Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.tuco.ai/llms.txt

Use this file to discover all available pages before exploring further.

The Knowledge Base is now part of AI SDR — find it at AI SDR → Knowledge tab.

​
What the Knowledge Base Does

The Knowledge Base (KB) is Tuco’s AI reply engine. When a lead replies to your message, the KB can:
  • Draft a reply using your business context, FAQs, and conversation history
  • Send automatically (Autopilot) or wait for approval (Copilot)
  • Route to the right person via Telegram and email notifications

​
Modes

​
Manual

No AI involvement. Replies are forwarded to the contact owner via Telegram and email. You reply manually from Unibox.

​
Copilot

AI drafts a reply and sends it to the contact owner for approval:
  • Telegram notification with Approve / Reject buttons
  • Email with the draft and approval link
  • The draft appears in Unibox for editing before sending

​
Autopilot

AI sends the reply automatically without human approval. Use for high-volume, low-risk conversations.

​
How It Works

  1. Lead replies → message is debounced (60 seconds of silence)
  2. All messages in the batch are combined
  3. KB generates a contextual reply using your uploaded knowledge
  4. Copilot: draft sent to owner for approval via Telegram + email
  5. Autopilot: reply sent immediately, notification sent to owner

​
Notifications

The contact owner receives notifications via:
  • Telegram (personal channel if linked, workspace channel as fallback)
  • Email (contact owner’s email, with support@tuco.ai CC’d)
If the contact owner can’t be found, notifications go to the earliest workspace member.

​
Daily Limits

Set a max auto-replies per day limit in Integrations → Telegram → Auto-Reply Limits (default: 50, max: 500). Once reached, new replies wait until the next day.

​
Multi-Bubble Replies

The AI Responder sends replies as multiple short iMessage bubbles instead of one long message. This looks more natural:
  • Replies split into 2-4 short messages
  • Typing indicator (”…”) appears between each bubble
  • 1-3 second delay between bubbles (based on message length)
  • All gaps and throttles bypassed for AI replies — leads don’t wait 30 seconds between bubbles

​
Reactions

The AI Responder detects iMessage reactions (tapbacks) and does NOT treat them as messages:
  • Heart, thumbs up/down, laugh, emphasize, question, and custom emoji reactions are recognized
  • No AI call is wasted on reactions
  • No notifications sent for reactions

​
Voice Notes

When a lead sends a voice note:
  • Audio file is downloaded and stored
  • Displayed as an audio player in Unibox
  • Optional: transcribe via Whisper AI (enable in workspace settings)
  • Transcription passed to AI Responder as the inbound message

​
Goal Bot Integration

When a lead is enrolled in a Goal Bot, the AI Responder uses the goal bot’s stage-specific prompt instead of generic knowledge base search. See Goal Bots for details.