How does it work?

Send a message or a voice note on Telegram right after the meeting.

n8n transcribes ( if it's a voice note ) and sends the text to GPT.

GPT generates a structured and professional meeting minutes report.

The report is automatically stored in Airtable.

Your team is instantly notified in Slack.

A formal email is sent via Gmail to the right recipients.

👉 Works for all types of meetings : client calls, team syncs, project updates… whether you type a message or send a quick voice memo.

---

✅ Requirements

Before running this workflow, you'll need:

• A Telegram account with a bot configured (to send your messages/voice notes)
• An OpenAI API key (for GPT and voice transcription)
• An Airtable account with a base containing these fields:
  • E-mail
  • Subject
  • Report
• A Slack account with the target channel for notifications
• A Gmail account connected to n8n (OAuth2)

🔧 Step-by-step setup

Step 1 – Telegram Trigger

• Node : Telegram Trigger
• Updates : message
  👉 Captures every message or voice note sent to the bot.

---

Step 2 – Detect text or voice

• Node : Code (“Message or Voice ?”)
• Expected output :
  • { type: "text", content } if message
  • { type: "voice", file_id } if voice note
    👉 Routes the workflow based on the input type.

---

Step 3 – IF Condition

• Condition : {{$json.type}} == voice
  👉 Directs to the transcription branch if it's a voice note.

---

Voice Branch 🎤

Step 4 – Download the voice file

• Node : Telegram → Voice note
• Parameter : fileId = {{$json.file_id}}

Step 5 – Wait

• Node : Wait (2–3s)
  👉 Lets Telegram prepare the file.

Step 6 – Voice note download

• Node : Telegram (file download)
  👉 Retrieves the audio file.

Step 7 – Transcribe to text

• Node : OpenAI (Transcription)
• Resource : audio
• Operation : transcribe
  👉 Converts the voice note into plain text.

Step 8 – Short wait

👉 Ensures continuity before sending to GPT.

---

Text Branch ✍️

Step 9 – Normalize

• Node : Code (“Content”)
• Return : { text: $json.content }
  👉 Standardizes the text as if it were already a transcription.

---

Step 10 – Detect email

• Node : Code (“Domain or Email detection”)
  👉 Extracts the target email (or builds a fallback contact@gmail.com ).

---

Step 11 – Generate meeting minutes

• Node : Agent (“Generate Meeting Message”)
• Prompt : specialized for “meeting minutes”
• Model : GPT-4.1-mini
• Output : { email, subject, body }
  👉 GPT creates a clean and structured meeting report.

---

Step 12 – Enforce clean JSON

• Node : Output Parser Structured
• JSON Example :
  json {"email": " address@gmail.com ","subject":"Subject","body":Email"}

👉 Ensures the output is always valid JSON.

Step 13 – Cleanup / Airtable mapping

• Node : Code
• Return : { Email, subject, Report }
  👉 Prepares the correct fields aligned with your Airtable table.

---

Step 14 – Store in Airtable

• Node : Airtable (Create Record)
• Mapping :
  • Email = {{$json.Email}}
  • subject = {{$json.subject}}
  • Report = {{$json.Report}}
    👉 Archives each meeting report in your Airtable base.

---

Step 15 – Notify in Slack

• Node : Slack (Send Message)
• Channel : your team channel
• Message : {{$json.fields.subject}} {{$json.fields.Report}}

Step 16 – Send the email

• Node : Gmail (Send Email)
• sendTo : {{$('Create a record').item.json.fields.Email}}
• subject : {{$('Create a record').item.json.fields.subject}}
• message : {{$('Create a record').item.json.fields.Report}}