esachdev12/CLINOVA

1import argparse2import asyncio3import json4import yaml5import time6import numpy as np7from datetime import datetime8from pathlib import Path9from tqdm import tqdm10from env.environment import FinomIQEnv11from env.agents import get_agent12from env.models import Action13from env.logging_config import setup_logger14 15logger = setup_logger("Runner")16 17 18class SimulationRunner:19    """Orchestrates automated runs for FinomIQ Hedge Fund Intelligence."""20 21    def __init__(self, config_path: str = "config.yaml"):22        with open(config_path, "r") as f:23            self.config = yaml.safe_load(f)24 25        self.env = FinomIQEnv(config_path)26        self.agent = get_agent(self.config["agent"]["type"], self.config)27        self.num_episodes = self.config["scenario"]["num_episodes"]28        self.results_path = Path(self.config["visualization"].get("persistence_path", "results/finomiq_run.json"))29        self.results_path.parent.mkdir(parents=True, exist_ok=True)30 31    async def run_episode(self, episode_idx: int) -> dict:32        """Execute a single episode simulation."""33        logger.info(f"═══════════════════════════════════════════")34        logger.info(f"EPISODE {episode_idx + 1}/{self.num_episodes} — START")35        logger.info(f"═══════════════════════════════════════════")36 37        observation = await self.env.reset()38        done = False39        episode_reward = 040        steps = 041        action_history = []42 43        while not done:44            action = self.agent.choose_action(observation)45            46            result = await self.env.step(action)47            observation = result["observation"]48            episode_reward += result["reward"]49            done = result["done"]50            steps += 151 52            action_history.append({53                "step": steps,54                "action": action.action_type,55                "asset": action.asset_name,56                "amount": action.amount,57                "reward": round(result["reward"], 4),58                "portfolio_value": observation["portfolio_value"],59                "unrealized_pnl": observation["unrealized_pnl"],60                "asset_prices": observation["asset_prices"].copy(),61            })62 63            if done:64                logger.info(f"───────────────────────────────────────────")65                logger.info(f"EPISODE {episode_idx + 1} RESULT: PnL=${observation['unrealized_pnl']:.2f} | Steps={steps}, Reward={episode_reward:.2f}")66                logger.info(f"───────────────────────────────────────────")67 68                return {69                    "episode": episode_idx + 1,70                    "reward": round(episode_reward, 2),71                    "steps": steps,72                    "final_portfolio_value": observation["portfolio_value"],73                    "unrealized_pnl": observation["unrealized_pnl"],74                    "action_history": action_history,75                    "observation": observation,76                    "history": self.env.history77                }78        return {}79 80    async def run_all(self):81        """Execute the batch of episodes and produce a rich summary."""82        start_time = time.time()83        logger.info(f"--- FinomIQ Autonomous Hedge Fund Simulation ---")84        logger.info(f"Market: {self.config['scenario']['market_type']} | Agent: {self.config['agent']['type']} | Episodes: {self.num_episodes}")85 86        results = []87        for i in tqdm(range(self.num_episodes)):88            res = await self.run_episode(i)89            results.append(res)90 91        elapsed = round(time.time() - start_time, 2)92 93        # ── Aggregate metrics ──94        rewards = [r["reward"] for r in results]95        pnls = [r["unrealized_pnl"] for r in results]96        steps_list = [r["steps"] for r in results]97 98        avg_reward = round(sum(rewards) / len(rewards) if rewards else 0, 2)99        avg_pnl = round(sum(pnls) / len(pnls) if pnls else 0, 2)100        avg_steps = round(sum(steps_list) / len(steps_list) if steps_list else 0, 1)101 102        summary = {103            "run_metadata": {104                "timestamp": datetime.now().isoformat(),105                "elapsed_seconds": elapsed,106                "market_type": self.config["scenario"]["market_type"],107                "agent_type": self.config["agent"]["type"],108                "num_episodes": self.num_episodes,109                "max_steps": self.config["scenario"]["max_steps"],110                "seed": self.config["scenario"]["seed"],111            },112            "metrics": {113                "avg_reward": avg_reward,114                "avg_pnl": avg_pnl,115                "avg_steps": avg_steps,116                "total_profit": sum(pnls),117            },118            "episodes": results,119            "config": self.config,120        }121 122        with open(self.results_path, "w") as f:123            json.dump(summary, f, indent=2)124 125        # ── Console summary ──126        logger.info(f"")127        logger.info(f"╔══════════════════════════════════════════════════════════╗")128        logger.info(f"║           FinomIQ SIMULATION SUMMARY REPORT             ║")129        logger.info(f"╠══════════════════════════════════════════════════════════╣")130        logger.info(f"║  Market Regime: {self.config['scenario']['market_type']:<42}║")131        logger.info(f"║  Agent Model:   {self.config['agent']['type']:<42}║")132        logger.info(f"║  Episodes:      {self.num_episodes:<42}║")133        logger.info(f"║  Avg PnL:       ${avg_pnl:<41}║")134        logger.info(f"║  Runtime:       {elapsed}s{' ' * (40 - len(str(elapsed)))}║")135        logger.info(f"╚══════════════════════════════════════════════════════════╝")136        logger.info(f"Results saved to: {self.results_path}")137 138 139if __name__ == "__main__":140    parser = argparse.ArgumentParser()141    parser.add_argument("--config", type=str, default="config.yaml")142    args = parser.parse_args()143 144    runner = SimulationRunner(args.config)145    asyncio.run(runner.run_all())146