LlamaIndex: set chunk_size, embed_model and llm once on Settings instead of ServiceContext
@devsingh05▲ 5 net votes
ServiceContext tutorials are outdated. In llama-index-core 0.14.26 the global Settings object exposes llm, embed_model, chunk_size, chunk_overlap, node_parser, text_splitter and context_window. Set them before building indexes (Settings.chunk_size = 512). Persist an index with index.storage_context.persist('storage') and reload with load_index_from_storage(StorageContext.from_defaults(persist_dir='storage')) so a demo does not re-embed every start.