Notifications, inbox & voice settings
Settings reference for what appears in the unified inbox, how and when the app alerts the user, and how voice commands and text-to-speech behave.
On this page
- Alarm assertiveness
- Inbox analytics
- Inbox digest quiet hours
- Show sessions in inbox
- Show SMS in inbox
- Show Telegram in inbox
- Show daily digest in inbox
- Show calendar reminders in inbox
- Show cron jobs in inbox
- Show automations in inbox
- Show CLI approvals in inbox
- Show recipes in inbox
- Send new system alerts to the crew that owns them
- Send alerts no crew owns to the Inbox Alerts crew
- Crew that receives unclaimed alerts
- Inbox order
- Low disk space alerts
- Low disk space floor
- Daily weather report
- Weather report place
- Weather report delivery time
- Weather report archive time
- Weather report units
- SMS offline alerts
- Meeting Rooms
- Meeting call auto-detection
- Live transcript indicator
- Speaker attribution
- Summary language
- Transcript auto-delete
- Transcription language
- Wake word detection
- In-app notifications
- Error/warning display time
- Notification timing
- Notification history
- Notification sounds
- Needs-you sound
- Stalled sound
- Error sound
- Notification volume
- Mobile sound
- Text-to-speech
- Auto-read replies
- TTS daily cost cap
- TTS provider
- TTS speed
- TTS volume
- TTS character limit
- TTS summarization mode
- Show TTS button in session header
- Voice commands
- Voice feedback sounds
- Voice response style
- Speech-to-text provider
- Live talk-back voice provider
- Live talk-back model
- Live talk-back daily spend limit
- Let me cut in during live talk-back
- Gemini voice for the inbox walk
- Filter spam
- Filter spam in texts
- Filter spam in agent email
- Filter spam in calendar invites
- Remove invites from blocked senders
Controls what shows up in your unified inbox, how and when Omniscio alerts you, and how voice commands and text-to-speech behave.
Alarm assertiveness
Choose how forcefully an alarm interrupts you when it fires. Full-screen modal (the most assertive option) takes over the screen until you dismiss it — use this for alarms you cannot afford to miss. Banner toast pops up in the corner and fades; it still plays a sound. Silent banner shows the same corner pop-up but skips the sound entirely, so you see the alert without being interrupted mid-call or mid-focus. You can override the assertiveness for any individual alarm; this setting is the fallback default for alarms that do not specify their own. Accepted values: "modal", "banner", "silent-banner".
Inbox analytics
When turned on, Omniscio quietly tracks how you work with inbox items — how long things wait before you clear them, how often you open the inbox without acting, and roughly how much time you spend on each item. That data feeds the Stats → Inbox tab so you can spot patterns over time. Everything stays on this device; no message content, phone numbers, or identifiers are ever stored. Turn it off if you do not want that tracking at all. Default: on.
Inbox digest quiet hours
The overnight window in which no inbox digest is delivered, and the single summary time that carries whatever the night had to report. On by default: quiet from 22:15 to 08:15, with one summary at 08:15. Set under Overseers hub → Fleet settings → Quiet hours, and read by every digest producer, so the app and its agents agree on one window. The times are your own local clock — the hours you chose to sleep through — not a time zone. Turn it off to receive digests through the night. Accepted values: { enabled, start, end, summaryAt }, each time an HH:MM string.
Show sessions in inbox
Controls whether Claude Code sessions that need your attention appear in the unified inbox. When on, any session that has finished a turn and needs your reply — an answer, a question, a plan approval, a permission request, a mission, or a merge-conflict handback — shows up as an inbox row. Errored, stalled, stopped, recovery-failed, and rate-limit-waiting sessions never appear in the inbox; they wait in their project's Interrupted section in the sidebar instead. Turn it off to keep the inbox clear of session cards and rely only on the sidebar indicators. Default: on.
Show SMS in inbox
Controls whether incoming text messages routed via Pushbullet appear as rows in the unified inbox. When on, each incoming SMS creates an inbox card you can read and reply to without leaving Omniscio. Turn it off if you prefer to handle texts in your phone's own messaging app and want to keep the inbox free of SMS cards. Requires the SMS integration (Pushbullet) to be set up separately. Default: on.
Show Telegram in inbox
Controls whether incoming Telegram messages appear as rows in the unified inbox. Omniscio connects to Telegram via the MTProto protocol directly — no bot required. When on, new Telegram messages land in your inbox alongside your other communications. Turn it off to stop Telegram cards from appearing there. Requires a Telegram account to be connected in Settings. Default: on.
Show daily digest in inbox
Controls whether your daily briefing summary appears in the unified inbox. When on, the morning digest card — a summary of overnight AI session activity across all your hubs — shows up as an inbox row. Turn it off to stop the digest from cluttering the inbox; the digest is still generated and accessible elsewhere. Default: on.
Show calendar reminders in inbox
Controls whether Google Calendar reminders and agenda items appear in the unified inbox. When on, upcoming events and reminders land in your inbox alongside sessions and other channels. Turn it off to keep the inbox free of calendar cards and rely on the Calendar panel instead. Requires Google Calendar to be connected. Default: on.
Show cron jobs in inbox
Controls whether cron jobs waiting for your approval appear in the unified inbox. When on, a cron job with a pending approval creates an inbox row so you can approve or reject it there. Cron run results are not inbox rows, and cron failure alerts are CLI approval cards that follow the "Show CLI approvals in inbox" switch instead. Turn it off if you want to handle cron approvals only from the Cron Jobs hub. Default: on.
Show automations in inbox
Controls whether automation rules waiting for your approval appear in the unified inbox. When on, a rule with a pending approval creates an inbox card so you can approve or reject it there. Turn it off if you would rather handle automation approvals only from the Automation Rules hub. Default: on.
Show CLI approvals in inbox
Controls whether pending CLI approval requests appear in the unified inbox. When on, any change an agent or external tool has asked Omniscio to make through the local control server (and cron failure alerts) shows up as an inbox row, prompting you to act. These are not tool-permission prompts inside a session; those show on the session's own row. Turn it off to suppress these cards — but be aware that pending requests will then only be visible via the CLI approval panel. Default: on.
Show recipes in inbox
Controls whether recipe items that need you appear in the unified inbox. When on, recipe approvals, recipe-authoring approvals, and recipe runs that failed, stalled, or stopped incomplete create inbox rows. Successful runs do not. Turn it off if you prefer to check recipes in the Recipes panel rather than the inbox. Default: on.
Send new system alerts to the crew that owns them
On by default. When an alert type names the crew that owns it, a new card of that type is delivered to that crew's lead instead of to you, and the card is kept out of your inbox for as long as the crew holds the copy. Alert routes you set yourself always win: this only applies when none of your routes matches the type. Safety alerts — sign-in, backups, a lost git store, memory pressure — and the cards that exist to reach you alone (the dark-alarm summary, a cloud session reaching for your private data, a blind dispatch trace, an unlanded-branch backup failure, a box holding uncaptured work, the disk-space guardian, and the phone device-approval cards) are never sent this way. If the named crew has no live lead when the card is raised, nothing is routed and the card stays with you. Turn this off to keep every card in your inbox exactly as before this existed, and any card this rule already took out of your inbox comes back. Only you can change this — an agent's own token is refused. Accepted values: a boolean; the setting is alertRegistryOwnerRouting.
Send alerts no crew owns to the Inbox Alerts crew
On by default. A new system alert that no alert route claims and no alert type owns is delivered to the crew you name in the setting below instead of to you, and the card leaves your inbox for as long as that crew holds the copy. One card a day lists what was sent away, family by family, with how many of each and the crew that received it; that summary stays with you and is never sent away itself. Sign-ins, security, approvals, people waiting, disk warnings, the never-hide cards and your own digests and reports are all kept back for you, and so is any card that asks you a question. A card the crew cannot receive anywhere comes straight back to your inbox. With no crew named below, every one of those cards stays with you. Turn this off to keep every card in your inbox exactly as before this existed, and any card this rule already took out of your inbox comes back. Only you can change this — an agent's own token is refused. Accepted values: a boolean; the setting is alertUnclaimedCrewRouting.
Crew that receives unclaimed alerts
The id of the crew that receives new system alerts nobody owns. It belongs to one install — a crew id minted on one computer names nothing on another — so it ships empty and stays empty until you set it. Leave it empty and every unclaimed card stays in your inbox. Set it to a crew that has nobody live, or to an id that names no crew at all, and the cards stay with you too; the app writes one line to its log saying which id it could not reach. The id is the one shown on the crew in the Overseer panel, and it is never matched against a crew name — a name can be claimed by anyone, so only the id counts. Changing or clearing the field hands back any card the previous crew was holding as soon as nothing else keeps it away. The app also fills this in for you once, the first time it starts after an update: if the field is empty and a crew on this computer is still holding alerts that nothing else owns, that crew's id is written here so those cards are not dumped back on you — an install with no such cards is left empty. Only you can change this — an agent's own token is refused. Accepted values: a crew id, up to 64 characters; the setting is alertUnclaimedCrewId.
Inbox order
Choose how the inbox is ordered. Oldest first (the default) groups the inbox by project, with the project that has waited longest at the top and the oldest item first inside each project. Newest first drops the project groups and shows one list of every item by time, the most recent at the very top — so an old item is never pulled up just because its project has a newer one. In Newest first each item starts with its hub's name in brackets, such as [My App], in the hub's colour when it has one. The per-project menus that sit on group headers (approve all, dismiss all) are not shown there, because there are no group headers. Switching back to Oldest first restores the project order you had. Change it from the Sort section at the top of the inbox filter menu (the sliders icon in the inbox header), or here. Accepted values: "oldest", "newest".
Low disk space alerts
When on, Omniscio warns you in the inbox if free disk space drops below your configured floor and automatic cleanup cannot recover enough space. The warning is an inbox card, not a system notification. Turning this off silences only the warning — the automatic disk cleanup (which removes finished worktrees to protect your drive) keeps running regardless. Default: on.
Low disk space floor
The minimum amount of free space Omniscio tries to keep on your working drive, in gigabytes. When free space drops below this number, finished worktrees are automatically cleaned up. If cleanup cannot recover enough space and low-disk alerts are on, you receive an inbox warning. Range: 5 to 500 GB. Adjust upward if you work with very large repos or want a bigger safety buffer. The same number also decides whether a drive has room for a new worktree when a session starts, at every step of that create and on the emergency create road too. A worktree that needs its own full copy of the dependencies (rather than sharing them) still needs the larger built-in reserve. On any drive, Omniscio never asks for more than half of that drive to be free, so a number bigger than half a small drive acts as half of it rather than blocking new sessions there. Installs that saved 60 back when 60 was filled in automatically have that value cleared once, so they get the automatic floor, which scales with the drive size; if you really want 60, set it again and it stays.
Daily weather report
When on, Omniscio posts one forecast card to your inbox each day at your delivery time: tonight, tomorrow (high and low, conditions, chance of rain, wind, sunrise and sunset), and a 3-day outlook. The forecast comes from the free Open-Meteo service, which needs no account or key, and no AI is involved, so it costs nothing. If you have not opened the card by your "Archive if unopened at" time it archives itself; a card you opened stays until you put it away. Off by default. It needs a place to be set (see weatherReportLocation) and does nothing while that is empty. Settings → Notifications → Daily weather report. For the full behaviour see Daily weather report.
Weather report place
The place the forecast is for, filled in for you the first time the weather is needed: Omniscio asks a free public lookup (ipwho.is) roughly where the computer's internet connection is and writes in the ZIP code (in the US) or the town and country anywhere else. That lookup request carries your public IP address and nothing else. You can type over it any time, the way you would say it — for example Austin, TX, or a ZIP code such as 80302 (up to 200 characters) — and it is saved when you leave the field. If Omniscio cannot find the place, add a state or country; the Send a test report button tells you which place it found or why it could not. Only this text and the coordinates it resolves to are sent to Open-Meteo. Only shown while weatherReportEnabled is on.
Weather report delivery time
The time of day the weather card is posted to your inbox, in your computer's local time. Stored as a 24-hour HH:MM string; default is 06:00 (6:00 AM). If Omniscio is not running at that moment, the card arrives as soon as it is running again, but never after the tidy-up time (weatherReportArchiveTime) — a late launch skips that day rather than posting a card that would be archived at once. A card posted before noon leads with today; a card posted later leads with tomorrow. Only shown while weatherReportEnabled is on.
Weather report archive time
The time of day an unopened weather card archives itself. Stored as a 24-hour HH:MM string; default is 20:00 (8:00 PM). A card you opened — on the desktop or on your phone — is never archived by this; only cards still unopened are tidied away, and only weather cards. If this time is earlier than the delivery time, the tidy-up falls on the next day. The tidy-up keeps working even if you later switch the report off. Only shown while weatherReportEnabled is on.
Weather report units
Whether the forecast is written in Fahrenheit with miles per hour (fahrenheit, the default) or Celsius with kilometres per hour (celsius). Changes the next card; cards already in your inbox are not rewritten. Only shown while weatherReportEnabled is on.
SMS offline alerts
When on, Omniscio raises an inbox alert and a single notification if SMS is turned on but stops working — for example, if your Pushbullet token expires or the connection drops for a sustained period. The alert clears automatically once SMS reconnects, so you do not need to manually dismiss it. Turn it off if you are OK discovering SMS outages on your own. Only relevant while the SMS integration is enabled. Default: on.
Meeting Rooms
Shows the Meeting Rooms panel in the sidebar and enables the Ctrl+Shift+Z keyboard shortcut to open the quick-launch palette. Meeting Rooms let you save your recurring video call links (Zoom, Google Meet, Teams, etc.) for one-tap joining. Your saved rooms are preserved when you disable this — turning it back on restores them. Turn it off if you do not use saved meeting links and want to keep the sidebar uncluttered. Default: off.
Meeting call auto-detection
When on, Omniscio watches for an active audio call on your computer — including browser-based calls like Google Meet and Microsoft Teams — and automatically shows a heads-up overlay so you can start capturing notes without manually starting a recording. Turn it off if you always start meetings manually or find the automatic prompt distracting. Default: off.
Live transcript indicator
When on, a small floating window appears on screen while a meeting is being recorded, showing the live transcript as it is being captured. It gives you a real-time view of what is being transcribed so you can catch any errors immediately. Turn it off if you find the overlay distracting and prefer to review the transcript afterward. Default: on.
Speaker attribution
When on, the transcript labels each line with who said it — distinguishing your voice from other participants. This makes the transcript easier to follow when reviewing a meeting with multiple speakers. If your transcription provider does not detect multiple speakers cleanly, turning this off can produce a cleaner single-stream transcript. Default: on.
Summary language
The language Omniscio uses when writing the AI-generated notes and summaries for your meetings. This is the OUTPUT language — it controls what language the summary is written in, not what language is spoken. Set it to match the language you want to read your notes in. You can set this independently from the transcription language, so you can speak in one language and receive notes in another.
Transcript auto-delete
Automatically deletes the raw meeting transcript after a set period to save space and protect privacy. Options are to never delete, or to delete after 30 days, 90 days, or 1 year. Your AI-written notes and summaries are always kept — only the raw transcription text is removed. Set to Off if you want to keep raw transcripts indefinitely. Default: off.
Transcription language
The language spoken in your meetings, used to improve the accuracy of the live transcriber. Setting this correctly helps the speech recognition engine handle accents, vocabulary, and phrasing for the language you actually speak during calls. You can set it independently from the summary language.
Wake word detection
When on, Omniscio listens for one of two wake phrases — say "Hey Omniscio" or "Computer" — to activate voice commands without touching the keyboard. This keeps your hands free during tasks. The built-in engine (OpenWakeWord) runs locally at no cost and requires no account or API key. Requires voice commands to also be enabled. Default: off.
In-app notifications
Every pop-up message in the app belongs to a named kind — confirmations, deletions, notices — and each kind has its own on/off switch. You can reach them two ways. Open Settings → Notifications → In-app notifications and work down the list, or take the shortcut: every pop-up carries a small settings (gear) icon beside its close button, labelled with the kind it belongs to (for example "Notification settings: Clipboard copies"). Pressing it opens this same list already scrolled to that pop-up's kind, with the row highlighted and its switch ready — so you can turn a noisy kind off in one click, straight from the pop-up that annoyed you, then come back to what you were doing.
Turning a kind off silences that kind only, and the choice is remembered. Two things do not move: Errors and Warnings stay always on, whatever you do — a failure must never be silenceable — and pressing the gear never performs the pop-up's own action or closes it. Pressed from a popped-out window (Writer, KMS, Scratchpad), the gear opens the same row in the main window, because that is where Settings lives.
Error/warning display time
How long error and warning toasts stay on screen before auto-dismissing. Choose from preset durations: 5, 8, 10, 15, or 30 seconds. The default (8 seconds) matches the previous fixed duration. Success and info toasts are unaffected and always use a shorter fixed duration. If code passes an explicit duration when creating a toast, it overrides this setting. Number (seconds); accepted values: 3–30. Default: 8.
Notification timing
Controls when Omniscio sends OS-level notifications to your taskbar or notification center. Always sends a notification every time a session needs attention, even if Omniscio is in the foreground. Backgrounded (the default) sends notifications only when Omniscio is in the background or minimized — so you are not double-alerted while already looking at the app. Never disables all OS notifications from Omniscio entirely. Accepted values: "always", "backgrounded", "never". Default: "backgrounded".
Notification history
When on, Omniscio keeps a local log of every notification sound decision — whether a chime fired, was suppressed, or was rate-limited — so you can always look up "what was that sound?" in Settings → Notifications. The log stays on this device and is bounded in size. Turn it off to stop recording chime history; the notification sounds themselves are unaffected. Default: on.
Notification sounds
Master switch for all notification chime sounds in Omniscio. When off, no sounds play for any session event — needs-you, stalled, or error — regardless of the per-event sound settings below. Turn it off for a fully silent experience, or use a headphone-friendly alternative. Default: on.
Needs-you sound
The sound Omniscio plays when a session switches to "Needs You" — the agent has asked a question or is waiting for your reply. You can set a custom audio file path or use one of the built-in sounds. The default is builtin:alert-ping. Works only when notification sounds are on and the volume is above zero.
Stalled sound
The sound Omniscio plays when a session appears to have stalled — it has been running without producing output for longer than expected. You can set a custom audio file path or use one of the built-in sounds. The default is builtin:soft-bell, which is intentionally gentler than the needs-you alert. Works only when notification sounds are on.
Error sound
The sound Omniscio plays when a session hits an error state. You can set a custom audio file path or use one of the built-in sounds. The default is builtin:attention. Works only when notification sounds are on.
Notification volume
Sets how loud all notification sounds play, from 0 (silent) to 1 (full volume). The default is 0.5 (half volume). This applies to every notification chime — needs-you, stalled, and error alerts all share this volume setting. Alarms use the same volume by default but can be overridden per alarm.
Mobile sound
When on, Omniscio plays notification sounds on the mobile web app in addition to the desktop. Off by default because mobile browsers often restrict auto-playing audio. Turn it on if you use the mobile interface and want audio chimes there as well.
Text-to-speech
Master switch for Omniscio's read-aloud feature. When on, you can have session replies spoken aloud via a voice synthesis provider of your choice. Turning it on unlocks the TTS controls below and reveals the speaker button in the session header (if showSessionVoiceButton is also on). Off by default. Requires a supported TTS provider to be configured with a valid API key.
Auto-read replies
When on, Omniscio automatically speaks each new completed session reply aloud as it arrives, without you pressing the play button. Use this for a fully hands-free experience. Off by default — when off, replies are only read when you manually press the speaker button. Requires ttsEnabled to be on.
TTS daily cost cap
The maximum amount Omniscio will spend on text-to-speech audio synthesis each calendar day, in US dollars. Once the cap is reached, read-aloud stops for the rest of the day. Range: $0 (disables TTS) to $100. Default: $5. Adjust upward if you use TTS heavily and keep hitting the cap.
TTS provider
The voice synthesis service used to generate spoken audio. Each provider has different voice styles, latency characteristics, and pricing. Options: Fish Audio (the default), Grok (xAI), ElevenLabs, Speechify, Pika, and Soniox. Fish Audio and Soniox are company-funded (no personal API key needed); the others bill to your own provider key. Requires ttsEnabled to be on.
TTS speed
How fast the synthesized voice speaks, from 0.5× (half speed) to 2× (double speed). The default is 1.0 (normal speed). This setting bakes the tempo into the generated audio at synthesis time. If you want to adjust playback speed without re-synthesizing, use the narration playback rate setting instead. Requires ttsEnabled to be on.
TTS volume
The playback volume of spoken audio, from 0 (silent) to 1 (full volume). The default is 0.8. This is separate from the notification sound volume — it controls only the TTS read-aloud level. Requires ttsEnabled to be on.
TTS character limit
The maximum length of a single message that Omniscio will read aloud in full, measured in characters. Messages longer than this limit are handled according to your summarization mode setting. Range: 100 to 10,000 characters. Default: 2,000. Increase it if you often want very long replies read verbatim; lower it to keep spoken replies brief.
TTS summarization mode
What Omniscio does when a reply exceeds the character limit above. Verbatim reads the full text regardless of length. Summarize always condenses the reply before speaking it. Auto (the default) reads short replies in full and summarizes only when the reply exceeds the threshold below. Accepted values: "verbatim", "summarize", "auto".
Show TTS button in session header
When on, a speaker icon appears in the session header bar so you can trigger read-aloud from within a session without opening settings. Hidden by default to keep the session header uncluttered. Requires ttsEnabled to also be on — enabling this setting alone will not show the button until TTS is turned on. Default: off.
Voice commands
Master switch for voice input and commands. When on, Omniscio listens for your voice via the microphone — you can speak to send messages, issue commands, and navigate the app. Turning it on also unlocks wake-word detection and the voice command settings. Off by default. This is the top-level gate; all other voice features require this to be on. Enabling voice implies consent to send your audio to the configured speech-to-text provider.
Voice feedback sounds
When on, Omniscio plays small audio cues to confirm voice actions — for example, a brief tone when it starts listening, finishes processing, or completes a command. These sounds help you know your voice input was heard without looking at the screen. Turn it off for a fully silent voice experience. Requires voiceEnabled to be on. Default: on.
Voice response style
Controls how the AI responds when you send a message by voice. Off means the AI replies the same way it would for a typed message — no length adjustment. Short keeps replies brief and to the point, better suited for spoken responses. Conversational strikes a middle ground: friendly and natural, but not as terse as "short." Accepted values: "off", "short", "conversational". Default: "off".
Speech-to-text provider
The service that converts your spoken audio to text for voice input. Options: Deepgram Nova-3 (the default — fast, streaming), ElevenLabs Scribe v2 (lowest latency), Groq Whisper v3 Turbo (transcribes after you stop speaking), Grok/xAI (in development, requires xAI key), and Built-in/local (free, runs on your own computer — revealed by localSttEnabled). Each provider has different latency, accuracy, and cost characteristics. Requires voiceEnabled to be on. Default: "deepgram".
Live talk-back voice provider
Whose live voice is used when you press Talk it through during an inbox walk. Google Gemini is the default; OpenAI Realtime is the other choice. Each needs your own key for that provider, saved in Settings. A live walk uses only the provider you pick: if its key is missing or refused, the walk carries on with the ordinary voice and says so once out loud. Changing the provider also changes which models the next setting lists.
Live talk-back model
The model that listens and speaks during a live walk, chosen from a built-in list for the provider above. For Gemini that is Gemini 3.8 Live (the default) or its extended-thinking variant, which takes longer over each answer; for OpenAI it is GPT Realtime 2.1 or the cheaper GPT Realtime 2.1 Mini. A saved model that belongs to the other provider is ignored and that provider's default is used instead.
Live talk-back daily spend limit
The most live talk-back may spend in one day, in US dollars. The default is $1.00 and you can set anything from $0 to $100; 0 turns the limit off. Spend is counted by the second of audio, today's total is shown in this card, and when you pass the limit you are told out loud once. A walk already under way keeps going; after that, new live walks use the ordinary walk until tomorrow.
Let me cut in during live talk-back
When on (the default), you can talk over the live voice and it stops at once and listens. The microphone always uses echo cancellation, and the voice service's own signal decides when you have cut in, never how loud the room is. Turn it off for strict take-turns, where the microphone stays shut while the voice is speaking.
Gemini voice for the inbox walk
When on (the default), the inbox walk speaks in one natural Gemini voice at one steady speed: the walker's own lines, each session's recap, the read-back before a send, a session's answer and the answers to your follow-up questions. It is the voice you picked for Gemini text-to-speech (geminiTtsVoiceId, Kore until you choose another), and a live walk pins the same voice so the live conversation and the read-outs sound like one speaker. It needs your Google (Gemini) key. Once today's shared Gemini-voice spend limit (the live talk-back daily limit) is used up - or if the Gemini call itself fails - the walk uses your regular voice, at the same one speed, and tells you once. With no Gemini key at all there is no voice to lose, and setting that daily limit to 0 switches the voice off outright, so in either case the walk just uses your regular voice and says nothing about it. Turn it off to keep the walk on your regular voice (there is no Settings row for it yet - ask Omniscio or an agent to set it, or use the settings API). Regular narration and read-aloud outside the walk never change. The walk itself is still an in-development feature and stays hidden until it is revealed in Settings → Lab. Toggle; accepted values: true | false. Default: true.
Filter spam
Checks messages from people you don't know and keeps junk in Spam. Messages from people you know are never checked. Off by default, so nothing is filtered until you turn it on; the per-channel switches below are then live, and each is disabled until this one is on. Toggle; accepted values: true | false. Find it in the Inbox hub's Spam tab.
Filter spam in texts
Checks texts from numbers that aren't saved as contacts and that you have never texted. Short-code senders such as bank and login codes are never checked. On by default, but it does nothing until Filter spam is on.
Filter spam in agent email
Checks new email from strangers to your agent's address and your public support address. Caught mail starts no session and gets no reply, so junk sent to your agent cannot quietly spend a session. On by default, but it does nothing until Filter spam is on.
Filter spam in calendar invites
Checks calendar invites from organizers outside your own organization. A caught invite is hidden rather than announced — you get no notice for it. On by default, but it does nothing until Filter spam is on.
Remove invites from blocked senders
Also deletes an invite from your Google Calendar when it comes from a sender you have marked as spam. Invites the automatic check catches are only hidden, never deleted; this is the switch that removes them. Deleted invites stay in Google Calendar's trash for 30 days. On by default, but it does nothing until Filter spam is on.