Issue Creating a Property Graph Index

At a glance

The community member is facing an issue when creating a property graph index using the llama-index library. The error message indicates that the index_id attribute is not available on a coroutine object. The community members discuss potential solutions, including checking the version of llama-index, reproducing the issue without Neo4j, and running the code in a Google Colab environment.

One community member suggests adding import nest_asyncio; nest_asyncio.apply() at the top of the code and launching the FastAPI application with uvicorn.run(..., loop="asyncio"), which resolves the issue.

Another community member explains that the root cause is that Python does not allow async constructors on Python classes, and suggests that the index itself should have async insert methods to avoid any async nesting. They mention the need to make a pull request to add async entry points to the base index class.

Useful resources

ttharak#3

hi,

i am facing issue when creating a property graph index

Plain Text

    kg_extractor = SchemaLLMPathExtractor(
        llm = OpenAI(model="gpt-4o", temperature=0.1),
        possible_entities=entities,
        possible_entity_props=entity_properties,
        possible_relations=relations,
        kg_validation_schema=validation_schema,
        strict=True
    )

Plain Text

    graph_store = Neo4jPropertyGraphStore(
        username=os.getenv("username"),
        password=os.getenv("password"),
        url=os.getenv("url"),
        database=os.getenv("database")
    )

Plain Text

    pg_index = PropertyGraphIndex.from_documents(
        documents,
        kg_extractors=[kg_extractor],
        embed_model = OpenAIEmbedding(model="text-embedding-3-small"),
        property_graph_store=graph_store,
        vector_store=None,
        show_progress=True,
    )

when i run it getting error
File "/home/user/.cache/pypoetry/virtualenvs/llama-graph-Kulloq32-py3.12/lib/python3.12/site-packages/llama_index/core/storage/index_store/keyval_index_store.py", line 44, in add_index_struct
key = index_struct.index_id
^^^^^^^^^^^^^^^^^^^^^
AttributeError: 'coroutine' object has no attribute 'index_id'

8 comments

LLogan M

Do you have the latest version of llama-index? That seems low-key impossible 😅

ttharak#3

Yeah, I was running on latest version

LLogan M

Can you reproduce without neo4j? (I hate settting that up lol) -- would love to repro on google colab

ttharak#3

this is something weired that i was facing, it works fine in colab and throws error when runned in fastapi application.

colab link - https://colab.research.google.com/drive/1VNIpaOflVOgSSLwQd8ut6HHX9Vo39_QA?usp=sharing

ttharak#3

in colab it runs good, but when i run in fastapi i get runtime errors

LLogan M

Oh, I see

Put this at the top of your code

Plain Text

import nest_asyncio
nest_asyncio.apply()

And launch your app with

Plain Text

uvicorn.run(..., loop="asyncio")

ttharak#3

Thanks Logan.
This worked

LLogan M

The root issue here is, python doesn't allow async constructors on python classes 😅 Ideally, the index itself has async insert methods so that you can create an empty index and use async inserts to populate it. This would avoid any async nesting

But, need to make a PR to add async entry points to the base index class 🙂

Add a reply

Find answers from the community

Issue Creating a Property Graph Index