thepipe

An open-source tool that extracts clean data from complex documents and websites using vision-language models, for feeding into multimodal LLMs.