EDGAR (SEC Filings)

EDGAR (Electronic Data Gathering, Analysis, and Retrieval) is the SEC's public archive of the disclosures every US public company must file. It reaches back to 1993 and holds millions of documents: annual reports (10-K), quarterly reports (10-Q), material event notices (8-K), registration statements, insider trading forms, and shareholder proxy materials. If a listed US company had to tell its investors something, it is almost certainly in here.

You have several ways in. Full-text search covers filings from 2001 onward, the daily and quarterly index files list everything submitted in a given period, and the submissions and company facts APIs return clean JSON without needing a key. Individual filings come as HTML, plain text, and XBRL, a tagged XML format that exposes each figure in a financial statement as a labelled, machine-readable field. A sensible pipeline walks the index files, downloads the primary document, strips the HTML, and chunks by the filing's own section headings.

EDGAR suits any RAG system that answers questions about company financials, filing history, or regulatory compliance: an assistant that compares a firm's risk factors year on year, say, or one that surfaces the exact 10-K passage behind a number. Because the XBRL layer carries the actual figures, you can pair narrative retrieval with precise numeric lookups rather than trusting a model to read a table correctly.

A few things to watch. The SEC rate-limits requests (roughly 10 per second) and asks you to send a descriptive User-Agent header, so be polite or you will be blocked. Older filings can be messy ASCII with inconsistent formatting, and XBRL tagging quality varies between companies. On licence, everything is a US Government work in the public domain, so commercial reuse is fine with no attribution or share-alike strings attached. For non-US markets, pair EDGAR with legislation.gov.uk or EUR-Lex on the regulatory side; for the filings themselves it is the definitive source.