[Bug]: AttributeError: 'MultiModalVectorIndexRetriever' object has no attribute 'object_map'

[dodofung]

Bug Description

I Use MultiModalVectorIndexRetriever, and when it runs, this error happend:
AttributeError: 'MultiModalVectorIndexRetriever' object has no attribute 'object_map'

Version

main

Steps to Reproduce

the code like the following

index = MultiModalVectorStoreIndex.from_documents(image_documents,storage_context=storage_context)
retriever_engine = index.as_retriever( similarity_top_k=5, image_similarity_top_k=3 )
query_str = "....."
retrieval_results = retriever_engine.retrieve(query_str)
###the probable reseaon
MultiModalVectorStoreIndex class init method don't call super().init, I add it, the error disappered

Relevant Logs/Tracebacks

[dosubot]

Hi @dodofung! I'm Dosu and I'm helping the LlamaIndex team.

You're correct — this is a real bug. The MultiModalVectorIndexRetriever.__init__() never calls super().__init__() [1], which means BaseRetriever.__init__() is never invoked and attributes like object_map and _verbose are never set [2].

The error surfaces because BaseRetriever.retrieve() calls _handle_recursive_retrieval(), which accesses self.object_map [3]. Since the parent __init__ was skipped, that attribute doesn't exist.

Fix: Add a super().__init__() call at the end of MultiModalVectorIndexRetriever.__init__():

def __init__(self, ...) -> None:
    # ... existing attribute assignments ...
    super().__init__(callback_manager=callback_manager)

Your diagnosis is spot-on — adding the super().__init__() call is the correct fix.

To reply, just mention @dosu.

---

Share context across your team and agents. Try Dosu.