Implementation:HKUDS AI Trader Main Parallel

Knowledge Sources	AI-Trader
Domains	Concurrency, Backtesting
Last Updated	2026-02-09 14:00 GMT

Overview

Concrete tool for running multiple LLM trading agents in parallel subprocesses using asyncio.

Description

The main_parrallel.py module provides three key functions: main() loads config and orchestrates execution, _spawn_model_subprocesses() creates one asyncio subprocess per enabled model, and _run_model_in_current_process() handles single-model execution. The parallel runner uses asyncio.create_subprocess_exec() to spawn isolated Python processes, each re-invoking main_parrallel.py with a --signature flag.

Usage

Run from command line: python main_parrallel.py [--config path] to launch all enabled models in parallel. For programmatic use, call await main(config_path).

Code Reference

Source Location

Repository: AI-Trader
File: main_parrallel.py
Lines: L191-263 (main), L171-188 (_spawn_model_subprocesses), L100-168 (_run_model_in_current_process)

Signature

async def main(config_path=None, only_signature: str | None = None):
    """
    Main entry point for parallel agent execution.

    Args:
        config_path: Path to JSON config (default: configs/default_config.json)
        only_signature: If set, only run the model with this signature

    Returns:
        None. Each agent writes its own position.jsonl and logs.
    """

async def _spawn_model_subprocesses(config_path, enabled_models):
    """Spawn one subprocess per enabled model and wait for all to complete."""

async def _run_model_in_current_process(AgentClass, model_config, INIT_DATE, END_DATE, agent_config, log_config):
    """Run a single model in the current process (no subprocess)."""

Import

# CLI usage:
# python main_parrallel.py --config configs/default_config.json

# Programmatic:
from main_parrallel import main
import asyncio
asyncio.run(main("configs/default_config.json"))

I/O Contract

Inputs

Name	Type	Required	Description
config_path	Optional[str]	No	Path to config JSON (default: configs/default_config.json)
only_signature	Optional[str]	No	Single-model filter (used by subprocess mode)
INIT_DATE	str (env)	No	Override start date
END_DATE	str (env)	No	Override end date

Outputs

Name	Type	Description
position.jsonl	File (per agent)	Each agent writes its own position log
log/	Directory (per agent)	Each agent writes its own conversation logs

Usage Examples

Run All Enabled Models

import asyncio
from main_parrallel import main

# Run all enabled models from default config in parallel
asyncio.run(main())

From Shell

# Run with default config
python main_parrallel.py

# Run with custom config
python main_parrallel.py --config configs/astock_config.json

Related Pages

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment