Hi all

Hi all.

I have an api endpoint that does the following:

Plain Text

    ...
    parser = SimpleNodeParser()
    nodes = parser.get_nodes_from_documents(documents)

    # create (or load) docstore and add nodes
    docstore = MongoDocumentStore.from_uri(uri=URI, db_name='db', namespace='public')
    docstore.add_documents(nodes)

    # create storage context
    storage_context = StorageContext.from_defaults(
        docstore=docstore,
        index_store=MongoIndexStore.from_uri(uri=URI, db_name='db', namespace='public')
    )

    # build index
    index = GPTVectorStoreIndex(nodes, storage_context=storage_context)

On a separate endpoint, I wanna load the index and query it:

Plain Text

    storage_context = StorageContext.from_defaults(
        docstore=MongoDocumentStore.from_uri(uri=URI, db_name='db', namespace='public'),
        index_store=MongoIndexStore.from_uri(uri=URI, db_name='db', namespace='public')
    )
    index = load_index_from_storage(storage_context)
    query_engine = index.as_query_engine()
    result = query_engine.query(data.prompt)

this errors out with KeyError('1').

I believe that stems from a method on the storage context that load_index_from_storage calls deep in the code: storage_context.index_store.index_structs().
This is what errors out with KeyError('1'); I take it that means it couldn't find an index/index_struct? But https://gpt-index.readthedocs.io/en/latest/how_to/storage/index_stores.html says that, if using MongoDBIndexStore, you don't have to persist storage.

am I loading the index correctly? Am I fundamentally doing something wrong?

Find answers from the community

Hi all