With n8n, I'm implementing a feature that interprets any image on a webpage, converts it to a PDF, and sends it via Telegram.
It feels like stacking building blocks. It maintains programming concepts, but from a user's perspective, you only need to set up a very simple flow. It's like the expert mode of Scratch, a block coding tool used in middle and high schools these days.
It's not difficult when I set it up while asking Gemini how to use it.
The problem is configuring the part that connects AI services. Gemini AI Pro subscription has restrictions compared to the API pricing plan, so it seems I need to split questions and use time-spaced attacks. If I use the API pricing plan, I could just fire it off directly.
There's a big difference in image interpretation by Gemini model. Gemini-2.5-flash properly recognizes two men with permed hair, but gemini-3-27b-it recognizes them as a mother and son. ㅠ.
When using AI models, the waiting time is quite long. It's tedious. I'll need to either multitask with other work or watch a movie while waiting.
After setting up an API pricing plan and doing various things in the past, getting hit with $50 in half a day, I'm scared of API pricing plans now. ㅠ.
