UserSyncUI / tinytroupe /simulation_manager.py
AUXteam's picture
Upload folder using huggingface_hub
f285b41 verified
Raw
History Blame Contribute Delete
12.1 kB
from typing import List, Dict, Any, Optional
import uuid
import threading
import random
import requests
import json
from datetime import datetime
from tinytroupe.agent import TinyPerson
from tinytroupe.social_network import NetworkTopology
from tinytroupe.environment.social_tiny_world import SocialTinyWorld, SimulationResult
from tinytroupe.agent.social_types import Content
from tinytroupe.ml_models import EngagementPredictor
from tinytroupe.content_generation import ContentVariantGenerator
from tinytroupe.network_generator import NetworkGenerator
import tinytroupe.openai_utils as openai_utils
from tinytroupe import utils
config = utils.read_config_file()
from tinytroupe.agent import logger
class SimulationConfig:
def __init__(self, name: str, persona_count: int = 10, network_type: str = "scale_free", **kwargs):
self.name = name
self.persona_count = persona_count
self.network_type = network_type
self.user_id = kwargs.get("user_id")
class Simulation:
def __init__(self, id: str, config: SimulationConfig, world: SocialTinyWorld, personas: List[TinyPerson], network: NetworkTopology):
self.id = id
self.config = config
self.world = world
self.personas = personas
self.network = network
self.status = "ready"
self.created_at = datetime.now()
self.last_result: Optional[SimulationResult] = None
self.chat_history: List[Dict[str, Any]] = []
self.progress = 0.0
self.analysis_results: List[Dict[str, Any]] = []
class SimulationManager:
"""Manages simulation lifecycle and execution with remote load balancing"""
def __init__(self):
self.simulations: Dict[str, Simulation] = {}
self.focus_groups: Dict[str, List[TinyPerson]] = {}
self.predictor = EngagementPredictor()
self.variant_generator = ContentVariantGenerator()
self.remote_url = "https://auxteam-tiny-factory.hf.space"
def _call_remote_api(self, api_name: str, payload: List[Any]) -> Any:
"""Call remote backend API"""
try:
logger.info(f"Calling remote API: {api_name}")
response = requests.post(
f"{self.remote_url}/call/{api_name}",
json={"data": payload},
timeout=300
)
response.raise_for_status()
event_id = response.json().get("event_id")
# Poll for result
import time
while True:
res = requests.get(f"{self.remote_url}/call/{api_name}/{event_id}")
res.raise_for_status()
# Gradio SSE output parsing or simplified pooling
# This is a bit complex for a simple script,
# let's try the simpler /api/predict approach if available
# Or use the legacy /api/ route if it works
break
# For simplicity, let's assume a direct POST to /api/predict works for some endpoints
# or just use the local one if remote fails.
return None
except Exception as e:
logger.error(f"Error calling remote API {api_name}: {e}")
return None
def create_simulation(self, config: SimulationConfig, focus_group_name: str = None) -> Simulation:
if focus_group_name and focus_group_name in self.focus_groups:
personas = self.focus_groups[focus_group_name]
else:
# Decide whether to generate locally or remotely
if random.random() < 0.3: # 30% chance for remote persona generation
logger.info("Decided to generate personas remotely (Load balancing)")
# Placeholder for remote call logic
pass
from tinytroupe.factory.tiny_person_factory import TinyPersonFactory
factory = TinyPersonFactory(
context=config.name,
total_population_size=config.persona_count
)
personas = factory.generate_people(number_of_people=config.persona_count)
# Generate network
net_gen = NetworkGenerator(personas)
if config.network_type == "scale_free":
network = net_gen.generate_scale_free_network(config.persona_count, 2)
else:
network = net_gen.generate_small_world_network(config.persona_count, 4, 0.1)
# Create world
world = SocialTinyWorld(config.name, network=network)
for persona in personas:
world.add_agent(persona)
sim_id = str(uuid.uuid4())
simulation = Simulation(sim_id, config, world, personas, network)
self.simulations[sim_id] = simulation
return simulation
def run_simulation(self, simulation_id: str, content: Content, mode: str = "full", background: bool = False) -> Optional[SimulationResult]:
if simulation_id not in self.simulations:
raise ValueError(f"Simulation {simulation_id} not found.")
simulation = self.simulations[simulation_id]
if background:
thread = threading.Thread(target=self._run_simulation_task, args=(simulation, content))
thread.start()
return None
else:
return self._run_simulation_task(simulation, content)
def _run_simulation_task(self, simulation: Simulation, content: Content) -> SimulationResult:
simulation.status = "running"
simulation.progress = 0.1
initial_viewers = [p.name for p in simulation.personas[:5]] # Seed with first 5
# In a real async scenario, simulate_content_spread would update progress
result = simulation.world.simulate_content_spread(content, initial_viewers)
simulation.status = "completed"
simulation.progress = 1.0
simulation.last_result = result
# Automatically trigger analysis using alias-huge
self.analyze_simulation_opinions(simulation.id)
return result
def analyze_simulation_opinions(self, simulation_id: str) -> List[Dict[str, Any]]:
"""Analyze simulation results using Helmholtz alias-huge model"""
sim = self.get_simulation(simulation_id)
if not sim or not sim.last_result: return []
logger.info(f"Analyzing simulation {simulation_id} using Helmholtz alias-huge")
analysis_results = []
# We take a sample of engagements to analyze
engagements = sim.last_result.engagements
for eng in engagements:
if eng["type"] == "comment" or (eng["type"] == "none" and eng["feedback"]):
persona_name = eng["persona_id"]
opinion = eng["feedback"]
prompt = f"""
Analyze the following opinion from {persona_name} regarding the content.
Content: {sim.last_result.content.text}
Opinion: {opinion}
Provide a structured analysis and direct implications for the business.
Return ONLY a JSON object with the following keys:
- persona_name: the name of the persona
- opinion: the original opinion
- analysis: your psychological and social analysis
- implications: direct business implications
"""
try:
response = openai_utils.client().send_message(
[{"role": "user", "content": prompt}],
model=config["OpenAI"].get("FALLBACK_MODEL_HUGE", "alias-huge"),
temperature=0.7
)
# Try to extract JSON from response
content_str = response["content"]
if "```json" in content_str:
content_str = content_str.split("```json")[1].split("```")[0].strip()
elif "```" in content_str:
content_str = content_str.split("```")[1].split("```")[0].strip()
result = json.loads(content_str)
analysis_results.append(result)
except Exception as e:
logger.error(f"Error during alias-huge analysis for {persona_name}: {e}")
analysis_results.append({
"persona_name": persona_name,
"opinion": opinion,
"analysis": "Error during analysis",
"implications": "N/A"
})
sim.analysis_results = analysis_results
return analysis_results
def send_chat_message(self, simulation_id: str, sender: str, message: str) -> Dict[str, Any]:
sim = self.get_simulation(simulation_id)
if not sim: raise ValueError(f"Simulation {simulation_id} not found.")
msg = {
"sender": sender,
"message": message,
"timestamp": datetime.now().isoformat()
}
sim.chat_history.append(msg)
# Trigger persona responses if it's a "User" message
if sender == "User":
# For now, pick a random persona to respond
responder = random.choice(sim.personas)
response_text = f"As a {responder._persona.get('occupation')}, I think: {message[:20]}... sounds interesting!"
response_msg = {
"sender": responder.name,
"message": response_text,
"timestamp": datetime.now().isoformat()
}
sim.chat_history.append(response_msg)
return msg
def get_chat_history(self, simulation_id: str) -> List[Dict[str, Any]]:
sim = self.get_simulation(simulation_id)
if not sim: return []
return sim.chat_history
def get_simulation(self, simulation_id: str, user_id: str = None) -> Optional[Simulation]:
return self.simulations.get(simulation_id)
def list_simulations(self) -> List[Dict[str, Any]]:
return [
{
"id": sim.id,
"name": sim.config.name,
"status": sim.status,
"persona_count": len(sim.personas),
"created_at": sim.created_at.isoformat()
}
for sim in self.simulations.values()
]
def get_persona(self, simulation_id: str, persona_name: str) -> Optional[Dict[str, Any]]:
sim = self.get_simulation(simulation_id)
if not sim: return None
for p in sim.personas:
if p.name == persona_name:
return p._persona
return None
def list_personas(self, simulation_id: str) -> List[Dict[str, Any]]:
sim = self.get_simulation(simulation_id)
if not sim: return []
return [p._persona for p in sim.personas]
def save_focus_group(self, name: str, personas: List[TinyPerson]):
self.focus_groups[name] = personas
def list_focus_groups(self) -> List[str]:
return list(self.focus_groups.keys())
def get_focus_group(self, name: str) -> Optional[List[TinyPerson]]:
return self.focus_groups.get(name)
def delete_simulation(self, simulation_id: str) -> bool:
if simulation_id in self.simulations:
del self.simulations[simulation_id]
return True
return False
def export_simulation(self, simulation_id: str) -> Optional[Dict[str, Any]]:
sim = self.get_simulation(simulation_id)
if not sim: return None
return {
"id": sim.id,
"config": {
"name": sim.config.name,
"persona_count": sim.config.persona_count,
"network_type": sim.config.network_type
},
"status": sim.status,
"created_at": sim.created_at.isoformat(),
"personas": [p._persona for p in sim.personas],
"network": sim.network.get_metrics(),
"analysis_results": sim.analysis_results
}