Hi everyone,
We’re trying to set up a lightweight, AI-driven database to help manage a small business (a car garage), using ChatGPT or similar tools in combination with n8n (self-hosted on a VPS).
Here’s the idea:
I speak to an AI assistant (like a GPT-based chatbot), and I can say things like:
• “Add this item to the stock”
• “Remove one unit of item PRD-001”
• “How many units of this product do I have left?”
The AI should be able to understand the command, and trigger an n8n webhook that reads from or modifies the relevant database or file.
We’ve tried a few approaches so far:
• 
CSV files: GPT can read them, but writing/editing reliably has been difficult
• 
MySQL database: GPT couldn’t read from or write to the database in a structured way
• 
JSON files: this is more promising, but managing appends, updates, merges, etc., requires more logic
Here’s what we’re trying to figure out:
1. Does a ready-to-use workflow for this already exist? Has anyone built something like this?
2. What would be the best technical approach? JSON, SQLite, PostgreSQL, something else?
3. How can we make it so GPT can reliably read, write, and modify data, through API calls or n8n webhooks?
Additional info:
• We’re self-hosting n8n on a VPS
• We have access to a mounted folder (like /files) for reading/writing local files
• We can send POST requests (from GPT or elsewhere) that include natural language or structured JSON
We’d really appreciate any advice, examples, or best practices from the community.
Thanks a lot!
Jonathan