How to get documents ready for AI tools: what goes wrong, how to fix it with or without Agnos, and real examples from Agnos’s own output.
Remove headers, footers and page numbers from PDF text before RAGWhy page furniture ends up in your chunks, a real before and after, a short Python script, and how Agnos names each removal.
How much of a PDF’s text is page furniture? 25 public documentsHow much of the plain text in 25 public PDFs from US federal agencies is headers, footers and page numbers, what Agnos removed and kept, and where it was wrong.
PDF to JSONL for RAG and fine-tuningWhat JSONL is, the fields a RAG corpus needs, real records made from a PDF, and a sample file to download.
Load an Agnos RAG export into LangChain or LlamaIndexRead an Agnos RAG corpus into LangChain and LlamaIndex with a few lines of Python, keep every field as metadata, and run one query on a small index in memory.
Prepare documents for Custom GPTs, NotebookLM, Claude Projects and CopilotWhy cleaned text files make better knowledge files, how to make them without code, and each tool’s own limits.
Use an Agnos export in Custom GPTs, NotebookLM, AnythingLLM and n8nWhich Agnos export to give each tool, what each one accepts today, how to upload, and an n8n workflow that reads the JSONL.
Free tool: find the repeated lines in a PDFDrop a PDF to see its running headers, footers and page numbers, and any pages with no text. It runs in your browser, and nothing is uploaded.
About the examples
Every Agnos output in these guides is real. Most of it came from Agnos 2.3.0, run on the documents in the sample project that comes with it, plus a four-page PDF made from the sample’s staff handbook. The study used Agnos 2.3.2 on 25 public PDFs from US federal agencies. Excerpts are shortened or laid out across lines where the page says so, and are otherwise unchanged.
Most guides also show how to do the job without Agnos, by hand or with a short script, and what each way is good and bad at.
Using Agnos
For the app itself, step by step, see the documentation, starting with the quick start on the sample project. What this version does not do is in What Agnos doesn’t do. How Agnos compares with Docling, LlamaParse, Marker, MarkItDown, rag-scrubber and Unstructured, and when one of them is the better pick, is in Compare Agnos.
Try Agnos on your own documents
Try everything in Pro free for 14 days, with no card and no account. Agnos runs on Windows 10 and 11, and your documents never leave your computer.