For IT: installation, data folders, network

Where Dictate installs and keeps its data, which addresses it reaches, how updates are verified and which environment variables it reads.

Applies to: Windows, macOS
Version: Dictate 0.4.0-beta.6

Installation


Setup	`MediSync-Dictate-Setup.exe`, NSIS, 64-bit (x64)
Installation	Per Windows user, without administrator rights, into `%LOCALAPPDATA%\Programs\medisync-dictate`
Signature	The setup and the program files are signed by "MediSync GmbH".
Autostart	Per user, starts without a window in the notification area. The user turns it off under Settings›System›Start with Windows.
Uninstall	Through Windows Settings, per user. The data stays in place.

Dictate runs without administrator rights. If a user dictates into a program that runs with administrator rights, Windows blocks Dictate's input.


Package	`MediSync-Dictate-arm64.dmg`, Apple chip (arm64) only, macOS 14 or later
Installation	Drag `MediSync Dictate.app` into the Applications folder
Bundle ID	`me.medisync.dictate`
Signature	Signed by MediSync and notarized by Apple
Permissions	Accessibility, Input Monitoring, Microphone; for system audio in Notes the audio capture (macOS 14.2 or later)
Autostart	When the user logs in to the Mac. The user turns it off under Settings›System›Start when you log in to the Mac.
Uninstall	Drag the app to the Trash. The data stays in place.

macOS applies updates only when the app is in the Applications folder.

Data folders

All data is in the user's profile:

System	Folder
Windows	`%LOCALAPPDATA%\MediSync\Dictate` (local, non-roaming profile)
macOS	`~/Library/Application Support/MediSync Dictate`

The main files and folders in it:

Path	Contents
`settings.json`	Settings, dictionary, snippets, dictation counts of the last 14 days
`secrets.json`	Sign-in token, licence key, calendar and Slack access, encrypted with DPAPI (Windows) or the Keychain (macOS)
`dictation-history.bin`	Dictation history of the last 7 days, encrypted
`notes`	Notes, encrypted, without audio
`scratchpad`	Scratchpad, encrypted
`voiceprofile.json`	Voice profile
`documents`	Smart Find search index (`index.sqlite`)
`mcp`	State and connection data of the local MCP server
`slack`	Settings of the Slack connection
`team-wordbook.json`, `team-book.json`	Last loaded team words
`prebuilt-wordbooks.json`	Specialty vocabulary of the chosen profession
`wordbook-suggestions.json`, `wordbook-use.json`	Word pairs from corrections and how often entries are used
`usage-stats.json`	Numbers for Insights
`usage.json`	Counters for the licence key: dictations and minutes, signed
`models`	MediSync Local models
`logs\dictate.log`	Log. It contains no dictated text.

Dictate does not write dictation audio to disk. What Delete local data removes is in Privacy in Dictate.

Network

All connections use HTTPS on port 443.

Address	What for	Needed for
`app.medisync.me`	Sign-in, access, plan, team words, usage counts; with MediSync Cloud, Smart Dictation and the assistant; summary, questions and translation of notes	Every installation with an account
`ms-speech.medisync.me`	Speech recognition in the cloud, HTTPS and WebSocket (WSS)	MediSync Cloud only
`models.medisync.me`	One-time model download	MediSync Local with an account
`downloads.medisync.me`	Setup and updates	Installation and updates
`medisynchealth.ai`	Release notes, referral program, the film in the walkthrough, relay of the Slack sign-in	Optional
`accounts.google.com`, `oauth2.googleapis.com`, `www.googleapis.com`	Google Calendar, read only	Only when a user connects Google Calendar
`login.microsoftonline.com`, `graph.microsoft.com`	Outlook calendar (Microsoft 365), read only	Only when a user connects Outlook
`slack.com`	Slack connection	Only when a user connects Slack

The MCP server runs on 127.0.0.1 only, only when the user turns it on under Settings›AI apps (MCP), and only while Dictate is open. Other computers cannot reach it, see MCP.

A workstation that must never be online needs none of these addresses. It gets a licence key and the model package, which you load with Import model package, see Model packages. After the first-run setup, a computer with MediSync Local keeps dictating without a connection, with no time limit. MediSync Cloud stops when access could not be checked for 7 days.

Updates

Dictate checks downloads.medisync.me 30 seconds after it starts and then every hour, in the channel Stable (default) or Beta. It installs an update only if the signature of the update list matches MediSync's Ed25519 key and the checksum of the file is correct. It installs after 10 minutes without a dictation, otherwise when the app quits. Automatic updates cannot be turned off in the app. More in Updates.

Environment variables

These variables replace the default addresses. Set them only in agreement with MediSync, for example for a dedicated MediSync instance.

Variable	Default
`MEDISYNC_API_URL`	`https://app.medisync.me`
`MEDISYNC_SPEECH_URL`	`wss://ms-speech.medisync.me/api/v2/ws/transcribe`
`MEDISYNC_MODELS_URL`	`https://models.medisync.me/dictate`
`MEDISYNC_UPDATES_URL`	`https://downloads.medisync.me/dictate`, one folder per channel
`MEDISYNC_GOOGLE_CLIENT_ID`, `MEDISYNC_GOOGLE_CLIENT_SECRET`	Google sign-in for the calendar, built into the program
`MEDISYNC_MS_CLIENT_ID`	Microsoft sign-in for Outlook, built into the program
`MEDISYNC_SLACK_CLIENT_ID`	Slack app, built into the program
`MEDISYNC_SLACK_REDIRECT`	`https://medisynchealth.ai/api/dictate/slack/callback`, or `loopback` for a redirect straight to this computer

Did this article help you?