Glossary

Plain-language definitions of the terms used in Discovery Engine and in these docs.

#A–C

Agent (AI agent) An AI assistant that can run commands on your computer, such as ChatGPT desktop (Work) or Claude desktop. It works with your case through the discovery command-line tool.

Agent access A per-case switch (Settings → This case → Let AI agents use this case). When off, every agent command on that case is refused.

Agent log The screen listing every command your agent ran, with the kind of answer each got.

Agent privacy block A stricter mode built into the engine that withholds agent output holding detected personal details. It's part of stand-in mode for agents, which is off by default in this version.

Agent work folder The folder you open as your agent's project. It holds the agent's notes and the guide files AGENTS.md and CLAUDE.md.

AGENTS.md / CLAUDE.md The app's guide for AI agents, written into the work folder by Write guide files. Agent apps read them automatically.

App data folder Where the app keeps each case's database, the models and the logs. Holds real names; never give it to an agent.

Bates number An identifier stamped on each page of a formal production (for example HFC-000123). The app reads them per page; a Bates number opens the page it's stamped on.

Case A folder of files (plus any added to it) and everything the app learns from them.

Category A tag on documents, such as the attorney's own categories. The exhibit index lists them, and Search can filter by them.

Citation An exact pointer to a place in a document: DOC-0417 p.2, EMAIL-0007 msg 3, SHEET-0033 row 118, VID-0012 08:41.

Command-line tool (discovery) The command your agent runs to search, read and cite. Installed from the Start Up screen.

Connections study A set of people and events on the Connections screen, with every place each pair meets.

#D–M

.decase file A single file holding a whole case (its database and cache, not the originals), for moving it to another computer.

Display ID A document's short ID: DOC-, EMAIL-, SHEET-, IMG-, AUD- or VID- followed by a number, or a Bates number.

Document What the app reads a file as. Usually one file is one document; e-mail attachments and files in ZIP archives become documents of their own.

Exhibit index A spreadsheet (XLSX and CSV) listing every exported document with its ID, date, type, title, categories and more.

Export folder Where exports for the attorney go.

Highlight A passage proposed for yellow-highlighting, with a quote, a citation and a reason. Only approved highlights are exported.

Issue list "What this case is about": the short list on the Highlights screen that proposals are judged against.

Load file A .dat or .opt file that comes with a formal production, describing its documents and pages.

Merge Telling the app that two spellings are the same person or organization (People screen). Every merge can be undone.

Models (reading models) Speech recognition, speech detection, name detection and search-by-meaning models, downloaded once (about 1 GB).

#O–S

OCR Optical character recognition: reading text from scans and pictures. Built into the app.

Passage (segment) The unit a citation points at: a page, paragraph, spreadsheet row, e-mail message or stretch of a recording.

Production A set of documents handed over in discovery, often Bates-stamped and sometimes with load files.

Prompt The first message you paste into your agent, from Start Up → Copy prompt.

Review A record your agent keeps while working through every document that could answer a question: relevant, not relevant or unsure, with a reason and citations.

Rescan Looking for new, changed and removed files in a case's folders (Rescan folder on Progress).

Sanitized With personal details replaced by stand-ins, and blacked out on page images. Exports can be sanitized.

Search by meaning (semantic search) Finding passages that say the same thing in other words.

Show stand-ins The switch at the bottom of the left column that shows stand-ins instead of real names everywhere in the app, for screenshots and screen sharing.

Speed score How many five-page scanned PDFs this computer reads in ten minutes. Used for time estimates.

Stand-in A neutral label the app gives a detected personal detail: PERSON A, ORG 3, ACCOUNT 1, ADDRESS 2. One real person gets one stand-in across the whole case.

Start Up The screen that gets your AI agent connected.

#T–Z

Transcript The text of a recording, with timestamps, made on your computer.

Work product What you hand the attorney: highlighted PDFs, the exhibit index, connections spreadsheets.