Difficult integration with langchain

[pprados]

Hello,

Zerox's latest langchain integration does not comply with best practices. The proposed API only accepts a file, not a blob.

Normally, for various reasons, a loader should be split into two parts: the loader and the parser. The parser should parse a blob (not a file). The loader loads the file and gives it to the parser.

Why is this? To be able to use Zerox from files directly retrieved from cloud storage, via the GenericLoader and CloudBlobLoader classes, without going through a temporary file.

The api must accept an array of bytes as input.

Thanks