{ "message": "Error : An error occurred: 'str' object does not support item assignment." }

Question

I have a blob storage account where I dropped a single file. Then I want to add a record on pinecone based on this file using langchain:

@app.get("/BlobStorage")
def IndexContainer(storageContainer: str, indexName: str, namespace_name: str):
    logging.info('Python HTTP trigger function IndexContainer processed a request.')

    try:
        pinecone.init(
            api_key=os.getenv("pineconeapikey"),
            environment=os.getenv("pineconeenvironment")
        )

        connect_str = os.getenv('blobstorageconnectionstring')
        loader = AzureBlobStorageContainerLoader(conn_str=connect_str, container=storageContainer)
        embeddings = OpenAIEmbeddings(deployment=os.getenv("openai_embedding_deployment_name"),
                                        model=os.getenv("openai_embedding_model_name"),
                                        chunk_size=1)

        openai.api_type = "azure"
        openai.api_version =os.getenv("openai_api_version")
        openai.api_base =os.getenv("openai_api_base")
        openai.api_key =os.getenv("openai_api_key")

        documents = loader.load()
        texts = []
        metadatas = []

        for doc in documents:
            texts.append(doc.page_content)
            metadatas.append(doc.metadata['source'])

        docsearch = Pinecone.from_texts(
            texts,
            embeddings,
            index_name=indexName,
            metadatas=metadatas,
            namespace=namespace_name
        )


        return {
            "message":f"File indexed. This HTTP triggered function executed successfully."
        }
    except Exception as e:
        error_message = f"An error occurred: {str(e)}"
        logging.exception(error_message)
        return {
            "message":f"Error : {error_message}."
        }

I debugged this code line by line and all variables are setup correctly and the exception is only thrown until the from_texts method.

However I get this error:

An error occurred: 'str' object does not support item assignment.

loader.load is below:

[Document(page_content="long content", metadata={'source': 'C:\\Users\\xx\\AppData\\Local\\Temp\\tmpk5cqh4nd/abc/filenameAI sorting_Project Description.docx'})]

Stack trace

What am I missing?

I based my code on the unit tests of langchain:

https://github.com/hwchase17/langchain/blob/e27ba9d92bd2cc4ac9ed7439becb2d32816fc89c/tests/integration_tests/vectorstores/test_pinecone.py#L118

If it helps, the Blob Storage Container loader from Langchain is implemented here:

https://github.com/hwchase17/langchain/blob/e27ba9d92bd2cc4ac9ed7439becb2d32816fc89c/langchain/document_loaders/azure_blob_storage_container.py

Answer 1

Your problem is int the metadata variable. Looking at the source code it seems that metadata needs to be a list of dictionaries. You are passing a list of strings instead.

Just fix it by:

        for doc in documents:
            texts.append(doc.page_content)
            metadatas.append(doc.metadata)

Answer 2

The error 'str' object does not support item assignment typically occurs when you try to modify a string object, which is immutable in Python. Looking at your code, the issue seems to be with the assignment of metadatas in the for loop.

The documents list contains Document objects with a metadata attribute, but in your code, you're trying to access the metadata using doc.metadata['source']. However, since metadata is a string object, you cannot access its elements like a dictionary.

To resolve this, you need to ensure that metadata is in a dictionary format. If metadata is a JSON string, you can parse it using json.loads() to convert it into a dictionary. Here's an example of how you can modify your code:

import json

# ...

documents = loader.load()
texts = []
metadatas = []

for doc in documents:
    texts.append(doc.page_content)
    metadata_dict = json.loads(doc.metadata)  # Parse the 

metadata JSON string into a dictionary
        metadatas.append(metadata_dict.get('source', ''))  # Access the 'source' key from the metadata dictionary
    
    # ...

By parsing the metadata string into a dictionary, you can then access the 'source' key correctly. Make sure the metadata value is in a valid JSON format for this to work.

With this modification, you should be able to resolve the error and proceed with your code.