Team Chat (built-in channels and direct messages) (part 2)
This is part 2 of Team Chat: the message-surface polish and composer upgrades, saved items, custom status, scheduled send, attachments and custom emoji, plus the workspace story — per-channel roles and invitations, self-serve workspaces you create and share by invite code, and 1:1 connections to other organizations.
On this page
What it is
This is part 2 of the Team Chat page. It covers what came after the first version — message-surface polish, composer upgrades, saved items, custom status, scheduled send, attachments and custom emoji — and the workspace story: per-channel roles and invitations, self-serve workspaces you create beside your company workspace, and 1:1 connections across organizations.
Where to find it
The same surface as the parent page: once switched on, Team Chat appears as a built-in virtual project in the projects sidebar, with the channel and DM list in the sidebar pane and the open conversation filling the main panel. It ships dark until Settings → Lab → Team Chat is turned on; the parent page covers finding and enabling it.
Everything here sits inside the chat surface: the features appear in the conversation, the composer, the sidebar header and channel context menus; self-serve workspaces are the switcher rail down the side of the chat sidebar; cross-organization connections come from the same New-message picker. Per-channel notification controls and the bell menu are on part 3.
How it behaves
New in this wave (v2 — the "destroy Slack" overhaul)
These features are built on desktop; the phone PWA renders the text and ignores the richer parts until its own parity work. Invariants and locking tests: team-chat-v2-features-contract.md.
- Message-surface polish — Slack-style day dividers (a centered "Today / Yesterday / July 2" pill) and a bold "N new messages" line above the first unread, in the full Team Chat view and in the inbox conversation embed (which fetches your read marker on open). It marks where your unread run starts; nothing scrolls to it. Opening a conversation lands on the NEWEST message, positioned at its top so a long final message is read from its start, once per open, so a message arriving while you read never pulls you away (an agent thread counts as its own open). A floating jump-to-latest pill counts what arrived while you were scrolled up; the floating up/down message arrows an agent session has step back one message per tap, holding one (or Shift/Ctrl/Cmd+clicking) jumps to the top of loaded history, and Shift+Up / Shift+Down step through messages from the keyboard while the up arrow is on screen. A live typing indicator sits under the composer, and DM rows carry online/away dots.
- Composer power — drafts that survive channel switches and restart (device-local, LRU-capped) and mirror live into a popped-out Team Chat window (unsent text appears in the other window for the same channel, thread or DM, both ways, never replacing text you are actively typing there); up-arrow on an empty box to edit your last message;
:emoji:autocomplete as you type; and auto-capitalization of the first word, sentence starts (after. ! ?) and new paragraphs, skipping URLs, paths and abbreviations. A word that is the first content of its line counts as a paragraph start even behind a list marker, so2. run testsbecomes2. Run tests. It reuses the shared@shared/smart-capitalizeengine, runs live on each word plus a finalize pass on send, is default ON with an opt-out in Settings → Sessions → Composer & input (autoCapitalizeTeamChat), and is locked by composer contractauto-capitalize-as-you-type. - Search, wired in — the sidebar magnifier, or the magnifier icon in the header of every channel and DM (there is no always-visible search box), opens a search overlay over the existing search index; Ctrl+Shift+F opens it too. Type ≥2 chars for results (channel + author + snippet), click to jump. It searches every channel and DM you can read; to narrow it, pick a channel in Filters → Channel, or, when you opened it from a conversation, click Only in #channel at the top. Clear filters searches everywhere again. Typing
in:#namewith a channel's exact name narrows the same way. Degrades to "search isn't switched on yet" when the search Cloud Function cannot be reached. - Saved items — a bookmark on any message plus a sidebar "Saved" view (own-doc
saved/{uid}, cap 200), newest-first. Each save captures a snapshot (author name, message text, channel name) so the panel renders with zero Firestore reads and survives message deletion. A pre-snapshot save falls back to fetching the message: a failed read (denied, unavailable, timed out) shows "Could not load this message.", while a message genuinely gone shows "This message is no longer available.". The panel pages in 20 at a time (Load more shows a spinner until the next page is in), and while you search it says "Searching loaded messages only", because only the loaded pages are searched. Opening a saved message lands on that message, not the bottom of its channel. - Custom status — an emoji + text + auto-expiry (30m / 1h / 4h / Today / This week / Don't clear), shown on the profile card, the member directory and beside DM names. Global (
workspaceId: null): it shows in every workspace. Expired statuses are actively cleared from Firestore at theexpiresAtmoment (already-past expiries clear on mount) and hidden client-side as a fast visual gate; the field-scopedcustomStatusmerge can't clobber the presence heartbeat, and vice-versa. Reopening the dialog to edit an existing status shows a Keep (clears at …) choice and keeps that clear time unless you pick a different one, so fixing a typo never silently makes an expiring status permanent. An emoji-only status shows the emoji alone in your profile popover, never the "Set a status" prompt beside it. - Custom status lands immediately — Setting or clearing it applies to your own row immediately and the dialog closes at once — the write goes out in the background, comes back to the old status only if the server refuses it, and a write that merely timed out is kept because the app retries it.
- Desktop notifications + unread badge — an app-level watcher raises native OS toasts for new messages while you're elsewhere (honoring per-channel mutes and Focus Mode) and folds an unread count onto the Team Chat nav row. It baselines on first load (no start-up storm) and skips the channel you're looking at. The same watcher drops a row into your unified Inbox — one coalescing row per channel, titled with the sender's name (resolved from the roster, never a raw id), refreshed as messages arrive — so a missed message is reliably waiting even when the OS toast was suppressed. It applies the same audience rule (DMs + @mentions + unmuted channels, never your own, skipping the channel you're reading), but the inbox row is not dropped for being a few seconds old or arriving in a burst. Messages that arrived while Omniscio was fully closed are swept in the next time you open it (silently, no late toast, only for channels you'd genuinely left unread). Each row stays until you read it, reply, or dismiss it. Selecting it embeds the live conversation and a composer: a reply posts to the channel and clears the notice (a failed send keeps it, with Retry); viewing marks the channel read, and the row clears once you move on (back, another item, or closing it — on the phone too), never while you're still reading. "Open channel" stays as a secondary jump, and after you dismiss, the next new message raises a fresh row.
- Thread replies aimed at you — when someone replies in a thread to a message you wrote, or @-mentions you in a reply, you get a separate, targeted Inbox row ("Jordan replied to your message in #general"), distinct from the channel's new-message row, cleared when you read the channel or open that thread, and clicking it opens straight into that thread. It never fires for your own reply, a muted channel, or a reply that isn't aimed at you, and an on-by-default switch in Settings → Team Chat notifications turns it off. The phone push still treats a reply as an ordinary channel message. Invariant I8 in inbox-alert-contract.md; details in team-chat-threads-contract.md.
- Reactions to your messages — opt-in and OFF by default. With
teamChatNotifyReactionsON, someone reacting to a message you wrote raises a separate, targeted Inbox row ("Jordan reacted 👍 to your message"; a custom emoji shows as its:shortcode:, and anything stored that is not a real emoji is left out, so the row reads "Jordan reacted to your message"), inbox-row only (no OS toast — reactions are higher-frequency and lower-signal), its own coalescing row that clears when you read the channel; never for your own reaction, a muted channel, or a reaction on someone else's message. Toggle in Settings → Team Chat notifications. Details: team-chat-reactions-contract.md. - New reactions since you last looked — always on, no setting. When someone reacts to a message you wrote, the conversation's row in the Team Chat sidebar shows that reaction's emoji as a small chip (beside the unread count when there are new messages too, and read out as "new reaction" by a screen reader). Opening the conversation shows a small notice at the top of the message list — "Jordan reacted 👍 to your message", or "3 new reactions to your messages" — with Show, which scrolls to the reacted message, and × to close it; the reacted pills carry an accent ring for that visit. A reaction on a thread reply or on a message above the loaded history still gets the notice ("…to one of your messages") but no Show button. Opening the conversation never jumps away from the newest message, and the unread badge and the "N new messages" divider work exactly as before. What counts as seen is kept per device: a reaction you saw on your phone can still show once on the desktop. Known limits: the chip relies on the conversation's most recent reaction, so a reaction that was overwritten while this computer was not running is not shown; and because a person's reactions on one message share one timestamp, adding or removing one of several emoji marks all of that person's pills on that message as new. Rules: R20 in team-chat-reactions-part2-contract.md.
- Remind me later — on your note-to-self (a DM whose only participant is you), a "Remind me later…" item in the header's ⋯ More actions menu and in the conversation's right-click menu in the sidebar both open the same snooze time picker you already use in the Inbox. Picking a time creates a durable reminder for that note and hides it until then, so it pops back into your Inbox at the time you chose — even a note you'd already read. Clicking the reminder opens the note; you can re-snooze or dismiss it like any Inbox row, and setting one offers undo. It reuses the existing snooze storage and is scoped to the note-to-self for now. Desktop-only — there is no CLI route for this action. Mechanics and the
team-chat-reminder:notice kind: inbox-alert-contract.md. - Attachments — paperclip, paste or drag-drop a file (client pre-validated to 50 MB, 100 MB for video, against a content-type allowlist mirroring the storage rules, ≤10 per message, filename sanitized), uploaded to Storage with progress + cancel (removing a chip mid-upload cancels quietly, with no "Upload failed" toast; a real failure still shows one). Staged chips show a thumbnail of the image itself (fitted whole, never cropped to a square) or a plain file icon; images render inline in the sent message, files as download cards. Send is blocked while uploads are in flight. An attachment-only send auto-fills the text with the filenames, but that fallback body is hidden when every attachment name matches (the cards already show the name). Before you send, a staged chip is removed by its
Xor a middle-click on the chip, either undoable (a toast Undo, or Ctrl+Z / Ctrl+Y) — a completed upload's bytes are not purged until the undo window has passed, so undo genuinely restores the file. Once a message is sent, its attachments are not middle-click-removable; removing a sent message is the existing delete flow, which best-effort cleans up its files after the undo window (and undo restores them). An inline picture keeps its shimmering box until the picture itself has finished downloading — not merely until its link is minted — and a picture that then fails to download shows the Couldn't load … Retry tile instead of a broken-image glyph.- Known limit: a few images per message in a DM or private channel. The server checks every image against its channel and sender, and its rules may spend only 1,000 steps on one message. In a DM that fits two images with their previews (three plain files) per message; a message with more is refused as "permission denied". Send larger batches as several messages. Raising the cap is an open decision.
- Up to ten files in one message, anywhere. A message that carries files is sent through the server (the
teamChatSendWithAttachmentscallable) instead of written straight to the database, because the database rules spend a limited budget of checks on one message and refuse a send that runs past it. The server makes every check those rules make — that you belong in the conversation, may post there, and that each file is yours and uploaded to this conversation — so up to ten pictures or files work in a DM, a channel or a thread, on the desktop and the phone. A message with no files is still written directly (and keeps working offline); that direct write is the one the limit above describes. If the server function cannot be reached, the app falls back to the direct write, which the rules still bound. Details: team-chat-workspaces-contract.md ("Rules-layer abuse bounds").
- Files you staged stay with their message box — switching channels and coming back keeps the files you attached but have not sent, the same way your typed draft is kept; a thread reply keeps its own files too. An image that fails to load says so, with Retry and Download.
- Long messages — near the length limit the message box shows a character count, and above it Send and Send later are turned off, so an over-length message is refused where you type it rather than after you send. On a phone, a message's action sheet also opens the full emoji picker, custom emoji included.
- Scheduled send — "Send later" (presets in your local time plus a custom picker with a date input and hour/minute/AM-PM dropdowns; 24-hour locales get 0–23 with no AM/PM) writes an author-owned doc at
channels/{channelId}/scheduledMessages/{id}; a sidebar "Scheduled" view lists and cancels (a cancel confirms with a "Scheduled message cancelled" toast). Each scheduled message enqueues one Cloud Task at its owndeliverAt— nothing per-message polls — and a 15-minute recovery job re-arms a message whose task was lost, within the last 24 hours. The task delivers in a per-doc transaction (deterministic idsched-<id>so it is idempotent, re-checks the author can still post, drops to a replayable dead-letter if the channel is gone), then deletes the doc. Cancelling arms nothing; an edit that moves the time arms the new time and leaves the old task to expire harmlessly; a task that fires early is re-armed. Security by construction: the function derives the target channel from the scheduled doc's own parent — never from a client field — so a scheduled doc can only post to the channel it was created under. - Who deleted a message is recorded — pressing delete stamps your own uid on the message for the instant before it is removed, so the server's delete record names a person instead of
system:message-delete-trigger; a delete that arrives with no stamp (a console/admin action, or a client that never stamped) is still written down, flagged as an unattributed path, so unanswered "who erased these 400 messages" is at least countable. Every delete also emits one structured line carrying a salted actor hash, so a burst is visible as a per-actor counter, and a per-identity daily delete budget bounds the cleanup work one account can drive: past it the trigger still runs the correctness steps (audit, purge queue, pin, topic copy, saved-copy redaction, reactions, rev bump) and leaves the derived caches — channel denorm, shared-media sweep, storage purge — to the queued purge and the next message. A replayed delete event is claimed once (a per-message marker), so an at-least-once re-fire does not re-run the pipeline.
Scheduled messages are owned by the person who scheduled them (desktop and CLI): every list, cancel and retry is scoped to the signed-in account, so the next account on the machine never sees the previous one's scheduled text or recipients. Signing out now holds that account's pending messages instead of canceling them — they keep their delivery time and cannot fire while held — and signing back in resumes the ones still ahead; one that came due while the account was signed out is ended as failed with the reason "It was due while you were signed out, so it was not sent. Retry it to send it now.", so a missed scheduled message is never silent.
How far ahead you can schedule — a scheduled message can be set from one minute to exactly 30 days ahead; a later time is refused with "Cannot schedule more than 30 days ahead", the same limit an agent hits through the CLI schedule routes.
Custom emoji — all workspace members upload custom images by default (PNG/GIF/WebP ≤256 KB, no SVG) with a unique
:shortcode:— via the emoji picker's "+" button, the sidebar management dialog, or the bulk upload dialog (up to 50 files at once, drag-and-drop, per-file validation, auto-deduplicated shortcodes, sequential upload with per-file progress). A workspace holds at most 200 custom emoji; past that an upload is refused with a "delete one to add another" message. Members can delete their own uploads; admins can delete anyone's (that delete goes through a server step that checks the admin role, and that server step is deployed and live; the Firestore rules clause that lets an owner or admin delete someone else’s emoji is not in the live rules release yet — a rules deploy is coming). A second emoji with a name that is already taken is refused with "Already exists in workspace", even when two uploads race. You can also rename an emoji you uploaded: in the management dialog click the pencil beside it, type the new name, and press Enter (Escape cancels). The picture stays the same, a name another emoji already has is refused, and the rename happens all at once or not at all. Messages and reactions that used the old name keep showing the picture. Deleting removes the picture first, and if the picture cannot be removed the emoji stays so the delete can be tried again. Custom emoji work in the picker's Custom tab, in reactions, in custom status, and inline in message text (type:shortcode:and the rendered message shows the image). An emoji picture is only ever loaded from Omniscio's own storage: one pointing anywhere else — or a reaction carrying someone else's link — shows the plain:shortcode:chip instead, so nobody can use an emoji to make your app call a server of their choosing. The picture travels WITH the message: on send, each custom emoji you can see is attached, so a teammate who doesn't have it still sees the image rather than a grey chip — their own copy always wins. Messages sent before 2026-10-08 or from an older app carry no picture, and neither does an emoji added by a later edit or inside a scheduled message. A clock time like12:30:00stays plain text. Your own uploads are also kept in a personal emoji layer, persisted by Firebase uid, so they remain available in DMs and after switching workspaces; the active workspace's emoji still win on shortcode collisions. The identifier formatcustom:<shortcode>slots into every surface a native emoji uses, and inline rendering reuses the @mention remark-plugin pattern (text node → internal-scheme link → component override, no innerHTML). Invariants: team-chat-custom-emoji-contract.md.Who deleted a message is recorded — pressing delete stamps your own uid on the message for the instant before it is removed, so the server's delete record names a person instead of
system:message-delete-trigger; a delete that arrives with no stamp (a console/admin action, or a client that never stamped) is still written down, flagged as an unattributed path, so unanswered "who erased these 400 messages" is at least countable. Every delete also emits one structured line carrying a salted actor hash, so a burst is visible as a per-actor counter, and a per-identity daily delete budget bounds the cleanup work one account can drive: past it the trigger still runs the correctness steps (audit, purge queue, pin, topic copy, saved-copy redaction, reactions, rev bump) and leaves the derived caches — channel denorm, shared-media sweep, storage purge — to the queued purge and the next message. A replayed delete event is claimed once (a per-message marker), so an at-least-once re-fire does not re-run the pipeline.- How far ahead you can schedule — a scheduled message can be set from one minute to exactly 30 days ahead; a later time is refused with "Cannot schedule more than 30 days ahead", the same limit an agent hits through the CLI schedule routes.
Not yet proven live end to end: the OS toast firing (main-process) and scheduled-send delivery (its Cloud Tasks queue and deliverScheduledSendTask are deployed and running). Storage upload/download is deployed and live-verified; the logic is unit/emulator-tested.
Channel invitations + per-channel roles
Live (the channel-invitations flag is flipped, and the Firestore rules are deployed) — the channelMembers subcollection rules and the postingPermission admin-write clause are in the live rules release, so role and posting-permission changes persist in production. It follows the Slack model: instant add (no accept/decline flow), system messages for every membership change, workspace admin override.
- Per-channel membership — managers can add workspace members to specific channels (not just the whole workspace) and remove them. A member can leave any channel. Every change is instant and generates a centered system message (e.g. "Alice added Bob", "Charlie left").
- The person you add is told — adding someone (from the Add Members dialog, in bulk, or by an agent through the CLI) tags them in the "Alice added Bob" message, so they get an alert in their own Omniscio inbox and a push on their phone, even if they only get notified for mentions. Added while their app was closed, the alert appears next time they open it. Joining a channel yourself, a channel you created, or re-adding someone already in it raises nothing, and someone who muted all notifications stays muted.
- Three-tier role model — each channel member has a role: manager (add/remove/promote/demote), member (read + write), or readonly (read + react, no posting). The channel creator is automatically a manager; invited members default to
member. - Posting permissions — a channel-wide
postingPermissionsetting:everyone(default) ormanagers-only, set via the "Who can post" control in the Edit channel dialog (workspace admins only). When set to managers-only, non-managers — and anyreadonlymember in any channel — see a ReadOnlyBar banner in place of the message box, matching what the server-side message-create rule enforces. - Bulk invite — add up to 50 members at once via the Add Members dialog, each with an assigned role. The dialog shows all workspace members with a search filter, checkboxes and per-member role selectors.
- Member count + panel — the channel header shows a member count button (Users icon + number) that opens a ChannelMembersPanel slide-out, listing members sorted by role (managers first) with role badges and management actions for managers. A member the roster has no record of reads as Unknown member, never as their raw account id, and a refused remove or role change always says so in a toast, even when the cause is a permission or connection problem. On desktop the panel starts below the window's title bar, so its buttons never sit on top of the minimize/maximize/close controls.
- Bulk member actions — a manager sees a checkbox on every member except themselves, plus a bar with Select all, the selected count, Change role, Remove (one confirmation for the whole selection) and Clear. Each bulk change is saved in one all-or-nothing write and posts one channel notice ("Alex removed Ada, Bo, Cy +3"), not one per person. The buttons grey out, with a tooltip saying why, when the selection holds every manager (a channel always keeps at least one) or more than 50 people. You can never select yourself — leaving a channel stays on Leave channel.
- The conversation header — name and topic, a 🔒 lock on a private channel (hover it for what it means), member count, a Topics button that opens the topic list, a + button that starts a new topic, a magnifier that opens search, and a ⋯ More actions menu. Every control is the same plain button style. The menu holds Jump to date, Notifications › (one row showing the current setting — Default, All messages, Mentions only or Muted — that opens the full choices), call, Media gallery, Channel settings and, on a note-to-self, Remind me later. Jump to date, notifications and call each have a Show … in header row in the menu (checked while that control sits in the header): check it to put the control in the header row, uncheck it to send it back — nothing is removed, the choice is remembered on this computer, and the row always orders them the same way. Phone keeps only the bell beside the menu, and a narrow panel shows icons with tooltips instead of crowding the name. Invariants: team-chat-header-action-pins-contract.md.
- Last-manager protection — the last manager cannot leave, be removed, or be demoted. The system refuses the operation and surfaces an error.
- Lazy backfill — pre-feature members (in
memberUidsbut with no subcollection doc) are treated asrole: 'member'; the role lookup creates the subcollection doc on first check. - DMs excluded — all per-channel role operations skip DMs (DMs have no roles, no subcollection).
- New Channel dialog — when creating a private channel, a member picker appears to pre-select initial members with per-member role assignment. If the channel cannot be created, a toast says so and asks you to check your connection and your workspace permissions, then try again.
For engineers: membership lives in a Firestore subcollection channels/{channelId}/channelMembers/{uid} (role + addedBy + addedAt); memberUids stays as the fast-query index. Every mutation writes the subcollection doc, memberUids and the recency denorm in one writeBatch (atomicity invariant); the divider is posted afterwards through the teamChatPostSystemMessage relay, since no client may write a system-message doc (team-chat-system-messages-contract.md). Divider copy is recomposed at render from systemMessageKey + systemMessageParams, naming the actor live from authorUid; an older doc's frozen text is recovered by matching the same template table — never bake a name into text, because the auto-join writer runs before the member directory loads. postingPermission is writable only by a workspace owner/admin (alongside visibility). Those rules are deployed and live in production (emulator tests run in CI). Invariants and deploy checklist: team-chat-channel-invitations-contract.md.
When someone leaves the workspace (their messages are hidden from members)
When a person is no longer in a workspace (an admin removed them, they left, or they deleted their account), their old messages stop showing to regular members and guests. Before this, those messages stayed in every channel labeled "Unknown member".
A name that is only slow to load is never labeled as someone who left. While the team list is still loading (a slow or dropped connection), a message shows a softly shimmering bar where the name goes and a shimmering circle for the picture, and the real name fills in when the list arrives. If the list cannot load at all, the message says Name not loaded in gray instead. The same goes for an @mention inside a message (the chip shimmers, then reads "@Name not loaded" if the list fails) and for the profile card that opens when you click a person.
- Their direct message leaves your list — a DM whose only other person has left the workspace no longer appears under Direct messages, so the list stops collecting rows that can only read "Former teammate". A conversation one live person is still in stays.
- What members see — nothing from that person in the channel, its threads, the pinned bar, the files gallery or search, on desktop and the phone web app. If one of their messages started a thread that current members replied to, a slim "Message from a former member" row stays so the replies are still reachable.
- What owners and admins see — the same messages, dimmed, with a small Hidden from members label and "Former teammate" as the author.
- Show one message to everyone — an owner or admin right-clicks the message (or uses the phone action sheet) and picks Show to everyone. Every member then sees that one message, with "Former teammate" as the author; the person's other messages stay hidden. Hide again undoes it, and admins see a Shown by an admin label on it. The phone web app shows these choices but the action itself is on desktop.
- Nothing hides while the member list is loading, and system notices, imported Slack or Discord history, and messages still sending are never hidden.
- It is a view, not a lock — the messages stay in the workspace's data; nothing is deleted or edited. Agents reading history through the CLI still see them.
- Known edges — search and the workspace-wide files gallery only know which messages were shown again in the channel you have open, so a shown message from another channel stays hidden there until you open that channel. The inbox's embedded chat preview does not hide these messages.
For engineers: the rule lives in former-members.ts (isFormerMemberMessage, isFormerMemberDm, formerMemberView, visibleToViewer); hiding switches on only after a server roster snapshot below the listen cap — the sidebar passes the listener's own truncation flag, so a cut-short roster never reads a live teammate as gone. A shown message is a doc at channels/{cid}/unhiddenMessages/{messageId} (messageId, unhiddenBy, unhiddenAt) that only an owner or admin may create or delete, in both roots; the rules ship in firebase/firestore.rules and are deployed and live, so Show to everyone works for real users in both roots (a refused write still surfaces a friendly error, and hiding works either way). Invariants: team-chat-workspace-admin-controls-contract.md.
Self-serve workspaces (create + share your own chat spaces)
Alongside your governed company workspace, you can create and join your own chat workspaces — fully self-serve, with no admin to ask and no separate account to make. They live beside the company workspace in a switcher rail down the chat sidebar: the company workspace first (a building glyph), your workspaces after (a tile with its initial each), and only the active one carries the accent. A "+" right under the last workspace tile opens a small chooser to Create a new workspace or Join one by code. The tile list sizes to its tiles, so the "+" follows the last one; when there are more workspaces than fit, the list scrolls and the "+" stays visible just below it. It ships dark like the rest of Team Chat (same Settings → Lab toggle / AMC_SHOW_TEAM_CHAT=1); the backend is deployed — all fourteen workspace callables and the daily reaper are live in production.
- Create a workspace — name it and you become its owner; a
#generalchannel is created and you're switched straight into it. (There's a small per-account cap on how many you can own.) - Share an invite code — as an owner/admin, the sidebar's "Invite people" affordance opens a share-code dialog that mints a 10-character code on demand and copies it (owners/admins can also revoke). There is no list of existing codes: clients can't read the code docs, so the freshly-minted code is the only time one is ever shown. Two limits bound minting: a workspace can hold 25 live codes at once, and one person can mint 20 codes per day (resetting at midnight UTC) — so a workspace at its code limit uses the dialog's Turn off all invite codes (it asks first, and every code stops working right away) before minting again, and hitting the daily limit waits until tomorrow. For your company workspace the same affordance opens the by-email invite instead.
- Join by code — paste a code into the Join dialog and you're added as a member and switched in. A code that cannot be one (not 10 letters or numbers) is refused by saying what a code looks like — the difference between "you mistyped it" and "that code is gone" — while a valid-shaped code that was revoked / expired / used-up keeps its own plain refusal. The dialog stays open either way (never a false success).
- Refusals are readable and language-aware — a self-serve workspace failure (create, join, leave, restore, delete) reaches you as a plain sentence in your own language, and "that workspace is not available to you" is distinguishable from "you do not have permission" without reading the prose. Every refusal carries a machine code from the same vocabulary the rest of Team Chat uses.
- Leave a workspace — right-click (or long-press on mobile) a self-serve workspace tile and choose "Leave workspace", behind a confirm that names what you lose (access to all its channels and messages; you can rejoin later with a code). Shown only to non-owner members — an owner can't leave (it would orphan the workspace), and the server refuses it too. Leaving the workspace you're viewing switches you away first so nothing flashes an error.
- A mute stays muted when someone leaves and comes back — when an owner or admin mutes a member, that is kept as a per-person record on the workspace (
chatWorkspaces/{wsId}/muteRecords/{uid}), not just an entry in the live roster. Leaving drops the person from the roster's muted list, but a rejoin by invite code re-applies the mute, so a muted member cannot lift their own mute by leaving and rejoining. Only an owner or admin unmutes; removing someone is the stronger action and clears the record, so an unbanned member who rejoins comes back unmuted. - Delete a workspace (with a 30-day undo) — as the owner, right-click (or long-press) your tile and choose "Delete workspace", the owner's counterpart to Leave. It's a soft delete behind a confirm: the workspace and all its channels, messages and members are hidden immediately but kept for 30 days, and a toast offers an Undo that restores it. After the window a scheduled reaper permanently deletes the whole tree. Deleting the workspace you're viewing switches you away first.
- Leaving or deleting never sets off a burst of warnings — the moment you confirm, the workspace comes off your workspace list, before the server acts, and goes back on the list (and reopens if you were viewing it) when the delete or leave fails. Every background watcher for a workspace — its unread counts, its member list, the session starter bot's watch — follows that list, so none reads the workspace before the server hides it. The behavior lives in
withSwitchAwayAndRestorein use-chat-workspaces.ts; the open workspace falls back to the company workspace, else the next workspace on the list, else none. - Per-workspace identity — your display name in a workspace defaults from your account and is never blank. You can set a custom display name per workspace via the "Edit workspace name" button in your profile popover — it shows on all messages, past and future (resolved at render time), survives global profile edits, and can be reset to your account name with one click. A name-change system message is optionally announced in #general (default on, togglable in the dialog). Works in both company org and self-serve workspaces. Per-workspace avatar editing is a follow-on.
- Profile photos only load from where they should — a workspace shows a teammate's picture only when it comes from Omniscio's own storage or a known sign-in provider's photo service (Google, GitHub). Any other web address is treated as no picture at all, in the member list, on message rows and on a connection request. That stops someone pointing their photo at a tracking link which would learn who looked at them, and when.
- Workspace icon color — right-click (or long-press) a self-serve tile to pick a gradient color from an 8-color palette (green, blue, red, purple, orange, pink, slate, teal), so you can tell your workspaces apart at a glance. The color is stored locally on each device (server persistence is a follow-on) and shows immediately, including on the active tile with an accent outline ring. See workspace-icon-color.md.
- Workspace notification badge — each workspace icon in the desktop rail and the mobile switcher pill shows a notification count badge (Discord-style) when that workspace has something waiting on you. The count is that workspace's share of the all-workspaces Needs you list (the same rows the rail's inbox button opens), so it changes as you read, answer or dismiss items, and caps at 99+.
- What works in a self-serve workspace (v1): the core chat surface — public/private channels, DMs, live send / receive, paged history, @mentions, real names & avatars, edit/delete-your-own, presence/availability (user-level, tied to your company org, so your Online/Away/DND/OOO/Offline status persists across all workspaces), plus reactions, pins, threads, saved items, scheduled send, unread badges and desktop notifications — all identical in both roots. The remaining company-only extras are hidden here (not shown as dead buttons): typing indicators and per-channel notification prefs.
Cross-org connections (1:1 DMs across organizations)
Connect with any Omniscio user by email — same workspace or another organization — to start a 1:1 DM that forms only on the recipient's approval. Connections live in a separate top-level connections/{connId} collection (not under any org), with anti-enumeration privacy, server-only email lookups and a per-user opt-out toggle. Lifecycle, CLI routes, privacy controls and components: cross-org-connections.md; inbox approval flow: connection-request-inbox.md; contract: team-chat-connections-contract.md.
More of what you can do
- Self-profile & manual availability — your avatar and presence dot sit at the left of the one-row sidebar header (on a phone, before the workspace switcher). Click it for a popover with an availability selector (Online / Away / Do Not Disturb / Out of Office / Offline) that manually overrides automatic presence — useful because Omniscio runs 24/7 and you may want to pin "away" or "DND" while the app is focused. Clearing the override returns to automatic detection (online when focused, away when blurred, offline on quit). Manual "online" degrades to "away" if the heartbeat goes stale (past
PRESENCE_STALE_MS, 90 s in presence-dot.ts; the publisher heartbeats everyPRESENCE_HEARTBEAT_MS, 30 s). Away/DND/OOO/Offline persist across app restarts. The popover also has an Edit Profile button (available to all users) that opens the full profile editor (display name + avatar upload).manualAvailabilityis written field-scoped so it can't clobber the heartbeat; when set it wins over automatic presence. - Unread badges — each channel shows how many messages you haven't read; opening a channel clears its badge (marked read instantly, then reconciled with the server). You can also right-click a DM and choose "Mark as read" to clear its badge without opening it, with undo that restores the previous read marker and count. A channel also lights a badge when it has a durable inbox notice waiting (even if server read-state was reset on another device). The Team Chat hub badge counts only notices you can act on — one for a channel you can no longer open (you left it, were removed, or it was deleted) isn't counted and is auto-cleared from the Inbox. See team-chat-v2-features-contract.md
the-unread-badge-counts-inbox-rows/the-hub-badge-counts-only-surfaceable-notices. - Pinned messages — clicking a pinned row opens that message; a pin older than the loaded history offers Load older messages in place.
- A conversation that disappears — if the channel or DM you have open is deleted or you lose access, the panel says so and offers your unsent draft to copy, instead of loading forever.
- Alone in a new workspace — an owner or admin who is the only member sees Invite people in the empty channel.
- History — the live view shows the latest messages; "Load older messages" pages further back on demand. Scrolling into the top zone starts the next page (once per approach), so reading back is a scroll rather than a stop-click-wait at the button, which stays as the fallback.
- Instant feedback on your actions — "Mark as read" closes its menu the moment you click it (the Firestore write rides behind and the reads listener reconciles); the channel sidebar's first paint holds until BOTH queries have answered, so DMs are never spliced in under your pointer; and in an inline message edit the Save button shows its busy state while the write is in flight instead of looking dead for a round-trip.
- Jump to date — a calendar icon in the conversation header when it is pinned there, and always a row in the header's ⋯ More actions menu (either way, Ctrl+Shift+D) opens a date picker; selecting a date detaches the live listener and loads messages from that day in replace mode (the live tail is paused, not destroyed). In jumped mode you paginate bidirectionally (older + newer), a sticky "Viewing messages from [date]" banner shows where you are, and "Return to latest" re-attaches the live listener and scrolls to the bottom. Selecting today shortcuts straight to return-to-live. On mobile the picker renders in a modal instead of a portal popover. Search results also gain a "Jump to context" action that opens the surrounding conversation the same way. A jump lands on the first message of the window it loads, and a date after the last message shows the page just before it, with the banner naming the day actually shown.
- Quick switcher (Cmd/Ctrl+K) — while Team Chat is open, press Cmd/Ctrl+K to fuzzy-search your channels and DMs by name and jump to one: type to filter, arrow-keys to move (Home/End jump to the first/last row while the search box is empty), Enter to open, Esc to close. It only takes over Cmd/Ctrl+K while Team Chat is the active surface; everywhere else that key still opens global search. Desktop only. If your channel list failed to load, the switcher says "Couldn't load your channels" with a Try again button instead of claiming you have none, and a search with no hits echoes what you typed (
No channels match "...").
Related
This is one of five pages describing Team Chat. The parent page covers the feature and the first version; the companions are part 2 on the second wave of features and workspaces, part 3 on the phone client, notifications and the smaller surfaces, part 4 on the limits and the engineering, and part 5 on bringing people into a workspace.
Also worth having open: cross-org-connections.md for connecting with someone in another organization, and session-starter-bot.md for the bot that starts a Claude session from a channel message.
Last verified 2026-11-15