Samurai AppDocs

Using the WebUI — サム丸, the AI participant that answers by voice

Any screen not listed here does not exist. Headings quote the screen's own wording.

Part of the meetings guide. Other topics: conversations, the input box, and sign-in: references/webui.md; documents, files, tables, memory, channels, and apps: references/webui-workspace.md; editing a file in place (「編集」): references/webui-file-edit.md; SIROE pages, blocks, and boards: references/webui-siroe.md; settings, administration, models, and triggers: references/webui-settings.md; external service connections: references/webui-connections.md; /office: references/webui-office.md; /meetings: references/webui-meetings.md; /calendar: references/webui-calendar.md.

「サム丸招待」(声で答える AI の参加者)

「サム丸招待」 is the first button in the row above the call (in the app's top bar on wide screens), before 「メッセージ」, 「資料」, and 「AI」. Members only — hidden for guests, when 「AI アクセス」 is OFF for the meeting, and when captions and AI cannot be used in the call (usage limit, or not prepared). It joins the call as one more tile showing the samurai character, and the button turns into 「サム丸退出」. Only one サム丸 per meeting: if someone else has called it, you see 「ほかの人がサム丸を呼んでいます」; if this environment is not set up for it, 「サム丸はまだ使えません」 or 「この環境ではサム丸を呼べません」.

It talks with you directly by voice and answers right away: it starts as soon as you are done speaking, with no pause to wait for the rest of a sentence. Small talk and questions about the meeting are answered on the spot; when a question needs your organization's records or features, サム丸 says it will look it up and passes it to the meeting chat in 「AI チャット」, then says the answer. If that takes more than about 15 seconds, サム丸 says the result will appear in the chat, and the answer shows up there. If the connection drops for a moment (for example during an app update), サム丸 reconnects once by itself and keeps going, but it no longer remembers what was said just before. If サム丸 cannot talk (the voice service is not set up, the connection does not come back, a usage limit, or it has talked as much as one meeting allows), it leaves the call and shows the reason, such as 「この環境ではサム丸が話せません」 or 「サム丸との接続が切れました」; call it again to bring it back.

It answers only when called by name (e.g. 「サム丸、議題は?」) or through 「サム丸に話す」, aloud to everyone. For 20 seconds from that person's last line it took up (counted from when the line is transcribed, not from the end of its answer), the same person can follow up without the name. It stops speaking when someone says just a stop word — 「ストップ」「止まって」「黙って」「黙れ」「やめて」「わかった」「もういい」「うるさい」, with or without the name — or, while 「サム丸に話す」 is ON, when the caller talks over it. Otherwise talking over it, background noise, or its own voice coming back through a speaker does not stop it. Its face shows its state: blinks while waiting, nods when called, 「…」 while thinking, mouth moves while speaking (the wave shows the same).

If the name is missed, the caller can press 「サム丸に話す」 (only on their screen; it has no keyboard shortcut — Space no longer toggles it and now turns the mic on and off, as in Google Meet), and talk without the name: while it is ON, サム丸 hears the caller's microphone directly, answers each thing they say, and keeps listening after answering. Press it again to stop.

It hears being called by name through the captions' transcript (the same lines as 「字幕」). Its own voice is not recorded or transcribed.

Where the answers go: only the questions passed on to look things up are written in the meeting chat panel (「AI チャット」), as an ordinary chat of the person who called it; each one ends with 「by 」 so you can see who asked. Small talk is not written there. Answers are 1–3 short spoken sentences.

With a guest in the meeting, it uses only what was said in this meeting and general knowledge — no company files, memory, tools, or the chosen bot's instructions — and when a question needs company information it says it cannot use it while a guest is present. Once the guest has left, it answers normally again.

Operations that need approval cannot be done by voice — use 「AI チャット」 instead.

Settings (the ▾ joined to 「サム丸招待」, like 「AI ▾」; there even before calling it; pointing at it shows the current 「性能」 and bot):

  • 「ボット」 (a drop-down): 「使わない」 (default) or one of your enabled bots. The choice lasts only for this meeting — every other meeting (and a reload) starts again from 「使わない」. With a bot chosen, サム丸 answers as that bot — its instructions also shape small talk (except while a guest from outside is in the meeting), and its skills, knowledge, features, and memory setting apply to the questions passed to the chat, and 「使う機能」 and 「記憶」 below are replaced by 「機能と記憶は、ボット「…」の設定で動きます」. 「性能」 still applies. The voice exchange goes to a meeting chat tied to that bot, so switching bots starts a new chat in the panel. If the bot is deleted or disabled during the meeting, it shows 「使わない」 and サム丸 answers without a bot.
  • 「性能」 (side-by-side buttons; a note under them describes the chosen one): for the questions passed to the chat — 「速さ優先」 (default; answers fastest), 「バランス」, or 「じっくり」 (thinks deeper and takes longer).
  • 「使う機能」: a row of chips, one per group of features (the same groups as a bot's settings) plus 「つないだサービス」; press a chip to turn it on (it shows a check) or off. By default only 「探す」 is on. 「すべてオフ」 next to the heading makes it answer only from the meeting and general knowledge.
  • 「記憶」 (a switch): whether the caller's memory is used.
  • 「答えを文字起こしに残す」 (a switch): off by default. When on, each sentence サム丸 says is added to the captions, the transcript, and the minutes, under the speaker 「サム丸」. Only sentences it actually said are added (a sentence cut off by an interruption is left out); turning it on later does not add earlier answers.
  • 「姿」 (side-by-side buttons, at the bottom next to 「声」): 「キャラ」 (default) or 「波」.
  • 「声」 (a drop-down): 「サム丸の声」 (default, an energetic boy's voice) or one of 29 other voices, each listed by name with its gender and tone (e.g. 「Charon 男性・知的」, 「Kore 女性・きっぱり」).

Except 「ボット」, they are remembered on that device. 「ボット」 and 「声」 apply from the next thing サム丸 says (it reconnects; anything it was saying is cut off), 「性能」, 「使う機能」, and 「記憶」 from the next question passed to the chat, 「答えを文字起こしに残す」 from the next sentence; 「姿」 changes at once. Fewer features and 「速さ優先」 answer looked-up questions sooner. They affect voice only, not what is typed in 「AI チャット」. After サム丸 answers, the model and reasoning shown in that chat's input box switch to the ones 「性能」 used; changing them there affects only what you type.

Dismissing it: press 「サム丸退出」. It also leaves when the caller's tab is closed or the caller leaves the call; a tab behind other windows can make its voice stutter. Anyone else can then call it again.