Text Cleaner for LLMs & RAG

Why clean text before sending it to an LLM? Boilerplate, repeated headers, and formatting junk cost real money (every token is billed), pollute embeddings in RAG pipelines, and distract the model from the content that matters. The heuristics here are conservative: repeated-line removal only drops short lines that appear three or more times (classic page headers and footers), and signature detection looks for standard sign-off markers near the end of the text. Always skim the "after" pane before using the output; no heuristic is perfect on every document. Everything runs locally in your browser, and nothing you paste is uploaded anywhere.