Why this lives here. The capture server is a dedicated VM for infrastructure reasons only — it holds the device WebSocket fleet and the raw audio, and it has to keep recording when Data OS is down. It is not a separate product. Operations belong here, where the Salesforce roster, the session agenda and every enrichment surface already are. Listen links are public — a listen link streams live while the room is recording and plays back on demand after; revoke a link when it has served its purpose. Fleet MDM: on managed devices (HotMic is Android device owner) you can kiosk-lock, silently update, reboot, locate and sleep from here — one device or the whole fleet. Wake needs no provisioning and reaches every connected unit. Sleep blanks the screen only: a recording device keeps recording and keeps streaming. Factory wipe deliberately stays on the capture server, behind an operator login and a typed device-id confirmation. Storage shows what each unit is holding on its own disk and reclaims it — the cloud copy is never touched, and a file that has not been uploaded is flagged as the only copy before anything erases it.

All statuses Recording Completed Failed Needs certification only Needs master Shorter than sec min Transcribe with Server default

Delete is irreversible. A recording is the raw asset — Delete removes the audio on the capture server (and its cloud copy). There is no undo. Certification, by contrast, is always correctable: re-certify or use Change to rebind.

Defaults are pre-assigned. Email the PDFs to the contract primary contact — To = speaker + primary contact, CC = carbon copy (edit per sponsor from the full role list). The Event Overview attaches to every sponsor. Sent sponsors are flagged from the send log to prevent duplicates. Each send opens a one-click confirm — the platform requires a human confirm per email.

Incremental Sync Full Re-sweep

How the connection works. Data OS polls the GovTech Audio API (GET /api/v1/items, X-API-Key) with a since watermark, so each sync pulls only new or updated sessions; the capture server also pushes a finished transcript the moment it is ready. Records land in the shared hotmic_audio store, carrying salesforce_event_id / salesforce_session_id. Speaker attribution and the session-summary document are generated HERE (services/audio_refinery) — a nightly pull never overwrites a Data OS refinement. Set HOTMIC_API_KEY and HOTMIC_BASE_URL for the library sync, HOTMIC_ADMIN_KEY for fleet operations; run .\scripts\refresh_govtech_audio.ps1 for a scheduled pull.