Query Google Sheets/CSV Data via AI Agent using PostgreSQL

This repository contains an n8n workflow that demonstrates how to integrate Google Sheets or CSV data into a PostgreSQL database and query it using natural language via an AI agent. The workflow orchestrates data ingestion, persistence and an AI-powered query layer to make your tabular data accessible through human friendly questions.

Features

• Data Ingestion: Extracts rows from a Google Sheets spreadsheet or uploads a CSV file and loads the data into a PostgreSQL table.
• AI Powered Queries: Exposes an HTTP endpoint where you can ask questions in plain English. An AI agent translates the question into SQL, runs it against PostgreSQL and returns a concise answer.
• Extensible Workflow: Uses modular nodes for ingestion, SQL execution and AI summarisation so you can adapt the flow to your own data sources or prompt templates.
• Credential Management: Uses n8n’s credential system to store Google and database credentials securely.

Getting Started

1. Prerequisites

   • An n8n instance (self‑hosted or cloud).
   • A PostgreSQL database you control.
   • Google service account credentials with access to the spreadsheet (if using Google Sheets).
   • An API key for your AI provider (e.g. OpenAI) to translate questions into SQL.

2. Import the Workflow

   • Download the Query Google SheetsCSV data through an AI Agent using PostgreSQL.json file in this repository.
   • In the n8n UI, click Import workflow and upload the JSON file.

3. Configure Credentials

   • Set up a new PostgreSQL credential in n8n and point it at your database.
   • Create Google Sheets credentials if you plan to pull data from Sheets.
   • Add your AI provider API key in the corresponding credential node.

4. Ingest Data

   • If using Google Sheets, specify the spreadsheet ID and worksheet name in the relevant node.
   • To load a CSV file, place it where n8n can access it (e.g. via the File Trigger node).
   • Run the ingestion nodes to populate your Postgres table.

5. Query via AI

   • Send a POST request to the HTTP endpoint exposed by the workflow with a JSON body containing your natural language question.
   • The AI agent will translate the question, execute the generated SQL and return the answer.

Customisation

• Change the prompt used by the AI agent to influence the SQL generation.
• Modify the database schema or add indexes to optimise performance for your queries.
• Add additional data sources (e.g. API responses) by connecting new nodes upstream of the Postgres insert.

License

This project is licensed under the MIT License. See the LICENSE file for details.