I have been glancing through llama-index source code lately. I have come across their integrations and they seem to publish each integration as a different package. However, they modify the import path while publishing using poetry. I’m unable to understand how they do it.

[tool.llamahub]
contains_example = false
import_path = "llama_index.llms.azure_openai"

They use this in the pyproject.toml of their azure-OpenAI integration’s.

They seem to use llama hub for this. But couldn’t exactly figure out how they implement that on llamahub. Is there a way I can do the same easily.