The Ingest step has three ways to bring evidence in.
Drop or click — single files
Use the dropzone for one or many discrete files. Accepted: PDF, common image formats (PNG, JPEG, GIF, TIFF, BMP, WebP), Word (.docx), Excel (.xlsx), CSV, plain text, and ZIP. Per-file limit is 2 GB. Each file streams through SHA-256 + virus scan before storage; failures are recorded but never block the upload of unaffected files.
Choose folder — many files at once
Click Choose folder to pick an entire directory tree. The browser flattens it; the system preserves the original folder path as metadata on each document so you can see lineage in the documents list ("Q3-evidence/bank/march.pdf"). Up to 1000 files per folder pick.
ZIP archives — auto-expanded
Drop a .zip and the system expands it into individual child documents in the same case. The archive itself stays in the database for chain of custody but is hidden from the documents list. OOXML packages (Office files that happen to share the ZIP magic bytes), encrypted ZIPs, and nested ZIPs are not expanded. Decompression caps: 8 GB uncompressed, 50,000 entries max, 100:1 per-entry expansion ratio.
What happens after upload
Each new document goes through extraction (text, table detection, AcroForm field values on filled PDFs, OCR fallback on scanned PDFs), classification, entity extraction (SSNs, EINs, dollar amounts, IBANs, emails, phones, addresses), and tag matching. If the document contains a case number like ARC-2025-00042 in the text, the post-extraction linker auto-attaches it to that case.
If you upload via the global /documents page without picking a case or party, the document lands in the "Unattached only" bucket on that same page — useful for bulk triage.