How to Initialize the AutoHedge System with Custom Parameters

TLDR: The AutoHedge class in autohedge/main.py accepts optional constructor arguments—including name, description, output_dir, output_file_path, and output_type—that let you customize the system's identity, file locations, and return formats before running autonomous trading cycles.

The AutoHedge system from The-Swarm-Corporation/AutoHedge repository provides a fully autonomous hedge fund framework. Learning how to initialize the AutoHedge system with custom parameters allows you to tailor output destinations, naming conventions, and data formats to fit your deployment pipeline. The entry point is the AutoHedge class constructor located in autohedge/main.py, which configures the logger and conversation handlers before orchestrating the agent swarm.

AutoHedge Constructor Parameters

The AutoHedge constructor defined on lines 9-22 of autohedge/main.py exposes five optional parameters that control instance behavior:

• name – Identifier for the hedge fund instance. Defaults to "autohedge".
• description – Human-readable description of the system. Defaults to "fully autonomous hedgefund".
• output_dir – Directory path where generated files are stored. The system creates this directory automatically if it does not exist. Defaults to "outputs".
• output_file_path – Optional explicit file path for single-file output. When provided, the system writes to this specific location instead of the output_dir folder. Defaults to None.
• output_type – Controls the return format of the run() method. Accepts "list", "dict", or "str". Defaults to "list".

Initialization Internals

During instantiation, the constructor performs critical setup operations on lines 30-32 of autohedge/main.py. It initializes a Loguru logger for runtime event tracking and creates a Swarms Conversation object that logs dialogue between agents throughout the trading cycle. These components are essential for monitoring the autonomous decision-making process and debugging agent interactions.

Output Configuration and Return Formats

The output_type parameter directly influences the return value processing logic found on lines 52-60 of autohedge/main.py. Selecting "list" returns conversation turns as a list structure, "dict" returns a dictionary mapping participants to responses, and "str" returns a plain text transcript. When you specify output_file_path, the system writes results to that explicit location rather than generating filenames within output_dir.

Practical Initialization Examples

Use these patterns to instantiate AutoHedge with different configuration profiles.

Default initialization with no custom arguments:

from autohedge import AutoHedge

system = AutoHedge()
print(system.run("Analyze AAPL for 20k allocation"))

Custom identity with dictionary output format:

from autohedge import AutoHedge

system = AutoHedge(
    name="my_custom_hedge",
    description="My bespoke autonomous fund",
    output_type="dict",
    output_dir="my_outputs",
)
result = system.run("Analyze TSLA for 50k allocation")
print(result)  # → {'user': ..., 'director': ..., ...}

Directing output to a specific file path:

from autohedge import AutoHedge

system = AutoHedge(
    output_file_path="reports/tsla_report.txt",
    output_type="str",
)

report = system.run("Analyze TSLA for 50k allocation")
with open(system.output_file_path, "w") as f:
    f.write(report)

Integration with the Agent Pipeline

After initialization, the AutoHedge instance manages the task workflow by interfacing with the Director Agent defined in autohedge/workers.py (lines 1-20). This agent receives the task string and drives the remainder of the analysis pipeline. For command-line deployments, autohedge/cli.py provides the run_repl function (lines 73-77) which creates a default AutoHedge instance, though you can bypass the CLI entirely by instantiating the class directly in your Python scripts. Environment variables required for agent operation are loaded via autohedge/env_loader.py.

Summary

• The AutoHedge class accepts five key parameters—name, description, output_dir, output_file_path, and output_type—to customize initialization.
• The output_type parameter controls whether the run() method returns a list, dictionary, or plain string according to lines 52-60 of autohedge/main.py.
• Custom output_dir and output_file_path values let you redirect file operations to specific locations, with automatic directory creation handled by the constructor.
• The initialization process sets up Loguru logging and a Swarms Conversation object on lines 30-32 of autohedge/main.py.
• Initialized instances pass tasks to the Director Agent defined in autohedge/workers.py to execute the autonomous trading analysis.

Frequently Asked Questions

What is the default output format when initializing AutoHedge?

The default output_type is "list", which returns conversation turns as a list structure. You can override this with "dict" for dictionary formatting or "str" for plain string output suitable for file writing.

How do I specify a custom directory for AutoHedge output files?

Pass the output_dir parameter to the constructor with your desired path string. The system automatically creates the directory if it does not exist. Alternatively, use output_file_path to specify an exact file location instead of a folder when you need precise control over the destination.

What dependencies does AutoHedge initialize during construction?

According to autohedge/main.py lines 30-32, the constructor sets up a Loguru logger for runtime events and a Swarms Conversation object to track dialogue between agents. These dependencies are essential for the logging and agent coordination pipeline.

Can I use AutoHedge without the command-line interface?

Yes. While autohedge/cli.py provides a run_repl function for CLI usage (lines 73-77), you can instantiate AutoHedge directly in Python scripts with custom parameters and call the run() method programmatically to integrate the system into larger applications.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}