Simplify installation of Haystack

[julian-risch]

Problem Statement
As a software developer lacking strong NLP skills, I want to make some (small) NLP application, without spending a lot of time on dependency conflicts and switching OSes.

User Tasks

• Install a truly minimal installation, so small not to include even transformers or torch. (for users that want to use LLMs through API only, see Allow the installation of a minimal version of the library that only installs dependencies necessary for querying hosted models #3611)
• Install Haystack via conda
• ...

Existing Work

• Extraction of Audio nodes:
  • refactor: extract speech nodes from Haystack #3229
  • https://github.com/deepset-ai/haystack-extras

Tasks

Preview Give feedback

1. Extract dependencies from core list #4417
   +0
   

[mayankjobanputra]

Maybe this discussion will also help eventually: #3750

[ZanSara]

Closing as completed! 🎉

Optional dependencies management has improved drastically and the core dependency list has shrunk significantly in Haystack 1.18. Work will continue into making the installation as simple as possible, including extra features like much smaller Docker images, pure Python wheels for PyScript, etc