Classifying Incoming Support Requests
Overview
The first layer of the AI Agent Support Analyzer System is built to detect, classify, and route messages from multiple sources — Gmail, Telegram, voice, image, PDF, or plain text.
This is where the system determines:
- Who the message is from
- What kind of content it contains
- How it should be processed next
- Whether it needs to be escalated, auto-resolved, or summarized
Supported Input Sources
| Source |
Channel |
Content Type Handled |
| Gmail |
Email |
Text, attachments, inline images |
| Telegram |
Chatbot |
Voice, text, image |
| Webhooks |
(Optional) |
API-based support inputs |
Step-by-Step: Message Detection Setup
1. Trigger from Gmail (n8n Gmail Node)
- Monitors the inbox for:
- New unread messages
- Specific labels (e.g., “support”, “escalation”, “orders”)
- Extract:
sender emailsubjectbodythreadId (for reply chaining)
2. Trigger from Telegram Bot
- Captures:
- Voice notes (converted to text via Whisper or Gemini Audio API)
- Image uploads (optional: OCR for image-to-text)
- Text messages
- Automatically tags each incoming message with:
source_type: "telegram_voice", "telegram_image", "telegram_text"