easterbrooka teaches Mike to read emails and spreadsheets

The intake door opens to plain text, raw email, and Excel - and the AI actually knows what to do with them.

Mike's document upload used to quietly draw a line at PDFs and Word files. easterbrooka widened it to also accept text files, raw email exports, and Excel workbooks - but the real work was on the other side of the door. The team wired up server-side parsers (mailparser for emails, exceljs for spreadsheets) so the AI can actually reason over what's inside, and built three new inline viewers so users see the same thing the model sees.

There are sensible guardrails: text and email get capped at 200,000 characters and spreadsheets at 1,000 rows by 50 columns per sheet, with visible truncation notices when limits hit. Legacy .xls was deliberately left out because the only library that handles it has known security issues - users are expected to re-save as .xlsx. No database changes, no new infrastructure.