None)
if leader_agent:
for agent in agents:
agent.mode = "coalition"
leader_agent.is_leader = True
self.state.coalition_leader = leader_name
coalition = [a.model_name for a in agents if a.mode == "coalition"]
self.state.coalition_members = coalition
events.append(LeaderElectedEvent(leader=leader_name, coalition_members=coalition))
return events
async def _execute_actions(self, agents, decision_map, fire):
"""
Execute agent actions: search, collect water, extinguish, escape, vote, etc.
"""
events = []
chat_entries = []
for agent in agents:
decision = decision_map.get(agent.model_name, {})
action = decision.get("action", "escape")
message = self._normalize_message(decision.get("message"))
nearest_water = self._find_nearest_water(agent, self.state.water_sources)
dist_to_fire = math.dist((agent.x, agent.y), (fire.x, fire.y))
dist_to_water = None
if nearest_water:
dist_to_water = math.dist((agent.x, agent.y), (nearest_water.x, nearest_water.y))
# Guardrails to keep behavior consistent with visuals and objectives.
if dist_to_fire <= fire.radius + FIRE_SAFE_BUFFER:
action = "escape"
elif agent.water_collected:
action = "extinguish_fire"
elif dist_to_water is not None and dist_to_water <= WATER_PICKUP_RANGE:
action = "collect_water"
else:
action = "search_water"
if action == "collect_water":
water_source = nearest_water
if water_source and dist_to_water is not None:
dist_to_water = math.dist((agent.x, agent.y), (water_source.x, water_source.y))
if dist_to_water <= WATER_PICKUP_RANGE:
agent.water_collected = True
agent.status = "collecting_water"
events.append(WaterCollectedEvent(model=agent.model_name, water_source_id=water_source.id))
else:
agent.status = "searching"
self._move_toward(agent, water_source.x, water_source.y)
elif action == "extinguish_fire":
if agent.water_collected:
agent.status = "extinguishing_fire"
dist_to_fire = math.dist((agent.x, agent.y), (fire.x, fire.y))
target_dist = max(fire.radius + FIRE_SAFE_BUFFER, 0)
self._move_toward(agent, fire.x, fire.y, stop_distance=target_dist)
else:
agent.status = "searching"
message = self._normalize_message(decision.get("message"))
elif action == "search_water":
agent.status = "searching"
water_source = nearest_water
if water_source:
self._move_toward(agent, water_source.x, water_source.y)
elif action == "escape":
agent.status = "escaping"
# Move away from fire
dx = agent.x - fire.x
dy = agent.y - fire.y
dist = math.sqrt(dx**2 + dy**2) or 1
agent.x += int((dx / dist) * movement.MAX_AGENT_SPEED)
agent.y += int((dy / dist) * movement.MAX_AGENT_SPEED)
agent.x = max(0, min(agent.x, self.state.map_width))
agent.y = max(0, min(agent.y, self.state.map_height))
agent.last_message = message
events.append(MessageEvent(model=agent.model_name, content=message))
chat_entries.append(ChatEntry(agent_id=agent.model_name, message=message, tick=self.state.round))
return events
def _find_nearest_water(self, agent, water_sources):
"""Find the closest water source to an agent."""
if not water_sources:
return None
return min(water_sources, key=lambda w: math.dist((agent.x, agent.y), (w.x, w.y)))
def _check_extinguish(self, agents, fire):
"""Check if agents with water are extinguishing the fire."""
events = []
agents_with_water = []
for agent in agents:
if not (agent.water_collected and agent.status == "extinguishing_fire"):
continue
dist_to_fire = math.dist((agent.x, agent.y), (fire.x, fire.y))
if dist_to_fire <= fire.radius + EXTINGUISH_RANGE:
agents_with_water.append(agent)
if agents_with_water:
living_count = len([a for a in agents if a.alive]) or 1
scale = max(0.5, min(2.0, 2.0 / living_count))
per_agent_rate = BASE_EXTINGUISH_RATE * scale
per_agent_rate = max(MIN_EXTINGUISH_RATE, min(MAX_EXTINGUISH_RATE, per_agent_rate))
reduction = len(agents_with_water) * per_agent_rate
fire.intensity -= reduction
if fire.intensity < 0:
fire.intensity = 0
extinguisher_names = [a.model_name for a in agents_with_water]
events.append(FireExtinguishedEvent(extinguished_by=extinguisher_names, fire_intensity=fire.intensity))
for agent in agents_with_water:
agent.extinguish_score += per_agent_rate
agent.water_collected = False
return events
def _kill_agents_in_fire(self, agents, fire):
"""Check if agents are consumed by fire."""
events = []
for agent in agents:
if not agent.alive:
continue
dist_to_fire = math.dist((agent.x, agent.y), (fire.x, fire.y))
# Agent dies if inside fire radius
if dist_to_fire < fire.radius:
agent.alive = False
events.append(DeathEvent(model=agent.model_name))
events.append(MessageEvent(model=agent.model_name, content="No!!! The fire got me..."))
return events
|