Hello all, I am looking for some initial thoughts on a situation I want to solve. I constantly need to put curated employment guides in PDFs. This is a very manual process. I want to find a way with N8N to generate these PDFs from within a template document I have. In essence the flow would be: 1. The user provides the data 2. Agent fills in the template 3. PDF is extracted A nice to have: A return message from the agent saying that a section is empty, to provide data to fill it in... Not sure how to achieve this. I thank you all for your guidance!

I've been exploring ways to run Skool communities with minimal friction. Decided to share this simple but super useful setup, i was doing it a while back with my own communities. Here’s the setup: 1. Spin up your n8n instance (or do free self host using Google GCP, here's relevant vid) (Hostinger or Render is the cheapest route if you want something in 1 click). 2. Import my n8n workflow template 3. Use Chrome Developer Tools to grab your Group ID, Client ID, and Auth Token from the Network tab (follow the video for step-by-step). 4. Copy your Google Sheet ID and plug it into the workflow. Once it’s up and running, your Skool analytics (M.R.R, signups, retention, and more) will be logged and visualized in Google Sheets. Just add a chart and voilà—daily insights, automated. Perfect for: ✅ Community owners reporting to clients ✅ Teams managing multiple Skool groups ✅ Anyone tired of checking analytics manually Let me know if anything breaks. P.S. Ideas to use this: - Run this for multiple Skool communities at once - Turn this into a client reporting productized service - Add AI summaries of trends directly in Sheets - Push the analytics into Slack or Notion daily

Here’s the deal: I’ve got a sweet little JSON file – just some structured data about my family: names, birthdays, emails… you know, the usual stuff a smart assistant should remember so it doesn’t miss grandma’s birthday 🎂. I’m building a personal assistant using n8n, and I want it to have long-term memory via Qdrant. The goal? Context-aware responses, useful recall, and eventually a system that manages our family life better than I do 🧑‍💻➡️🤖. Now comes the fun (read: painful) part – getting that data into Qdrant the right way: Embed each object ✍️ Send it to Qdrant via Python or n8n 🔄 Structure the collection properly 🧱 And make sure nothing breaks along the way 🧨 Right now, it feels less like a “simple import” and more like onboarding a new team member 👔. So… if anyone here has a best practice, a code snippet, or a clean flow for uploading JSON data into Qdrant – ideally with custom fields and embeddings – I’d love to hear it! Even a meme or moral support is appreciated 😅🙌

I once heard from the Open AI people that bots hallucinate, and that's why it's dangerous to believe everything they say. At the agent level, it's essential to be able to control these types of problems, which can be very delicate in areas like healthcare. Do you believe hallucinations are inevitable, or are there simple models or solutions to achieve high accuracy?

Hey everyone, It's been a while since I was dreaming about a personal assistant (with very basic coding knowledge). and I discovered what truly was an AI agent and understood that it's something I can use to fulfil my wish. Though I have no budget. So I started to dig Youtube: - Self-host n8n: check - Using Ollama : check - Using Ngrok to go on the internet : check - fought with google Oauth 2 : I won And now my machine : an old macbook pro Intel. Ollama works and I find it already cool for conversation testing the different LLM's (Deepseek is giving me creepy vibe though with how it shows it's thought process). I'm using Nate's template for prompts foundation and I'm now reaching some limits: I'm managing to create events in my calendar and sometimes it's wrong in timing and duration. Though it can't delete or update my events and it just hallucinate the answer. My questions: - Is it a power problem because my computer is old? - Should I keep improving the prompt till It work or is there some trick by creating an agent for each calendar tool instead of giving all the tools to one agent? - Should I spend the money to get OpenAI or Anthropic solution to at least get some result and keep the motivation up and not give up on AI agent - Or buy a new computer adapted for my need? Does anyone have experience with getting a personal assistant working an intel processor i7 2,5GHz and AMD Radeon R9 M370X 2 Go? My thought is just get claude or chatGPT and then get yourself a VPS to get your first success with AI agent in general and go back to Open LLM's a bite later when you have power to run them Thank you very much for your advices, Byron PS: I placed it in general discussion because of how broad the topic seems. Let me know if I should place it in another category!