Odipie

Fast AI/ML Workflows Through Intelligent Lazy Loading

Reduce startup cost by loading heavy AI/ML libraries only when a code path actually uses them.

Odipie is a lightweight Python toolkit and project template for lazy-loading optional AI/ML dependencies. The base install stays small. Frameworks such as TensorFlow, PyTorch, scikit-learn, Transformers, Pandas, NumPy, Matplotlib, and OpenCV can be installed as optional extras.

Features

• Lazy proxy objects for common AI/ML libraries.
• Optional dependency extras instead of one oversized required install.
• Compatibility shim for older examples that use import lazy_init_py as odipie.
• Small CLI and smoke-check app for verifying the package install.
• Docker and Compose workflow that builds from the real repository tree.
• Wiki source files under docs/wiki-src/ for longer engineering notes.

Project Tree

odipie/
├── odipie/
│   ├── __init__.py
│   └── __main__.py
├── ROProj/
│   ├── README.md
│   ├── app.py
│   ├── config.py
│   ├── docker-compose.yml
│   └── requirements.txt
├── docker/
│   ├── Dockerfile
│   ├── .dockerignore
│   ├── entrypoint.sh
│   └── README.md
├── docs/
│   └── wiki-src/
├── tests/
│   └── test_imports.py
├── .dockerignore
├── app.py
├── config.py
├── docker-compose.yml
├── lazy_init_py.py
├── pyproject.toml
├── requirements.txt
└── README.md

Installation

Create and activate a virtual environment, then install the package in editable mode:

python -m venv .venv
source .venv/bin/activate  # Windows: .venv\Scripts\activate
pip install -e .

Install optional stacks only when you need them:

pip install -e .[data]
pip install -e .[tensorflow]
pip install -e .[torch]
pip install -e .[ml]

Usage

import odipie

print("App ready")
print(odipie.get_loaded_modules())

# TensorFlow imports only when this attribute chain is used.
model = odipie.tensorflow.keras.Sequential([...])

# scikit-learn imports only when this code path runs.
processed = odipie.preprocess_data(data, method="standard")

Older examples still work:

import lazy_init_py as odipie

CLI Smoke Checks

python -m odipie --loaded
python -m odipie --versions
python app.py

Tests

pip install -e .[dev]
pytest

Docker

docker compose build
docker compose up

See docker-setup.md for the Docker walkthrough.

Documentation

Resource	Description	Link
Wiki Home	Complete knowledge base	Wiki
Getting Started	Install, optional extras, Docker, and troubleshooting	docs/wiki-src/Getting-Started.md
Docker Guide	Build and run the package in Docker	docker-setup.md
AI Terminology & FAQ	Plain-language AI and agentic-system reference	docs/wiki-src/AI-Terminology-and-FAQ.md
Agents, MCP & Orchestration	Agents, graph workflows, MCP, and WebMCP	docs/wiki-src/Agents-MCP-and-Orchestration.md
Prompt Engineering	Prompt ideation and context engineering	adv_promptGuide.md
Wiki Sources	Source markdown for wiki pages	docs/wiki-src

Security Note

Do not use wildcard imports in project code. For PyTorch model files, only load artifacts from trusted sources; Odipie uses safer torch.load(..., weights_only=True) defaults when supported by the installed PyTorch version.

License

This project is licensed under the Apache License 2.0.