The community member asked for examples using the new API, and other community members provided helpful information. They discussed a specific example on GitHub that demonstrates loading documents and creating an index with the new API. The main changes mentioned are the new ServiceContext object and the use of nodes as the main inputs to indexes. The community members also discussed adding metadata to documents and expressed interest in experimenting with the new API features, but noted that their jobs limit the time they can devote to it.
The main difference is the new ServiceContext object that wraps the llm predictor, prompt helper, chunk_size_limit, and a few other things
Another major change is that nodes are the main inputs to indexes. You can still load from documents using the from_documents() function i.e. (index = GPTListIndex.from_documents(documents, service_context=service_context)
