Hi everyone,
I’m developing a Task Manager Agent in n8n that processes natural language commands to add tasks to a Google Sheet. The sheet has five columns: Task ID, Task Description, Status, Due Date, and Date Created. Each new task should get a unique Task ID and be added without overwriting any previous entries.
Environment Context:
- n8n Version: 1.76.1
- Hosting: n8n Cloud Hosted
- AI Models Tried: GPT-4o and GPT-4o-mini
The Challenges:
- Overwriting Tasks:
When I send a command like “Add apples and bananas to my list,” the agent appears to add the first task (e.g., “apples”) but then replaces it with the second task (“bananas”). I was expecting both tasks to be added without interference. - False Confirmations:
Sometimes, the agent confirms that it has added items to the Google Sheet, but when I check the sheet, those items are missing. This inconsistency is quite frustrating and makes troubleshooting harder.
What I’ve Tried So Far:
- Re-reading the Sheet for Each Task:
I initially had the agent re-read the entire sheet before processing each task. However, this approach ended up using the same highest Task ID for multiple tasks, which led to overwrites. - Bundling Tasks into a JSON Array:
I then adjusted the prompt to bundle multiple tasks into a single JSON array for one update, hoping that this would prevent one task from overwriting the other. Unfortunately, the problem persisted. - Single Read Before Processing All Tasks:
Finally, I modified the prompt so that the agent reads the sheet once at the start, calculates new Task IDs for all tasks, and outputs them together as a JSON array. Yet, even this approach hasn’t solved the issue.
Given these challenges, I’m wondering if switching from Google Sheets to Airtable might help. Has anyone encountered similar problems with Google Sheets in n8n and found that Airtable works more reliably? Would moving to Airtable address the overwriting issue and false confirmations I’m experiencing, or is there a better workaround that keeps me on Google Sheets?
I’d really appreciate any feedback, experiences, or ideas the community might have.
Thanks in advance for your support!