Convert documents to RAG online, in the browser
Converting documents to RAG online means turning a file into overlapping passages and vectors in the page you already have open - not uploading it to a queue and waiting for a hosted parser to send JSONL back.
Online does not have to mean uploaded
People search 'RAG convert online' when they want the job done in a tab, this afternoon, without installing Python. The usual result is a form that takes the file and a cloud that parses it. The better reading of the same query is a page that is the converter: WebAssembly reads the bytes, chunks them, embeds them, and the download is yours. RAG Converter is built for that reading. The file never has to become someone else's request body.
What you actually do
Open the converter, drop a file or a folder, wait for the chunk count per name, download chunks.jsonl and manifest.json. Load the JSONL into SpaceDB, Chroma, Qdrant, pgvector or LanceDB. That is the whole motion. There is no project, no corpus id, no second product you have to wire before the first file is useful. If a tool needs an API key before it will convert one PDF, it is not answering this query.
What converts in a tab, and what does not
Word, PowerPoint, Excel, Markdown, HTML, CSV and a PDF with a text layer extract as text and convert here. A scanned PDF and a photograph need OCR; the signed-in app runs that on the device too. Audio and video need models a tab cannot hold, so those go to the API if you want them. An online converter that claims every format equally is either uploading or lying about the empty results.
Keep the download; that is the product
The JSONL is the corpus. One object per line, each with text and a vector, is what every vector database already imports. You do not need the site to store it for you, and you should not want it to. Convert online, take the file, close the tab. The next question is asked against your own index, not against a vendor's leftover copy of this afternoon's upload.