Skip to content

YaronKoresh/definers

Repository files navigation

Definers

Definers is a modular Python platform for teams that build AI, media, and data products and want one serious codebase instead of a pile of disconnected utilities.

It brings together workflow-grade audio, text, image, video, data preparation, runtime compatibility, and launcher surfaces behind a concrete module structure that is designed to stay credible under real project pressure.

What Definers Is Built For

  • Shipping AI and media workflows without rewriting the surrounding infrastructure every time the stack changes.
  • Keeping CPU-only environments usable while still unlocking GPU-backed acceleration where the runtime can support it.
  • Giving teams one place for preparation, inference, automation, and runtime.

Capability Areas

  • Audio workflows for mastering, stems, analysis, cleanup, and generation.
  • Data workflows for preparation, tokenization, vectorization, and dataset assembly.
  • ML workflows for text processing, guided and advanced training inside the train workbench, inference, and retrieval-oriented features.
  • Image and video workflows for generation, composition, enhancement, and rendering.
  • System and runtime workflows for installation, process control, download handling, and compatibility.
  • Focused launcher surfaces for domain-specific applications instead of a single overloaded interface.

Architecture Direction

  • The repository is standardizing on feature-owned packages instead of broad routing layers.
  • import definers is a stable lazy discovery surface, not a signal to eagerly import every optional runtime.
  • Package facades such as definers.data, definers.chat, definers.text, and definers.media stay lazy discovery surfaces, while implementation work belongs in concrete owner modules.
  • CLI, launcher, and download behavior are being tightened around explicit contracts so runtime behavior stays predictable in local, CI, and hosted environments.

Documentation

Project Standards

  • Python support targets 3.10 through 3.14.
  • The package is structured to keep runtime policy, optional dependencies, and user-facing launchers explicit.
  • Contributor workflow and validation guidance live in CONTRIBUTING.md.

License

Definers is licensed under the MIT License. See LICENSE.