Skip to content

Segmentation Fault when Initializing Chroma Vector Store in LangChain #15

@godofecht

Description

@godofecht

I am encountering a segmentation fault when trying to initialize a Chroma vector store using langchain_community.vectorstores.Chroma. The issue occurs specifically at the point where I call Chroma.from_texts to create the vector store. Here is a minimal code snippet to demonstrate the issue:

import numpy as np
from langchain_openai import OpenAIEmbeddings
from langchain_community.vectorstores.chroma import Chroma

# Simplified data for testing
texts = ["This is a test document.", "This is another test document."]
metadatas = [{"title": "Test Document 1", "summary": "Summary of test document 1"},
             {"title": "Test Document 2", "summary": "Summary of test document 2"}]

# Initialize embeddings model
embeddings_model = OpenAIEmbeddings()

# Debugging information to check data integrity
print(f"Number of documents: {len(texts)}")
print(f"First document text: {texts[0]}")
print(f"First document metadata: {metadatas[0]}")

# Attempt to initialize Chroma Vector Store
try:
    print("Initializing Chroma Vector Store...")
    docsearch = Chroma.from_texts(texts=texts, embedding=embeddings_model, metadatas=metadatas)
    print("Chroma Vector Store initialized successfully.")
except Exception as e:
    print(f"Error initializing Chroma Vector Store: {e}")

Expected Behavior:

The Chroma vector store should initialize successfully, and the subsequent print statements should execute without errors.

Actual Behavior:

The script encounters a segmentation fault immediately after attempting to initialize the Chroma vector store with Chroma.from_texts.

Interestingly, the print statements prior to the call to Chroma seem to get swallowe by the crash, potentially diverting std.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions