Hello,
I’m new to n8n.
I’ve set up a simple workflow in n8n to analyze an Excel sheet uploaded to my Google Drive. The sheet contains three columns: Date, Expenses/Income, and Expense Description. The workflow extracts data, aggregates it, and summarizes it before uploading the processed data to Supabase.
The goal is to use a GPT agent to perform operations like:
• Summing up expenses for a specific date range.
• Listing all expenses for a specific day.
However, the GPT agent’s responses are inconsistent:
• For small queries (e.g., sum expenses for 2-3 days), it sometimes works.
but more often say “I can’t find info on document”
• When summing expenses for larger ranges or the entire dataset, the agent often fails, saying, “I can’t find info on document” .
• If I ask ‘give me the date we have in the document’, it sometimes give me some dates missing one or two dates (random), some time say “I can’t find info on document”
Here’s the workflow structure:
-
Data Extraction Node: Reads the Google Sheets data.
-
Aggregation Node: Groups or processes the data.
-
Summarize Node: Prepares data before uploading to Supabase.
-
The processed data is uploaded to Supabase for GPT agent consumption.
I suspect the issue might be related to how the data is aggregated, summarized, or formatted before being processed by the GPT agent.
It seems agent can’t do simple operations and loose random data / row on the excel/csv sheet. have you fixed or find the same issue?
What could be causing this inconsistency? How can I ensure the GPT agent recognizes and processes all data correctly for queries like summing expenses over large ranges or providing specific day data?
Thank you for any advice!
- n8n version: Version 1.68.0
- Database (default: SQLite): Supabase
- n8n EXECUTIONS_PROCESS setting (default: own, main):
- Running n8n via (Docker, npm, n8n cloud, desktop app): npm
- Operating system: Mac os 14.5
