lmmpg

lmmpg, short for Lightweight Multi-Modal Processing Graph, is an open-source software framework for building and executing modular data-processing pipelines that combine multiple data modalities such as text, images, and audio. It adopts a graph-based execution model in which processing components are implemented as nodes and data flow is represented by edges between nodes.

The project began in 2020 as a community effort to provide a simple, extensible platform for rapid experimentation with multi-modal workflows. It offers bindings for Python and C++, and aims to be language-agnostic through a plugin system. The initial release introduced a graph engine, a plugin registry, and basic I/O adapters; subsequent releases added streaming support and richer data types.

lmmpg consists of a core graph interpreter, a plugin registry for data sources and processors, an execution scheduler, and a set of data type definitions. Nodes implement a common interface for input validation, error handling, and logging. The framework supports both batch and streaming processing and can serialize graphs to a portable format for reuse or sharing.

Typical applications include preprocessing pipelines for machine learning, multi-modal feature extraction, and experimental research into pipeline architectures. The project emphasizes simplicity and readability, prioritizing ease of use over aggressive optimization; for large-scale workloads, users may integrate with external schedulers or write optimized plugins.

lmmpg is released under an open-source license and maintained in a public repository with documentation, tutorials, and example graphs. It is a community-driven project with ongoing contributions and discussions in issue trackers and forums.