Exports and reports
PDF and Word answers, card exhibits as images and CSV, and the Expert pack that reruns eligible calculations on the same snapshot.
The answer, as PDF or Word
Export, under each answer, turns it into a PDF or an editable Word file. You choose whether every consulted source goes with it or only the cited ones. Kelsey can also draft a memo or an email from the cards on screen; you edit the draft in the app and download it as Word on your office letterhead.
Each card, as an exhibit
Every card has "Copy or download": Copy as image, Download PNG, Download SVG for supported charts, and Copy data (CSV) or Download data (CSV) for a card with a table. A CSV row keeps the identifier behind it:
| Card | Columns in the CSV include |
|---|---|
| Case reports | caseid, primaryid, fda_date, event_date, age, sex, country, reporter, outcomes, reactions |
| Signal over time | quarter, the measure, lower_95, upper_95, reports_not_lawyer_filed, reports_lawyer_filed |
| Literature | pmid, date, title, journal, publication_types, retracted, finding, quote |
| Foreign regulators | date, regulator, products, section, quote, document, url |
| Label versions | version, date, section, mentions_reaction |
| Court decisions | caption, court, date, citation, holding |
The Expert pack
Under an answer whose cards carry a run stamp, the Expert pack is a zip. It holds:
exhibits.json: every card in the answer, each stamped with its data snapshot, tool and arguments- the methods appendix, in Markdown and PDF
VERIFY.md: the command that opens the snapshot read-only, runs each eligible tool again and compares the numbers, counts exactly and decimals to a relative difference of 1e-9
The check covers the exhibits whose numbers come from a snapshot. Exhibits built from documents, case law and literature, and from live lookups, cite their sources and are listed as not checked. It compares each exhibit's data, not every number in the written answer. Running it needs the code at the version named in the appendix, with its Python environment, and the data snapshot each exhibit names.
The run behind each calculated figure
The "Method and runs" card dates the data under the answer: "Every figure on these cards comes from the data snapshot of FAERS through 2026Q2, held from 2004Q1; snapshot 27 Oct 2026, PubMed metadata, documents corpus, ClinicalTrials.gov results, fetched 25 Oct 2026, built 25 Oct 2026; API data as of 23 Oct 2026 and 15 saved searches; open any card's method to repeat one." The answer text ends with the same stamp, for example "Data: FAERS through 2026Q2, held from 2004Q1, snapshot 2026-10-27T020806; primary-suspect reports; lawyer-filed reports excluded."
Example questions
Which terms make up "eye damage", and which query and data version sit behind each number?
What comes back: The term list, the query run and the data snapshot behind each calculated figure.
List the published studies on ciprofloxacin and tendon rupture.
What comes back: The literature card, with each study's date, journal, type and quoted finding. Download data (CSV) gives one row per study, with its PMID.