--- okf_version: "0.2" --- # COVID-19 Records An OKF bundle over 5,247 pages of records released by the U.S. Senate Committee on Homeland Security and Governmental Affairs — documents the Committee subpoenaed and then published — segmented into 18,094 individual communications spanning 2000-02-05 to 2026-04-10. ## What this bundle is, and is not **This is a derived layer, not the records.** It carries the analysis over the corpus: people, topics, organisations, outlets, sources, a chronology, and transcripts for the larger conversations. It does **not** carry the text of the 17,260 individual records, and it holds transcripts for only 301 of 2,021 conversations. That distinction matters more than it sounds. **Searching these files is not searching the corpus.** A grep across the bundle returns a fraction of the true hits and nothing here marks it as a fraction, so an agent reading only these files can produce a confident, well-cited, incomplete answer. For anything that needs completeness, use the API — see [Documents](documents/index.md). ## Collections * [Chronology](chronology/) - the record by date; the only time-based entry point * [People](people/) - 253 individuals, with correspondents, topics and statement timelines * [Topics](topics/) - 22 subject categories, each with its principal voices * [Organizations](organizations/) - 81 institutions * [Outlets](outlets/) - 54 media outlets and press-direction evidence * [Conversations](conversations/) - 301 chat sessions and email threads * [Sources](sources/) - 14 released document packages * [Documents](documents/) - how to retrieve the records themselves * [Glossary](glossary.md) - acronyms, institutional shorthand and strain names ## How to use this bundle Start from a [person](people/), a [topic](topics/) or a [date](chronology/) and follow the links. Person concepts carry a chronological statement timeline answering *what did this individual say, and when*. Topic concepts carry the same organised by subject. Then verify. Every quoted statement cites a page, and page text is retrievable at `https://randpaulcovid.org/api/page//`. The authoritative copy is the Committee's own PDF, linked from every [source](sources/) concept. ## What the fields mean `grade`, citations, chat audience, topic tags and the limits of coverage each carry a specific meaning that a reader taking the numbers at face value will get wrong. They are defined once, in machine-readable form, at `https://randpaulcovid.org/api/guidance` — not restated here, because an agent reading `/llms.txt`, this file and that endpoint would otherwise meet the same paragraph three times before reaching a record. ## Provenance Generated by `randpaul-corpus-pipeline/1.0` at 2026-08-01T12:42:27Z from the released PDF packages, which are public and hosted by the Committee at . This bundle adds no source documents. Text was recovered by OCR where the productions were scanned images. Extraction is rule-based and deterministic: re-running the pipeline reproduces this bundle. Statement timelines are machine-extracted and carry `confidence` below 1.0 in the database; they are evidence pointers, not adjudicated findings, and should be read alongside the cited page.