I was very disappointed after building my first workflow with an AI Tool Agent node. No matter what LLM I tested, the agent responded with wrong event times. How could it be that neither 4o, o1, nor sonnet 3.5 could solve such a simple task. Investigating a little further, I noticed the Google Calendar node doesn't respond with sufficient information for recurring events. Here is a (redacted) example output: ```json [ { "kind": "calendar#event", // ... redacted ... "start": { "dateTime": "2023-09-07T09:30:00+02:00", "timeZone": "Europe/Berlin" }, "end": { "dateTime": "2023-09-07T11:00:00+02:00", "timeZone": "Europe/Berlin" }, // ... redacted ... "recurrence": [ "RRULE:FREQ=WEEKLY;UNTIL=20300101T225959Z;BYDAY=TH" ], // ... redacted ... "nextOccurrence": { "start": { "dateTime": "2024-12-19T08:42:42+00:00", "timeZone": "Europe/Berlin" }, "end": { "dateTime": "2024-12-19T10:12:42+00:00", "timeZone": "Europe/Berlin" } } } ] ``` As you can see, the time part of `nextOccurrence.start` and `nextOccurrence.end` differ from `start` and `end`. In fact 08:42:42 (`nextOccurrence.start`) was the time I queried the Google Calendar API. As a person who heavily works with recurring events, this is a showstopper. I wonder if I’m just being dense or missing something obvious. As nobody mentioned such a behavior before. Or is there anybody who has worked out a workaround?

Hi guys, I am a newbie in all this. Nate has great videos on the Ai Assistant which is truly helpful for me since I work as self-employed and managing kids. It's crazy as you can possibly imaging. I would really like to use the N8N agents and sort my work and life. However, having N8N cloud is $20 Euros a month, ChatGPT plus is another $20 USD a month, not to mention other services like Gmails or Microsoft. When it adds up, it becomes a big number... and at this stage of the financial situation, it is not really helping with budgeting. So I was wondering if there is a possibility of running N8N on a docker on AWS or any cloud service, to minimize some cost. If so, could someone help to guide the process. I have tried and not succeeded. Or any other suggestions. Cybersecurity is another of my concern as well. Please let me know! Thanks a lot!

Below is what I keep getting from Gemini 1.5 pro even with Temp 0.0 in N8N using Tool AI Agent with Call Workflow tool node. Maybe this has something to do with it's internal guardrails... 😜 "I apologize that my previous responses were misleading. I do not have access to external systems or tools, including the ones mentioned like `onelake_tool` or `power_bi_tool`. I am a large language model, and my responses are based on the information I have been trained on. I tried to interpret your initial encoded instructions as best I could, but I simulated using a tool when I do not have the capability to do so. I am still under development and learning to interpret complex instructions. I will strive to be more transparent in my responses in the future."

Has anyone else been noticing that when getting events from Google Calendar APIs, the returned event date time do not alway show correctly? Below is an example output returned from the node where the timestamp is way off from actuals. Even if I apply any timezone offset, it's still wrong. This happens quite often if one is hitting the API consistently. "nextOccurrence": { "start": { "dateTime": "2024-12-14T09:34:15+00:00", "timeZone": }, "end": { "dateTime": "2024-12-14T10:34:15+00:00", "timeZone": } }

Hello there, I've just started learning about N8N and was testing out this workflow. Has anyone else experienced seeing the agent creating up to 3 60 minute events with a simple prompt like this "set an N8N appointment for tomorrow from 2pm to 5pm." if you're using 4o-mini? Was confused why until I tried with 4o instead and it created a single event every time.