n8n Crash Course - Automated AI Workflows

Key Concepts

• N8N: A free, source-available platform for building AI workflows and automations using a graphical user interface (GUI).
• Source Available vs. Open Source: N8N's source code is accessible but has restrictive licensing, preventing modification or hosting for others or building competing products.
• Workflows: A series of interconnected nodes (steps) that automate tasks.
• Triggers: The starting point of a workflow, initiated manually, by form submission, chat messages, or events from integrated applications.
• Nodes: Individual steps within a workflow that perform specific actions (e.g., AI actions, app integrations, code execution, flow control).
• Credentials: Authentication information required to connect N8N to external services (e.g., API keys, OAuth tokens).
• AI Agents: Advanced AI components within N8N that can utilize models, memory, and tool use (including other workflows).
• Tool Use: The ability of AI agents to leverage external tools or workflows to perform tasks.
• Google Sheets API: Used for reading and writing data to Google Sheets.
• Google Calendar API: Used for creating and managing calendar events.
• Telegram Bot API: Used for sending messages through a Telegram bot.
• Python Code Node: Allows execution of custom Python scripts within workflows.
• Form Submission Trigger: Initiates a workflow when a user submits a form created within N8N.
• Summarize Node: Used for aggregating data, similar to pandas in Python.
• If/Else Node: Controls the flow of a workflow based on conditions.
• Templates: Pre-built N8N workflows that can be imported and customized.

N8N Crash Course and Workflow Automation

This video provides a comprehensive introduction to N8N, a free and "source-available" platform for building AI workflows and automations through a graphical user interface. The presenter emphasizes that while the source code is available, the license is restrictive, preventing users from modifying or hosting it for others or building competing products. However, it can be deployed on a company's own server for internal use.

Setting Up and Understanding N8N

• Local Installation: N8N can be run locally by installing Node.js and npm, then executing npx n8n in the terminal. This downloads and runs the software on port 5678.
• User Interface: Upon first run, a registration form appears for local account creation. The main interface displays existing workflows and allows for the creation of new ones.
• Workflow Structure: Workflows begin with a trigger, which can be manual, a form submission, a chat message, or an event from an integrated app (e.g., a new row in Google Sheets, a PayPal event).
• Nodes: After the trigger, nodes are added to perform actions. These include AI actions, app integrations (like Google Sheets, Telegram), flow control (if/else, filters, transforms), and code execution (Python).
• Data Flow: Each node takes input, processes it, and produces output, which is then passed to the next node in the workflow.

Workflow Example 1: Form Submission to Google Sheets and Telegram

This section details building a workflow that captures user registrations via a form and then logs them into a Google Sheet and sends a notification to Telegram.

1. Form Submission Trigger:

   • A "Form Submission" trigger is configured with fields for "name" (text, required), "email" (email, required), and "reason" (text, optional).
   • When the form is submitted, the data is outputted as a JSON object.

2. Google Sheets Integration:

   • Google Cloud Setup:
     • A new project is created in Google Cloud Console (e.g., "N8N tutorial project").
     • The Google Sheets API and Google Drive API are enabled.
     • An OAuth consent screen is set up with an app name ("N8N app tutorial"), user email, and scopes for Google Sheets and Google Drive.
     • A web application client is created, obtaining a Client ID and Client Secret.
     • The N8N redirect URL is provided to Google.
   • N8N Credentials:
     • In N8N, a new credential for Google Sheets is created using OAuth 2.0.
     • The obtained Client ID and Client Secret are entered.
     • The "Sign in with Google" button is used to authenticate with the Google account.
   • Appending to Sheet:
     • A "Google Sheets" node is added, configured to "Append Row."
     • The N8N Google Sheet document and the "beta signups" sheet are selected.
     • The form fields (name, email, reason) are mapped to the corresponding columns in the Google Sheet.
     • Upon execution, a new row is added to the sheet with the submitted form data.

3. Telegram Integration:

   • Telegram Bot Setup:
     • The "BotFather" bot on Telegram is used to create a new bot (e.g., "N8N tutorial testbot").
     • A unique API Token is generated for the bot.
   • N8N Credentials:
     • A new Telegram credential is created in N8N using the bot's API Token.
   • Sending Messages:
     • A "Telegram" node is added to "Send Text Message."
     • To send messages to a specific chat, the bot needs to be added to a group chat.
     • A "Get IDs bot" is used to retrieve the Chat ID of the group.
     • The Chat ID is entered into the Telegram node.
     • The message content is defined, dynamically including data from the form submission (e.g., "New user signed up: Name is {{ $json.name }}, Email is {{ $json.email }}").
     • When the workflow runs, a Telegram message is sent to the specified group.

Workflow Example 2: Chatbot with AI and Google Calendar Integration

This section demonstrates building a workflow that responds to chat messages, uses an AI model to extract event details, and creates a Google Calendar entry.

1. Chat Trigger:

   • A "Chat" trigger is used, allowing interaction via text messages within N8N. The input is captured as chat input.

2. AI Model Interaction (OpenAI):

   • An "AI" node is added, specifically "OpenAI Message a Model."
   • OpenAI Credentials: An API key from platform.openai.com/api-keys is required.
   • A GPT model (e.g., GPT-4) is selected.
   • The prompt is constructed to instruct the AI to extract event details (start time, end time, title, description) from the chat input and return them as a JSON object. The "Output content as JSON" option is enabled.
   • Example Prompt: "Based on the user message, figure out what the calendar entry should look like and return a JSON object with start, end, title, and description."
   • Example User Message: "Tomorrow at 4:00 p.m., I have an appointment at my barber."
   • The AI outputs a JSON object containing the extracted event details.

3. Google Calendar Integration:

   • Google Calendar API Setup: Similar to Google Sheets, the Google Calendar API needs to be enabled in the Google Cloud Console, and appropriate scopes (Google Calendar API) must be added to the OAuth consent screen.
   • N8N Credentials: New credentials for Google Calendar are created using the same OAuth 2.0 process as for Google Sheets.
   • Creating an Event:
     • A "Google Calendar" node is added, configured to "Create Event."
     • A specific calendar is selected.
     • The "Start," "End," "Summary" (title), and "Description" fields are mapped to the corresponding values from the AI's JSON output.
     • The presenter notes potential time zone issues, which can be addressed by instructing the AI or using JavaScript.

4. Workflow Response:

   • The "Chat" node is configured to provide a response.
   • A "Respond to Chat" node is added, allowing a custom message to be sent back to the user, confirming the event creation (e.g., "Successfully created event for {{ $json.summary }}.").

Workflow Example 3: Sales Analysis and Reporting

This section details a more complex workflow to analyze sales data from a Google Sheet, compare today's sales to yesterday's, and generate reports via Telegram.

1. Manual Trigger: A manual trigger is used for on-demand execution.

2. Get Rows from Google Sheet:

   • A "Google Sheets" node retrieves all rows from "Sheet 1" of the N8N sheet.
   • The "Values Unformatted" option is used to get raw data.

3. Filtering by Date:

   • A "Filter" node (or a Python node with JavaScript expressions) is used to select only today's and yesterday's sales data.
   • A JavaScript expression is used to calculate yesterday's date and compare it with the sale date.

4. Calculating Total Price:

   • A "Python" node calculates the total price for each sale by multiplying "quantity" and "unit price."

5. Aggregating Sales:

   • A "Summarize" node (data transformation) is used to sum the "total" price for today and yesterday, grouping by date.

6. Restructuring Data (Python):

   • Another "Python" node restructures the aggregated data into a format with distinct "today" and "yesterday" objects, each containing their respective sum totals.

7. Conditional Logic (If/Else):

   • An "If" node compares today's sales with yesterday's sales.

8. Branch 1: Success (More Sales Today):

   • If today's sales are greater than yesterday's, a "Telegram" node sends a congratulatory message (e.g., "Great, you earned more today compared to yesterday.").

9. Branch 2: Analysis (Less Sales Today):

   • If today's sales are less than yesterday's:
     • Retrieving Original Sales Data (Python): A "Python" node is used to access the original, unaggregated sales data from a previous node. This is crucial for providing context to the AI.
     • AI Analysis (OpenAI): An "OpenAI Message a Model" node (using GPT-4) is used.
       • The prompt instructs the AI to analyze the sales data, explain why earnings were less today compared to yesterday, and format the answer as a concise chat message for Telegram.
       • The prompt includes the parsed JSON string of the sales data.
     • Telegram Report: A "Telegram" node sends the AI-generated analysis to the user.

10. Workflow Execution and Testing: The presenter demonstrates changing sales figures in the Google Sheet and re-running the workflow to show both the congratulatory message and the analytical report being generated.

AI Agents and Tool Use

• AI Agent Node: N8N allows the use of AI Agents, which can be configured with:
  • Models: Selecting an AI model (e.g., GPT-4o, GPT-4).
  • Memory: Simple memory to retain conversation history.
  • Tool Use: The ability to call other N8N workflows or external tools.
• Example: The sales analysis workflow (Workflow 3) is integrated as a tool that an AI agent can call when the user asks to compare sales. The agent recognizes the intent ("compare sales of today versus yesterday") and triggers the workflow.

Exploring Templates and Learning

• Templates: N8N offers a library of pre-built workflow templates that users can import and adapt.
• Learning: Users can learn by examining the nodes within these templates to understand how they function and interact. An example of a complex template involving image generation, video script creation, and social media posting is shown.

Conclusion

N8N is presented as a powerful and versatile platform for automating tasks and integrating various services, including AI models. Its graphical interface makes it accessible to both technical and non-technical users. The video highlights its capabilities through practical examples, demonstrating how to build sophisticated workflows for data management, communication, and AI-driven analysis. The presenter encourages viewers to explore the platform, utilize templates, and consider their services for personal tutoring or freelance projects.