docs: add domain glossary (CONTEXT.md)

CONTEXT.md

@@ -0,0 +1,29 @@
+# VoiceFlow — Domain Glossary
+
+The shared language for VoiceFlow. Terms here are the canonical names used in code,
+DB schemas, RPC methods, logs, and tests. UI strings may diverge (see "Display label"
+where it does).
+
+## Recording
+
+A captured audio file plus its optional transcript and optional AI-generated summary.
+The unit of work for the meeting-notes feature.
+
+A Recording is created in one of two ways:
+1. **Live capture** — user starts the recorder in the app, audio is streamed to disk from
+   one or more sources (mic, system-audio loopback), then stopped.
+2. **Import** — user picks an existing audio file from disk; it is copied into VoiceFlow's
+   recordings directory and treated as if it had been live-captured.
+
+A Recording is the entity in the DB (`recordings` table), in RPC methods (`recordings_*`),
+in services (`RecordingService`, `RecordingsRepository`), in the audio folder
+(`~/.VoiceFlow/recordings/`), and in logs (`domain=recording`).
+
+**Display label**: the UI surfaces this feature as "Meetings" (the dashboard nav item, page
+titles, button copy) because the dominant use case is multi-party meeting capture. A
+Recording is shown to the user as a "meeting" in the interface, but it is a `recording`
+everywhere else. This split is intentional — see `docs/adr/0002-recording-vs-meeting-naming.md`.
+
+Distinct from **History** (one-shot push-to-talk dictations, `history` table) — Recordings
+are long-form (minutes to hours), have a structured lifecycle (recording → transcribing →
+summarizing), and live in their own table.