OpenAIEmbeddings Initialisation error on LangGraph

[lluisCM]

Hello!

I'm having trouble using "from langchain_openai import OpenAIEmbeddings" If I initialize it or use it inside an async function and I call it through a tool node, I get the following error message: "Blocking call to os.listdir"

The code I use in the tool function is something like the following, and it errors on the initialization. As Far as I understood from your documentation, it should be compatible with an async call but doesn't seem to work, I'm probably doing something wrong.

Example code:

async def search_profiles_by_semantic_distance(
    job_query: str, 
    location_query: str,
    job_threshold: float = 1.1,   # Set to None to disable job distance filtering
    location_threshold: float = 0.9,  # Set to None to disable location distance filtering
    top_k: int = 5
) -> List[Dict]:
    # It fails in the next line.
    embeddings = OpenAIEmbeddings(model="text-embedding-3-small")
    job_embedding = await embeddings.aembed_query(job_query)  
    location_embedding = await embeddings.aembed_query(location_query)

Can anyone help me with it?
thank you very much.

[hinthornw]

I believe if you upgrade your langgraph-cli[inmem] to latest, this will already be handled as expected.

You can also run langgraph dev --no-blocking

We added block-checks in the local development server to ensure your server performs optimally in production, but there were a few common calls that we've relaxed, since they are less likely to cause actual issues and are fairly common in popular libraries (such as openai)

[lluisCM]

Thank you!