Chat engine from multiple (knowledge graph) indices #13233
Replies: 3 comments 1 reply
-
To initialize a chat engine with multiple indices, you can leverage the For merging loaded indices into a single index, consider using the def load_graph_from_storage(
storage_context: StorageContext,
root_id: str,
**kwargs: Any,
) -> ComposableGraph:
indices = load_indices_from_storage(storage_context, index_ids=None, **kwargs)
all_indices = {index.index_id: index for index in indices}
return ComposableGraph(all_indices=all_indices, root_id=root_id) This method loads indices and merges them by creating a graph structure, allowing for unified operations across the combined indices. Additionally, the While these methods provide a foundation for working with multiple indices, the direct initialization of a chat engine with a
|
Beta Was this translation helpful? Give feedback.
-
@dosu-bot, thanks for the response. What happens if the graph has disconnected sub-graphs? Also, where do I get the On the other hand, does the |
Beta Was this translation helpful? Give feedback.
-
Perhaps, there is no need to merge indices if an agentic RAG can be built to use multiple indices that can be looked up as and when needed. |
Beta Was this translation helpful? Give feedback.
-
While
load_indices_from_storage
allows for loading multiple (say, knowledge graph) indices from some storage context, is there any way to initialise one single chat engine from all the indices?Alternatively, is there a way to merge the loaded indices into one single index so that
index.as_chat_engine()
can be invoked on that merged single index?Beta Was this translation helpful? Give feedback.
All reactions