Synapse Digital is a multilingual AI business assistant and intelligent automation workflow built with n8n for an electronics repair and sales business.
The assistant operates through Telegram, supports both text and voice interactions, and can handle customer appointments, repair requests, complaints, appointment modifications and cancellations.
Rather than acting as a simple chatbot, the AI Agent is connected to real business tools and follows predefined business rules before performing important operations.
The workflow combines:
AI Agents + Business Rules + Databases + Calendar Automation + Voice AI + Telegram
The same bot, same logic, three languages — it detects the customer's language automatically and never mixes languages mid-sentence:
| English | Français | العربية |
|---|---|---|
 |
 |
 |
The assistant supports three languages:
- 🇸🇦 Arabic
- 🇫🇷 French
- 🇬🇧 English
The customer can communicate naturally in any of the supported languages.
The assistant detects the conversation language and responds accordingly.
The same business logic is applied regardless of the selected language.
The assistant can understand customer requests and guide the conversation step by step.
It does not require the customer to know a specific command structure.
For example, a customer can simply ask to:
- Book a repair appointment
- Report a problem
- Check an existing booking
- Modify an appointment
- Cancel an appointment
- Ask about a repair
- Ask about an estimated repair price
The AI Agent determines the appropriate workflow and uses the required tools.
The assistant supports both:
Telegram message
↓
AI Agent
↓
Telegram text response
Telegram voice message
↓
Speech-to-Text
↓
AI Agent
↓
Text-to-Speech
↓
Telegram audio response
Voice messages are transcribed before being processed by the AI Agent.
The generated response can then be converted back into speech and sent to the customer.
This allows the system to operate as a voice-enabled virtual receptionist.
The assistant can manage the complete appointment booking process.
It collects the required information step by step, including:
- Full name
- Phone number
- National ID / identification number
- Device brand and model
- Device problem
- Preferred date
- Preferred time
The assistant does not immediately create the appointment.
Instead, it validates the information and checks the required business conditions first.
The assistant uses explicit validation rules defined inside the AI Agent system prompt.
For example, the current configuration can require:
- A specific number of digits
- Specific allowed starting digits
- A specific phone-number format
- A specific national-ID format
The current example uses an 8-digit format with configured prefixes.
However, these are not hard-coded universal rules.
The user deploying the workflow can modify the corresponding validation instructions directly inside the AI Agent prompt according to:
- Country
- Local phone-number format
- National identification system
- Business requirements
Current configuration:
8 digits
Specific allowed prefixes
Can be changed to:
9 digits
10 digits
Different prefixes
Different validation rules
No change to the overall workflow architecture is required.
The assistant does not freely invent repair prices.
It follows a predefined accepted repair catalog contained in the AI Agent instructions.
The catalog can contain supported:
- Smartphones
- iPhones
- Tablets
- Laptops
- MacBooks
- Desktop computers
- GPUs
- Monitors
- PlayStation 5
- Controllers
- Other supported devices
For a recognized device and problem, the assistant provides the corresponding configured price or price range.
If the requested repair is not part of the accepted repair list, the assistant does not pretend that the repair is supported.
For complaint processing, the assistant can verify the customer's product against a MongoDB database.
The workflow searches the products collection using the product information provided by the customer.
Conceptually:
Customer provides product
↓
MongoDB search
↓
Product found?
/ \
No Yes
↓ ↓
Ask customer Continue
to verify complaint
This prevents the complaint workflow from automatically accepting a product that cannot be found in the business records.
If the product cannot be found, the assistant asks the customer to verify the information or provide the appropriate proof of purchase.
The assistant can also manage customer complaints.
The complaint flow collects information such as:
- Customer identity
- Phone number
- National ID
- Product
- Problem type
- Detailed description
Before proceeding, the product is checked against MongoDB.
Once the required information has been collected and confirmed by the customer, the complaint can be registered.
The assistant generates a private complaint reference such as:
REC-123456
The customer is instructed to keep the reference and present it when visiting the business, together with the required proof of purchase.
The assistant can also apologize for the inconvenience as part of the customer interaction.
After a successful appointment confirmation, the workflow generates a private booking reference.
Example:
SYNP12345
The customer is instructed to keep this reference.
When visiting the business, the customer may be required to provide:
- Booking reference
- Identification document
The booking reference is also stored in the Google Calendar event description so that the appointment can later be identified.
One of the important characteristics of the system is that the AI Agent does not immediately execute a booking after collecting the information.
The assistant first summarizes the customer's information.
For example:
Name
Phone number
National ID
Device
Problem
Estimated price
Date
Time
The customer must confirm the information.
Only after confirmation can the workflow create the appointment.
This reduces the risk of creating appointments from incomplete or misunderstood information.
Appointment scheduling is controlled by explicit business rules and Google Calendar checks.
Before creating or modifying an appointment, the workflow checks the required conditions.
Sunday is configured as the weekly closing day.
Appointments are therefore rejected for Sunday.
The workflow checks whether the requested date corresponds to a holiday.
The current configuration includes:
- A configurable holiday list
- Google Calendar holiday verification
The Google Calendar holiday source can be changed according to the country where the workflow is deployed.
The current example uses:
| Day | Opening Hours |
|---|---|
| Monday | 08:00 – 18:00 |
| Tuesday | 08:00 – 18:00 |
| Wednesday | 08:00 – 18:00 |
| Thursday | 08:00 – 18:00 |
| Friday | 08:00 – 16:30 |
| Saturday | 08:00 – 18:00 |
| Sunday | Closed |
These rules are defined in the AI Agent instructions and can be changed according to the business.
Before creating an appointment, the workflow checks Google Calendar.
If the requested time slot is already occupied:
Requested slot
↓
Calendar availability check
↓
Already occupied?
↓
Yes → Reject requested slot
↓
Suggest available alternatives
The assistant can then provide available alternatives within the configured working hours.
Existing appointments can be modified.
The workflow first searches Google Calendar for the existing booking.
The booking can be identified using information stored inside the event, including:
- Booking reference
- National ID
Before moving the appointment to a new date/time, the workflow checks:
- The existing booking
- Holiday restrictions
- Business hours
- New slot availability
- Customer confirmation
The original booking reference remains associated with the appointment.
Customers can also request cancellation.
The workflow verifies the existing appointment before deleting it.
Cancellation requires explicit confirmation from the customer.
The workflow therefore avoids immediately deleting an appointment based on an ambiguous request.
The AI Agent is connected to several specialized tools.
Used for:
- Creating appointments
- Reading appointments
- Updating appointments
- Cancelling appointments
- Checking availability
- Checking holidays
Used for:
- Product verification
- Searching business product records
Used for:
- Storing complaint information
- Creating structured records for future reporting and analytics
Used for:
- Speech-to-text
- Text-to-speech
Used as the customer-facing interface.
The workflow supports configurable LLM connections through providers such as:
- OpenRouter
- Qwen Cloud
The model can be changed without redesigning the entire workflow.
The workflow contains separate processing paths for text and voice interactions while using the same overall business logic and tools.
TELEGRAM
│
┌──────────┴──────────┐
│ │
TEXT VOICE
│ │
│ Speech-to-Text
│ │
└──────────┬──────────┘
↓
AI AGENT
│
┌─────────────────┼─────────────────┐
│ │ │
↓ ↓ ↓
Google Calendar MongoDB PostgreSQL
│ │ │
│ │ │
Scheduling Product check Complaints
│ │ │
└─────────────────┼─────────────────┘
↓
Business Rules
│
Customer Confirmation
│
┌───────┴───────┐
│ │
TEXT VOICE
│ │
│ Text-to-Speech
│ │
└───────┬───────┘
↓
TELEGRAM
Confirmed complaints can be stored in PostgreSQL using the reclamations table.
This creates structured business data that can later be used for:
- Reporting
- Business intelligence
- Complaint statistics
- Historical analysis
- Customer service analytics
- Dashboards
The workflow therefore provides a foundation for extending the system beyond conversational automation into business intelligence and analytics.
The workflow uses n8n memory nodes to maintain conversation context.
This allows the assistant to remember information already provided by the customer during the current conversation instead of repeatedly asking for the same information.
For example:
Customer: I want to repair my iPhone.
AI: What is your name?
Customer: Ahmed.
AI: What is your phone number?
Customer: 12345678.
AI: What is the problem with your iPhone?
The assistant can maintain the context of the conversation while collecting the required information.
The AI Agent is not intended to operate as an unrestricted general-purpose chatbot.
Its instructions define the role of the assistant and the business operations it is allowed to perform.
The workflow can therefore be configured to:
- Stay within the business scope
- Follow predefined validation rules
- Use specific tools for specific operations
- Require confirmation before critical actions
- Reject unsupported requests
- Avoid inventing repair prices
- Verify products before processing complaints
- Verify calendar availability before creating appointments
This creates a controlled AI Agent + deterministic business workflow architecture.
| Layer | Technology |
|---|---|
| Automation | n8n |
| AI Agent | n8n AI Agent |
| Customer Interface | Telegram |
| Speech-to-Text | ElevenLabs |
| Text-to-Speech | ElevenLabs |
| LLM | OpenRouter / Qwen Cloud |
| Scheduling | Google Calendar |
| Product Database | MongoDB |
| Complaint Database | PostgreSQL |
| Conversation Memory | n8n Memory |
synapse-digital-bot/
│
├── README.md
├── LICENSE
│
├── screenshots/
│ ├── welcome_en.png
│ ├── welcome_fr.png
│ ├── welcome_ar.png
│ └── workflow_overview.png
│
└── workflow/
└── Synapse_Digital_Assistant.json
Install or run your own n8n instance.
The workflow is designed to be imported into n8n.
In n8n:
Workflows → Import from File
Then select:
workflow/Synapse_Digital_Assistant.json
You must connect your own credentials for the services used by the workflow.
Required integrations may include:
- Telegram Bot API
- ElevenLabs
- Google Calendar OAuth2
- MongoDB
- PostgreSQL
- OpenRouter
- Qwen Cloud
The repository does not provide production credentials.
After importing the workflow, configure the credentials and business-specific values.
Connect your Telegram Bot credentials.
Create a bot using Telegram's official bot management interface and connect its credentials to the workflow.
Connect your Google Calendar account.
Configure the calendar used for appointments.
If required, configure the holiday calendar for the country where the business operates.
Configure the MongoDB connection.
The workflow currently uses a:
products
collection for product verification.
Adapt the collection and fields according to your database structure.
Configure the PostgreSQL connection.
The workflow currently writes complaint records to:
reclamations
Adapt the database schema if your business uses different fields.
Connect your ElevenLabs credentials for:
- Speech-to-text
- Text-to-speech
Configure the desired LLM provider.
The workflow includes support for OpenRouter and Qwen Cloud connections.
The model can be changed according to your requirements, availability and provider configuration.
A major design principle of this project is that many business rules are written directly inside the AI Agent prompt.
This means the workflow can be adapted without rewriting the entire automation architecture.
You can modify:
- Phone number format
- National ID format
- Required prefixes
- Required customer information
- Opening time
- Closing time
- Friday schedule
- Weekly closing days
- Holidays
- Supported devices
- Supported problems
- Price ranges
- Product verification
- Required information
- Required documents
- Customer instructions
- Appointment duration
- Confirmation requirements
- Booking reference format
- Cancellation requirements
- Modification rules
The current workflow supports:
Arabic
French
English
Additional languages can be added by extending the corresponding AI Agent instructions and customer-facing responses.
The workflow is not limited to the current country configuration.
For example, the current configuration may use:
Phone / ID:
8 digits
Specific prefixes
Business:
08:00 → 18:00
Friday → 16:30
Sunday → Closed
A user deploying the project in another country can modify these rules directly inside the AI Agent prompt.
For example:
Country A:
8 digits
Country B:
9 digits
Country C:
10 digits
The same concept applies to:
- Phone numbers
- National IDs
- Working hours
- Weekly holidays
- Public holidays
- Supported services
- Repair prices
The provided values are therefore example business rules for the current deployment, not universal requirements.
Before publishing the workflow, always verify that no private credentials or sensitive deployment information are present.
This repository is intended to contain:
- Workflow logic
- Configuration placeholders
- Documentation
- Screenshots
It should not contain:
- API keys
- Bot tokens
- OAuth access tokens
- OAuth refresh tokens
- Database passwords
- Private database connection strings
- Private production credentials
- Customer personal data
- Private chat IDs
n8n credential references are not the same thing as exposing the credential secret itself.
When deploying this workflow, reconnect your own credentials in n8n rather than publishing them inside the repository.
This project demonstrates an AI-powered business automation architecture.
Before using it in a real production environment, review:
- Personal-data protection
- Customer consent
- Data retention
- Database permissions
- Authentication
- Authorization
- Backup strategy
- Calendar permissions
- API permissions
- National-ID handling
- Complaint-data protection
- Logging and monitoring
The appropriate legal and privacy requirements depend on the country and business where the workflow is deployed.
The architecture is modular, so it is easy to build on. If you would like to take it further, here are some ideas:
- WhatsApp integration alongside Telegram
- Automatic appointment reminders
- Human-agent escalation for complex cases
- Receipt / image recognition with multimodal AI
- Document processing
- Customer authentication
- Real inventory management
- Payment integration
- CRM integration
- Customer notification system
- Role-based administration
- Advanced analytics and business intelligence dashboards
- Automatic reporting
- Customer history
Synapse Digital demonstrates how an AI Agent can be integrated into an actual business workflow rather than being used only as a conversational interface.
The project combines:
Artificial Intelligence
+
AI Agents
+
Workflow Automation
+
Business Rules
+
Databases
+
Calendar Automation
+
Voice AI
+
Business Intelligence
The objective is to create a system capable of understanding customers naturally while still operating under controlled business rules and real backend verification.
The complete n8n workflow is included in this repository.
This project is licensed under the Apache License 2.0.
See LICENSE for the complete license text.
You are free to use, modify and adapt the project according to the terms of the license.
Synapse Digital — AI Business Assistant
An AI-powered business automation system built with:
n8n • AI Agents • Telegram • Google Calendar • MongoDB • PostgreSQL • ElevenLabs • OpenRouter / Qwen
From conversational AI to real business automation.