/ (Root)
├── app/
│   ├── work/               <-- MOUNTED from $(pwd) on Host
│   │   ├── your_script.py
│   │   └── data_files/
│   └── lib/                <-- MOUNTED from ${BUILD_DIR}/lib on Host
│       └── aiesda/         <-- Your Python Package Root
│           ├── __init__.py
│           ├── VERSION
│           ├── requirements.txt
│           ├── pylib/      <-- Added to PYTHONPATH
│           ├── pydic/      <-- Added to PYTHONPATH
│           ├── nml/
│           └── yaml/
├── usr/
│   └── local/
│       └── [JEDI STACK]/   <-- Pre-installed in the Base Image
│           └── ufo/        <-- The UFO C++/Python bindings
└── etc/
    └── bash.bashrc         <-- Contains the JEDI path exports

AIESDA acts as a central hub, ensuring that data from any source is authenticated and standardized before it reaches the data assimilation cycle.

Understand the decoupled design of AIESDA, which consist of five major components as listed below.

The gatekeeper of the framework, verify the fingerprint and footprint of the model credential. Before any file is processed, it must present its "credentials."

The Passport verifies:

Identity: Is this file recognized by aidadic.py?

Biometrics: Do the vertical pressure levels match the model's known fingerprint?

Integrity: Is the resolution correct and is the data free of NaN values?

Located in dynlib.py, these interfaces handle the specific quirks of NWP systems like Bharat and Mithuna. They manage settings for the complexity domain and resolution variants of ocean-atmosphere coupled model families using the concept of class and inheritance.

Located in ailib.py, these bridges handle the output of foundation models. Since AI models often use different naming conventions (e.g., 2t vs air_temperature), these interfaces use the Registry to perform automated renaming and level standardization.

The dalib.py module contains the JEDI Model Bridge. It translates standardized AI/NWP states into GeoVaLs and IODA formats, allowing the data to be ingested by the JEDI/UFO observation operators.

The scilib.py module provides the evaluation layer. Once assimilation is complete, this toolbox is used to calculate bias, RMSE, and anomaly correlation coefficients to verify the "value-add" of the AI background.