What happens internally when you call index.as_query_engine()?

Calling index.as_query_engine() looks like a single simple call, but it assembles several components behind the scenes using sensible defaults so you don't have to wire them up manually.

1. It constructs a retriever appropriate to the index's type, for example a VectorIndexRetriever for a VectorStoreIndex, configured with any retrieval kwargs you passed, such as similarity_top_k.
2. It builds a response synthesizer using the LLM from Settings.llm (or an explicit llm kwarg) and the requested response_mode, defaulting to compact if none is specified.
3. If you passed node_postprocessors, they're attached to run between retrieval and synthesis.
4. All of this is wrapped into a query engine object, typically a RetrieverQueryEngine, that exposes the .query() and .aquery() methods you actually call.

Understanding this matters when defaults stop being good enough: since the pieces are just a retriever, postprocessors, and a synthesizer glued together, you can always drop down to the low-level API and construct a RetrieverQueryEngine yourself with custom versions of any of those three components.