Master 80% of n8n in 36 Minutes

Key Concepts

• Automation: Fixed sequence of actions (A to B to C).
• AI Agent: Dynamic, can reason, make decisions, and choose actions based on context.
• Nodes: Building blocks of NADN workflows, each performing a specific task.
• Triggers: Initiate workflows (e.g., schedule, form submission, chat input).
• Actions: Pre-built integrations with apps and services (e.g., Google Sheets, Gmail, Open Weather).
• HTTP Request: Universal adapter for connecting to any API.
• Logic Nodes: Control data flow (e.g., switch, filter, merge).
• AI Agent Node: Connects to a language model, adds memory, and attaches tools.
• JSON: JavaScript Object Notation, a data organization system used for data transfer between nodes.
• API: Application Programming Interface, how software talks to software.
• System Prompt: Custom instructions given to the AI agent to guide its behavior.
• RAG (Retrieval Augmented Generation): Uses a vector database to allow agents to pull from a larger knowledge base.

1. Introduction to NADN

• NADN is a node-based automation platform that allows users to build automations and AI agents.
• The video aims to teach the 80% of NADN that users will actually use by building three automations: a simple weather email, a form-driven workflow, and a full AI personal assistant.
• NADN can be used for personal productivity, marketing, e-commerce, data reporting, home automation, and AI integrations.

2. Automation vs. AI Agents

• Automations are fixed sequences of actions, following a predefined path.
• AI Agents are dynamic and can reason, make decisions, and choose actions based on context.
• Node-based platforms like NADN can be intimidating at first, but become intuitive with practice.

3. Building a Simple Weather Automation

• Goal: Create an automation that runs daily, checks the weather, and emails it to the user.
• Steps:
  1. Create a new workflow and name it.
  2. Add a Schedule Trigger node to run every day at 6 a.m.
  3. Add an Open Weather action node to get the weather data.
     • Connect credentials by creating an account on openweather.org and pasting the API key into the access token field.
     • Select desired options (e.g., current weather, Imperial units, zip code).
  4. Add a Gmail action node to send the weather data via email.
     • Connect credentials by signing in with Google and approving access.
     • Enter the recipient email, subject, and message body.
     • Use drag-and-drop to insert dynamic weather data from the Open Weather node into the email message using JSON.
     • Disable "append nad attribution" in options.
  5. Test the workflow and check the email.
• Canvas Tips:
  • Use the broom icon to tidy up the workflow.
  • Zoom in and out and move around the canvas.
  • Use "zoom to fit" to center the view.
  • The "executions" tab shows previous workflow runs and their status.

4. NADN Account Setup

• Cloud Version: Sign up on NADN's site (paid subscription after free trial).
• Self-Hosting:
  • Local Install with Docker: More technical, requires configuration.
  • Virtual Private Server (VPS): Easier middle ground, self-hosted but simpler to set up.
• Hostinger VPS Setup:
  1. Select the self-hosted NADN option under VPS services.
  2. Choose a plan (KVM2 is sufficient).
  3. Use coupon code Futuredia.
  4. Access the NADN installation through the Hostinger dashboard.
  5. Set up the NADN account and enter the free license key from email.

5. Core Node Types

• Triggers: Initiate workflows (schedule, form submission, chat input, webhook).
• Actions: Pre-built integrations with apps and services (Google products, weather tools, Notion, Airtable, Spotify, Stripe).
• HTTP Request: Connects to any API when not on the pre-built list.
• Logic Nodes: Control data flow (filter, branch, merge, loop).
• AI Agent Node: Powers AI agents, connects to language models, adds memory, and attaches tools.

6. Building a Form-Driven Workflow

• Goal: Route sponsorship inquiries based on budget, auto-respond by email, and log data to a Google Sheet.
• Steps:
  1. Create a new workflow and add a Form Submission trigger node.
  2. Define form fields (company name, contact name, email, website, budget, notes).
  3. Add a Switch logic node to route inquiries based on budget.
     • Create rules for budget less than $1 million and budget greater than or equal to $1 million.
  4. Add a Gmail action node to each branch to auto-respond by email.
     • Connect Google credentials (client ID and client secret).
     • Use drag-and-drop to insert dynamic data from the form into the email message.
     • Customize the email message based on the budget.
     • Disable "append nad attribution" in options.
  5. Add a Google Sheets action node to each branch to log the inquiry data.
     • Connect Google Sheets credentials.
     • Create a Google Sheet with columns for each form field and an "action" column.
     • Configure the Google Sheets node to append a row with the form data and the appropriate action (e.g., "decline" or "follow up sent").
  6. Rename nodes for clarity.
  7. Test the workflow by submitting a form and verifying the email and Google Sheet.

7. Expanding and Refining Workflows

• Enrich Data: Use an AI agent node to research the company and generate a summary or qualification report.
• Merge Nodes: Combine data from multiple streams into one.
• Set Node: Clean or reshape data before passing it on (e.g., strip symbols from a budget field).

8. Understanding JSON

• JSON (JavaScript Object Notation) is a way to organize information in text.
• $json refers to data from the previous node.
• $json.fieldName refers to a specific field in the previous node's data.
• item.json refers to a JSON item.
• math.round() can be used to round numbers.
• ChatGPT can be used to write and explain JSON snippets.

9. Building an AI Personal Assistant

• Goal: Create an AI agent that can handle tasks like checking calendar, managing emails, researching topics, and triggering other workflows.
• Core Components of an AI Agent:
  • Brain: Large language model (LLM) like ChatGPT, Claude, or Gemini.
  • Memory: Context for the agent (past interactions, conversation history, external memory sources).
  • Tools: How the agent interacts with the outside world (retrieving data, taking action, orchestrating workflows).
• Steps:
  1. Create a new workflow and add a Chat Trigger node.
  2. Add an AI Agent node.
     • Select OpenAI as the LLM and connect credentials (API key).
     • Choose a model (e.g., GPT-4).
     • Set up memory (e.g., simple memory with a limit of 10).
     • Add tools (e.g., Google Calendar, Gmail, Google Sheets).
     • Add a System Prompt to guide the agent's behavior.
       • Include the role, instructions, tools, relevant context, and restrictions.
       • Use ChatGPT to generate the system prompt.
  3. Test the agent by asking it questions and verifying its responses.
• Calling Other Workflows:
  • Use the "Call NADN Workflow" tool to connect the agent to other workflows.
  • Set up separate workflows for specific tasks (e.g., calendar management, email management).
  • Use webhooks to trigger workflows and receive responses.

10. HTTP Requests and APIs

• API (Application Programming Interface) is how software talks to software.
• HTTP Request node is used to connect to any public API, even when it's not officially integrated.
• GET: Pulls information (e.g., checking the weather).
• POST: Sends information (e.g., submitting a form).
• Example: Getting air quality data from airnow.gov using an HTTP request node.

11. Final Tips

• Expect errors and use them as learning opportunities.
• Use ChatGPT as a build partner.
• Explore retrieval augmented generation (RAG) with a vector database like Pinecone.

12. Conclusion

• NADN is a powerful platform for building automations and AI agents.
• By understanding the core concepts and following the steps outlined in the video, users can create custom workflows to automate tasks and improve productivity.
• The video encourages users to experiment, iterate, and continue learning to unlock the full potential of NADN.