Org policies

Everything that captures or sends content starts off

Recording, saving audio or video, transcription, captions, saved transcripts, personal transcription, quest suggestions and AI suggestions all default to off. This holds in the database schema (every flag is default false, personal transcription is off), and in the code fallback orgPolicy() for an org that has never saved a policy. Nothing is captured until an Org owner or Org admin allows it, and then a Team lead or a person still has to start it. Nothing in the policy forms is pre-ticked.

• Off every capture and AI flag, until allowed
• Org owner / Org admin the only roles that change a policy
• 4 dashboard pages: Recording policy, Transcription, Sign-in & channels, Suggestions

Org policies are set per organization in the web dashboard. Only an Org owner or Org admin can change them; any member can read the recording policy. A member who opens a page they cannot change sees the server’s refusal.

Recording policy

Dashboard page: Recording policy (sidebar: Recording Policy).

Capture switches

Setting	What it does	Default
Allow recording	Master switch. Off turns every capture mode off for the whole org, and disables the boxes below it.	Off
Save audio	Per-track audio and a mixdown for playback.	Off
Save video	Camera and screen tracks. Needs Save audio.	Off
Transcribe	Live captions and a saved transcript.	Off
Suggest quests from conversation	While a call is transcribed, phrases like “we should make a quest for” offer a one-click quest. Needs Transcribe. Rules are per team.	Off
Let AI fill in suggestion details	Sends the line that triggered a suggestion, and the few before it, to the AI model. Off means rules only, and nothing goes to a model. Needs the setting above.	Off
Suggest actions from team chat	Quest suggestions from messages.	Off

Retention, consent and models

Setting	What it does	Default
Keep media for (days)	0 to 365. 0 deletes audio and video when the meeting ends. Transcripts are kept.	30
Live captions (1st pass)	Live, word by word (runs on this server, English only) or Off (captions arrive a sentence at a time, about a second after each pause). Blank follows the deployment default.	Deployment default
Transcript (2nd pass)	The model that replaces the live words at each pause and produces the saved transcript. Blank follows the deployment default.	Deployment default

Dependent boxes follow: video needs audio, suggestions need transcription, and everything except the chat-suggestion box needs “Allow recording”.

The policy has more fields on the server than the dashboard form shows: captions, captions on by default, saved transcript, languages, personal transcription (off by default), the assistant name, and voice-command settings. All of the capture flags among them default off too. The form does not edit them.

Plan features

Starting audio or video capture also requires the recording feature on the org’s plan, and transcription requires transcripts. Without them the server answers 402 upgrade_required even when the policy allows it.

Keep a team’s call recorded

Further down the same page, each team has a Keep a team’s call recorded box. Ticking it shows the parts to capture (live captions, quest suggestions, keep transcript, save audio, save video). Nothing is on until you tick a part, and unticking the last part turns the setting off. The call then records while it has people in it and stops when it empties. Everyone is still told what is captured. Guests, and anyone who has not accepted the TeamHUD terms, are also asked to consent, and anyone who declines is removed.

A Team lead, Org owner or Org admin can change it. The server rejects any part the org policy does not allow.

Transcription provider

Dashboard page: Transcription (sidebar: Transcription).

• In use now shows the provider actually in force and whether it comes from this org’s settings or the deployment default. If none is configured, recording works but transcription does not.
• Providers: DeepInfra (zero retention: audio is transcribed and not stored by the provider) and OpenAI-compatible (your own endpoint: audio goes to the server you name).
• Fields are base URL (OpenAI-compatible only), API key, model and language. The key is write-only: the server returns it masked, and an empty box on save keeps the stored key.
• Test connection transcribes a two-second clip, so a wrong key, URL or model shows up while you are looking at the form.

The deployment default comes from TRANSCRIPTION_PROVIDER in Configuration. Unset, each org must configure its own.

Sign-in and channels

Dashboard page: Sign-in & channels (sidebar: Sign-in & Channels). It can only narrow what the server offers.

• Where chat lives: TeamHUD channels, or Slack channels. Slack is available only if the server lists it in CHANNEL_PROVIDERS and offers slack_oidc sign-in. Everyone then signs in with Slack. Teams channels are not available yet; an org already on them sees that and can switch away.
• How people sign in: untick a method to stop accepting it for this org. People who can only sign in a way you turned off cannot join by invitation or workspace auto-join. All ticked means no restriction, at least one method must stay on, and a native-channels org keeps its product’s sign-in. A method the server stops offering stops counting for the org as well; if none is left, nobody new joins until an Org owner or Org admin picks again.
• Slack workspace and Microsoft tenant: appears when the org was created by someone signing in from them. Switch on auto-join and everyone who signs in from the same workspace or tenant joins as a member, with no invitation.

Suggestion rules

Dashboard page: Suggestions, per team. Only a System admin’s sidebar has a link to it; anyone else opens /account/suggestions in the dashboard. Members read and test; Team leads edit.

While a call is transcribed, each sentence is checked against these rules, instantly and with no AI. A match offers everyone in the call a one-click quest, and later sentences can add an owner or due date. The page has a wake phrase, built-in rules you can switch off, the team’s own rules (name, what it does, a phrase template, wake-phrase-only, and whether AI may fill in details), and a Try a sentence box. The AI option applies to quest-creating rules only and is further gated by the org’s “Let AI fill in suggestion details” setting above. Suggestions run only if the org policy has them on.