I am also building an app that converts web pages to Markdown. How do you handle messy scans and tables?

How does this differ from open source cli tools such as Marker (https://github.com/datalab-to/marker)?

Also, it seems disingenuous to have “locally” in the title when your default is cloud inference providers and I don't see you naming a recommended LLM that could be hosted locally and is proven to work well for this task. You list “Your files” under “Stays on your machine” and then require an API key for a cloud provider.