Monitor Emails & Send AI-Generated Auto-Replies with Ollama & Telegram Alerts

Description

Workflow Overview

This advanced n8n workflow provides intelligent email automation with AI-generated responses. It combines four core functions:
Monitors incoming emails via IMAP (e.g., SOGo)
Sends instant Telegram notifications for all new emails
Uses AI (Ollama LLM) to generate contextual, personalized auto-replies
Sends confirmation notifications when auto-replies are sent

Unlike traditional auto-responders, this workflow analyzes email content and creates unique, relevant responses for each message.

Setup Instructions

Prerequisites

Before setting up this workflow, ensure you have:
An n8n instance (self-hosted or cloud) with AI/LangChain nodes enabled
IMAP email account credentials (e.g., SOGo, Gmail, Outlook)
SMTP server access for sending emails
Telegram Bot API credentials
Telegram Chat ID where notifications will be sent
Ollama installed locally or accessible via network (for AI model)
The llama3.1 model downloaded in Ollama

Step 1: Install and Configure Ollama

Local Installation

Install Ollama on your system:
Visit https://ollama.ai and download the installer for your OS
Follow installation instructions for your platform

Download the llama3.1 model:
ollama pull llama3.1

Verify the model is available:
ollama list

Start Ollama service (if not already running):
ollama serve

Test the model:
ollama run llama3.1 "Hello, world!"
Remote Ollama Instance

If using a remote Ollama server:
Note the server URL (e.g., http://192.168.1.100:11434)
Ensure network connectivity between n8n and Ollama server
Verify firewall allows connections on port 11434

Step 2: Configure IMAP Credentials

Navigate to n8n Credentials section
Create a new IMAP credential with the following information:
Host: Your IMAP server address
Port: Usually 993 for SSL/TLS
Username: Your email address
Password: Your email password or app-specific password
Enable SSL/TLS: Yes (recommended)
Security: Use STARTTLS or SSL/TLS

Step 3: Configure SMTP Credentials

Create a new SMTP credential in n8n
Enter the following details:
Host: Your SMTP server address (e.g., Postfix server)
Port: Usually 587 (STARTTLS) or 465 (SSL)
Username: Your email address
Password: Your email password or app-specific password
Secure connection: Enable based on your server configuration
Allow unauthorized certificates: Enable if using self-signed certificates

Step 4: Configure Telegram Bot

Create a Telegram bot via BotFather:
Open Telegram and search for @BotFather
Send /newbot command
Follow instructions to create your bot
Save the API token provided by BotFather

Obtain your Chat ID:
Method 1: Send a message to your bot, then visit:
https://api.telegram.org/bot<YOUR_BOT_TOKEN>/getUpdates
Method 2: Use a Telegram Chat ID bot like @userinfobot
Method 3: For group chats, add the bot to the group and check the updates
Note: Group chat IDs are negative numbers (e.g., -1234567890123)

Add Telegram API credential in n8n:
Credential Type: Telegram API
Access Token: Your bot token from BotFather

Step 5: Configure Ollama API Credential

In n8n Credentials section, create a new Ollama API credential
Configure based on your setup:
For local Ollama: Base URL is usually http://localhost:11434
For remote Ollama: Enter the server URL (e.g., http://192.168.1.100:11434)
Test the connection to ensure n8n can reach Ollama

Step 6: Import and Configure Workflow

Import the workflow JSON into your n8n instance
Update the following nodes with your specific information:

Check Incoming Emails Node
Verify IMAP credentials are connected
Configure polling interval (optional):
Default behavior checks on workflow trigger schedule
Can be set to check every N minutes
Set mailbox folder if needed (default is INBOX)

Send Notification from Incoming Email Node
Update chatId parameter with your Telegram Chat ID
Replace -1234567890123 with your actual chat ID
Customize notification message template if desired
Current format includes: Sender, Subject, Date-Time

Dedicate Filtering As No-Response Node
Review spam filter conditions:
Blocks emails from addresses containing "noreply" or "no-reply"
Blocks emails with "newsletter" in subject line (case-insensitive)
Add additional filtering rules as needed:
Block specific domains
Filter by keywords
Whitelist/blacklist specific senders

Ollama Model Node
Verify Ollama API credential is connected
Confirm model name: llama3.1:bf230501 (or adjust to your installed version)
Context window set to 4096 tokens (sufficient for most emails)
Can be adjusted based on your needs and hardware capabilities

Basic LLM Chain Node
Review the AI prompt engineering (pre-configured but customizable)
Current prompt instructs the AI to:
Read the email content
Identify main topic in 2-4 words
Generate a professional acknowledgment response
Keep responses consistent and concise
Modify prompt if you want different response styles

Send Auto-Response in SMTP Node
Verify SMTP credentials are connected
Check fromEmail uses correct email address:
Currently set to {{ $('Check Incoming Emails - IMAP (example: SOGo)').item.json.to }}
This automatically uses the recipient address (your mailbox)
Subject automatically includes "Re: " prefix with original subject
Message text comes from AI-generated content

Send Notification from Response Node
Update chatId parameter (same as first notification node)
This sends confirmation that auto-reply was sent
Includes original email details and the AI-generated response text

Step 7: Test the Workflow

Perform initial configuration test:
Test Ollama connectivity: curl http://localhost:11434/api/tags
Verify all credentials are properly configured
Check n8n has access to required network endpoints

Execute a test run:
Click "Execute Workflow" button in n8n
Send a test email to your monitored inbox
Use a clear subject and body for better AI response

Verify workflow execution:
First Telegram notification received (incoming email alert)
AI processes the email content
Auto-reply is sent to the original sender
Second Telegram notification received (confirmation with AI response)
Check n8n execution log for any errors

Verify email delivery:
Check if auto-reply arrived at sender's inbox
Verify it's not marked as spam
Review AI-generated content for appropriateness

Step 8: Fine-Tune AI Responses

Send various types of test emails:
Different topics (inquiry, complaint, information request)
Various email lengths (short, medium, long)
Different languages if applicable

Review AI-generated responses:
Check if topic identification is accurate
Verify response appropriateness
Ensure tone is professional

Adjust the prompt if needed:
Modify topic word count (currently 2-4 words)
Change response template
Add language-specific instructions
Include custom sign-offs or branding

Step 9: Activate the Workflow

Once testing is successful and AI responses are satisfactory:
Toggle the workflow to "Active" state
The workflow will now run automatically on the configured schedule

Monitor initial production runs:
Review first few auto-replies carefully
Check Telegram notifications for any issues
Verify SMTP delivery rates

Set up monitoring:
Enable n8n workflow error notifications
Monitor Ollama resource usage
Check email server logs periodically

How to Use

Normal Operation

Once activated, the workflow operates fully automatically:

Email Monitoring: The workflow continuously checks your IMAP inbox for new messages based on the configured polling interval or trigger schedule.

Immediate Incoming Notification: When a new email arrives, you receive an instant Telegram notification containing:
Sender's email address
Email subject line
Date and time received
Note indicating it's from IMAP mailbox

Intelligent Filtering: The workflow evaluates each email against spam filter criteria:
Emails from "noreply" or "no-reply" addresses are filtered out
Emails with "newsletter" in the subject line are filtered out
Filtered emails receive notification but no auto-reply
Legitimate emails proceed to AI response generation

AI Response Generation: For emails that pass the filter:
The AI reads the full email content
Analyzes the main topic or purpose
Generates a personalized acknowledgment
Creates a professional response that:
Thanks the sender
References the specific topic
Promises a personal follow-up
Maintains professional tone

Automatic Reply Delivery: The AI-generated response is sent via SMTP to the original sender with:
Subject line: "Re: [Original Subject]"
From address: Your monitored mailbox
Body: AI-generated contextual message

Response Confirmation: After the auto-reply is sent, you receive a second Telegram notification showing:
Original email details (sender, subject, date)
The complete AI-generated response text
Confirmation of successful delivery

Understanding AI Response Generation

The AI analyzes emails intelligently:

Example 1: Business Inquiry
Incoming Email: "I'm interested in your consulting services for our Q4 project..."
AI Topic Identification: "consulting services"
Generated Response: "Dear Correspondent! Thank you for your message regarding consulting services. I will respond with a personal message as soon as possible. Have a nice day!"

Example 2: Technical Support
Incoming Email: "We're experiencing issues with the API integration..."
AI Topic Identification: "API integration issues"
Generated Response: "Dear Correspondent! Thank you for your message regarding API integration issues. I will respond with a personal message as soon as possible. Have a nice day!"

Example 3: General Question
Incoming Email: "Could you provide more information about pricing?"
AI Topic Identification: "pricing information"
Generated Response: "Dear Correspondent! Thank you for your message regarding pricing information. I will respond with a personal message as soon as possible. Have a nice day!"

Customizing Filter Rules

To modify which emails receive AI-generated auto-replies:

Open the "Dedicate Filtering As No-Response" node
Modify existing conditions or add new ones:

Block specific domains:
{{ $json.from.value[0].address }}
Operation: does not contain
Value: @spam-domain.com

Whitelist VIP senders (only respond to specific people):
{{ $json.from.value[0].address }}
Operation: contains
Value: @important-client.com

Filter by subject keywords:
{{ $json.subject.toLowerCase() }}
Operation: does not contain
Value: unsubscribe

Combine multiple conditions:
Use AND logic (all must be true) for stricter filtering
Use OR logic (any can be true) for more permissive filtering

Customizing AI Prompt

To change how the AI generates responses:

Open the "Basic LLM Chain" node
Modify the prompt text in the "text" parameter
Current structure:
Context setting (read email, identify topic)
Output format specification
Rules for AI behavior

Example modifications:

Add company branding:
Return only this response, filling in the [TOPIC]:

Dear Correspondent!
Thank you for reaching out to [Your Company Name] regarding [TOPIC].
I will respond with a personal message as soon as possible.
Best regards,
[Your Name]
[Your Company Name]

Make it more casual:
Return only this response, filling in the [TOPIC]:

Hi there!
Thanks for your email about [TOPIC].
I'll get back to you personally soon.
Cheers!

Add urgency classification:
Read the email and classify urgency (Low/Medium/High).
Identify the main topic.

Return:

Dear Correspondent!
Thank you for your message regarding [TOPIC].
Priority: [URGENCY]
I will respond with a personal message as soon as possible.

Customizing Telegram Notifications

Incoming Email Notification:
Open "Send Notification from Incoming Email" node
Modify the "text" parameter
Available variables:
{{ $json.from }} - Full sender info
{{ $json.from.value[0].address }} - Sender email only
{{ $json.from.value[0].name }} - Sender name (if available)
{{ $json.subject }} - Email subject
{{ $json.date }} - Date received
{{ $json.textPlain }} - Email body (use cautiously for privacy)
{{ $json.to }} - Recipient address

Response Confirmation Notification:
Open "Send Notification from Response" node
Modify to include additional information
Reference AI response: {{ $('Basic LLM Chain').item.json.text }}

Monitoring and Maintenance

Daily Monitoring

Check Telegram Notifications**: Review incoming email alerts and response confirmations
Verify AI Quality**: Spot-check AI-generated responses for appropriateness
Email Delivery**: Confirm auto-replies are being delivered (not caught in spam)

Weekly Maintenance

Review Execution Logs**: Check n8n execution history for errors or warnings
Ollama Performance**: Monitor resource usage (CPU, RAM, disk space)
Filter Effectiveness**: Assess if spam filters are working correctly
Response Quality**: Review multiple AI responses for consistency

Monthly Maintenance

Update Ollama Model**: Check for new llama3.1 versions or alternative models
Prompt Optimization**: Refine AI prompt based on response quality observations
Credential Rotation**: Update passwords and API tokens for security
Backup Configuration**: Export workflow and credentials (securely)

Advanced Usage

Multi-Language Support

If you receive emails in multiple languages:

Modify the AI prompt to detect language:
Detect the email language.
Generate response in the SAME language as the email.

If English: [English template]
If Hungarian: [Hungarian template]
If German: [German template]

Or use language-specific conditions in the filtering node

Priority-Based Responses

Generate different responses based on sender importance:

Add an IF node after filtering to check sender domain
Route VIP emails to a different LLM chain with priority messaging
Standard emails use the normal AI chain

Response Logging

To maintain a record of all AI interactions:

Add a database node (PostgreSQL, MySQL, etc.) after the auto-reply node
Store: timestamp, sender, subject, AI response, delivery status
Use for compliance, analytics, or training data

A/B Testing AI Prompts

Test different prompt variations:

Create multiple LLM Chain nodes with different prompts
Use a randomizer or round-robin approach
Compare response quality and user feedback
Optimize based on results

Troubleshooting

Notifications Not Received

Problem: Telegram notifications not appearing

Solutions:
Verify Chat ID is correct (positive for personal chats, negative for groups)
Check if bot has permissions to send messages
Ensure bot wasn't blocked or removed from group
Test Telegram API credential independently
Review n8n execution logs for Telegram API errors

AI Responses Not Generated

Problem: Auto-replies sent but content is empty or error messages

Solutions:
Check Ollama service is running: ollama list
Verify llama3.1 model is downloaded: ollama list
Test Ollama directly: ollama run llama3.1 "Test message"
Review Ollama API credential URL in n8n
Check network connectivity between n8n and Ollama
Increase context window if emails are very long
Monitor Ollama logs for errors

Poor Quality AI Responses

Problem: AI generates irrelevant or inappropriate responses

Solutions:
Review and refine the prompt engineering
Add more specific rules and constraints
Provide examples in the prompt of good vs bad responses
Adjust topic word count (increase from 2-4 to 3-6 words)
Test with different Ollama models (e.g., llama3.1:70b for better quality)
Ensure email content is being passed correctly to AI

Auto-Replies Not Sent

Problem: Workflow executes but emails not delivered

Solutions:
Verify SMTP credentials and server connectivity
Check fromEmail address is correct
Review SMTP server logs for errors
Test SMTP sending independently
Ensure "Allow unauthorized certificates" is enabled if needed
Check if emails are being filtered by spam filters
Verify SPF/DKIM records for your domain

High Resource Usage

Problem: Ollama consuming excessive CPU/RAM

Solutions:
Reduce context window size (from 4096 to 2048)
Use a smaller model variant (llama3.1:8b instead of default)
Limit concurrent workflow executions in n8n
Add delay/throttling between email processing
Consider using a remote Ollama instance with better hardware
Monitor email volume and processing time

IMAP Connection Failures

Problem: Workflow can't connect to email server

Solutions:
Verify IMAP credentials are correct
Check if IMAP is enabled on email account
Ensure SSL/TLS settings match server requirements
For Gmail: enable "Less secure app access" or use App Passwords
Check firewall allows outbound connections on IMAP port (993)
Test IMAP connection using email client (Thunderbird, Outlook)

Workflow Not Triggering

Problem: Workflow doesn't execute automatically

Solutions:
Verify workflow is in "Active" state
Check trigger node configuration and schedule
Review n8n system logs for scheduler issues
Ensure n8n instance has sufficient resources
Test manual execution to isolate trigger issues
Check if n8n workflow execution queue is backed up

Workflow Architecture

Node Descriptions

Check Incoming Emails - IMAP: Polls email server at regular intervals to retrieve new messages from the configured mailbox.

Send Notification from Incoming Email: Immediately sends formatted notification to Telegram for every new email detected, regardless of spam status.

Dedicate Filtering As No-Response: Evaluates emails against spam filter criteria to determine if AI processing should occur.

No Operation: Placeholder node for filtered emails that should not receive an auto-reply (spam, newsletters, automated messages).

Ollama Model: Provides the AI language model (llama3.1) used for natural language processing and response generation.

Basic LLM Chain: Executes the AI prompt against the email content to generate contextual auto-reply text.

Send Auto-Response in SMTP: Sends the AI-generated acknowledgment email back to the original sender via SMTP server.

Send Notification from Response: Sends confirmation to Telegram showing the auto-reply was successfully sent, including the AI-generated content.

AI Processing Pipeline

Email Content Extraction: Email body text is extracted from IMAP data
Context Loading: Email content is passed to LLM with prompt instructions
Topic Analysis: AI identifies main subject or purpose in 2-4 words
Template Population: AI fills response template with identified topic
Output Formatting: Response is formatted and cleaned for email delivery
Quality Assurance: n8n validates response before sending