Where to Find OpenMed Example Configurations: A Complete Guide to Setup Files

OpenMed stores its example configurations as TOML files in the openmed/profiles/ directory and ships a root-level config.toml.example that you can copy to ~/.config/openmed/config.toml to initialize your environment.

The maziyarpanahi/openmed repository provides ready-to-use configuration templates that demonstrate how to structure the medical NLP library for different deployment scenarios. These OpenMed example configurations control critical runtime parameters—including device selection, authentication tokens, and logging levels—through a profile-based system defined in the source code.

Configuration Reference and Schema

The authoritative definition of all supported configuration keys resides in openmed/core/config.py. This module contains the OpenMedConfig class, which specifies valid fields including default_org, cache_dir, device, hf_token, log_level, timeout, use_medical_tokenizer, medical_tokenizer_exceptions, backend, and profile. The file also defines PROFILE_PRESETS, which supplies fallback default values when specific keys are omitted from user-defined configuration files.

For detailed explanations of each field and advanced setup scenarios, consult docs/configuration.md in the repository root. This documentation walks through the complete configuration layout and profile inheritance mechanism.

Pre-Built Profile Templates

The repository ships three environment-specific profiles in the openmed/profiles/ directory that you can copy directly into your local configuration:

Development Profile (dev.toml)

Located at openmed/profiles/dev.toml, this example enables debug-level logging and extended timeouts. It is optimized for local development workflows where verbose output and longer execution windows facilitate troubleshooting.

Production Profile (prod.toml)

The openmed/profiles/prod.toml file provides deployment-ready settings with warning-level logging and shorter timeouts. This configuration minimizes log volume and ensures strict timeout enforcement for production workloads.

Test Profile (test.toml)

Found at openmed/profiles/test.toml, this minimal configuration demonstrates how to disable the medical tokenizer toggle via the use_medical_tokenizer key. The project's test suite uses this profile to ensure deterministic behavior during automated testing.

The Root Configuration Template

The openmed/config.toml.example file serves as the master template for the main configuration file. You should copy this file to ~/.config/openmed/config.toml (or the path specified by the OPENMED_CONFIG environment variable) to establish your base settings. This top-level file determines which profile is active by default and can override any values defined in the built-in presets.

How Configuration Loading Works

When you import OpenMed, the openmed.core.config.get_config() function automatically searches for the configuration file at ~/.config/openmed/config.toml by default. You can override this location by setting the OPENMED_CONFIG environment variable to point to a specific file path.

The system supports profile activation through two mechanisms: setting the OPENMED_PROFILE environment variable (e.g., export OPENMED_PROFILE=dev), or programmatically calling OpenMedConfig.from_profile("<name>") in your Python code. The loader searches for profile files in the profiles/ subdirectory within your configuration directory and applies them over the base configuration.

End-to-End Implementation Examples

To observe how these configurations integrate into working applications, examine the example scripts that demonstrate the complete initialization flow:

• examples/privacy_filter_unified.py demonstrates loading a configuration and executing inference pipelines
• examples/privacy_filter_studio/app.py shows a web application implementation that reads configuration files and applies active profiles at runtime

These scripts automatically detect the configuration directory and initialize the OpenMed runtime with the specified device, logging, and tokenizer settings.

Quick Start: Copy and Configure

Set up your local environment by copying the bundled example files and selecting a profile:

# Create the configuration directory structure

mkdir -p ~/.config/openmed/profiles

# Copy the example configuration to the default location

cp openmed/config.toml.example ~/.config/openmed/config.toml

# (Optional) Copy a specific profile configuration

cp openmed/profiles/dev.toml ~/.config/openmed/profiles/

# Activate the profile via environment variable

export OPENMED_PROFILE=dev

# Run an example script

python examples/privacy_filter_unified.py

Summary

• OpenMed uses TOML configuration files stored in ~/.config/openmed/ by default, controlled by the OpenMedConfig class in openmed/core/config.py
• Three ready-to-use environment profiles are available: dev.toml, prod.toml, and test.toml in the openmed/profiles/ directory
• Copy openmed/config.toml.example to ~/.config/openmed/config.toml to establish your base configuration
• Activate specific profiles using the OPENMED_PROFILE environment variable or the OpenMedConfig.from_profile() method
• Example scripts in examples/ demonstrate complete configuration loading and model execution workflows

Frequently Asked Questions

Where does OpenMed look for configuration files by default?

OpenMed searches for config.toml in the ~/.config/openmed/ directory. You can override this location by setting the OPENMED_CONFIG environment variable to point to a specific file path on your system.

What is the difference between the dev and prod profiles?

The dev.toml profile configures debug-level logging and extended timeouts to facilitate troubleshooting during local development. In contrast, prod.toml uses warning-level logging and shorter timeouts to optimize performance and reduce log volume in production deployments.

How do I activate a specific configuration profile?

You can activate a profile by setting the OPENMED_PROFILE environment variable to the profile name (e.g., export OPENMED_PROFILE=dev), or programmatically by calling OpenMedConfig.from_profile("dev") in your Python code. The system loads the corresponding TOML file from the profiles/ subdirectory.

Can I use the medical tokenizer without loading a configuration file?

While OpenMed provides built-in defaults through PROFILE_PRESETS in openmed/core/config.py, you must explicitly set use_medical_tokenizer in a configuration file or profile to control this feature. The test.toml example shows how to disable it, while production configurations typically enable it for specialized medical text processing.

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