Zingle AI — Transfer your business knowledge to your data platform's AI (Databricks, Snowflake, BigQuery and Redshift).

AI assistants in your data platforms can write SQL — but they don't understand your unique schema and business. Zingle helps you build context and transfer it to your data platform, enabling AI chatbots to work with very high accuracy.

Zingle AI is an open-source VS Code/Cursor extension (by Zingle AI Lab) that scans your application’s source codebase (services, business logics, etc.) to automatically generate business-ready context for your database tables and columns. Add this real business context to your data platforms — so text-to-SQL tools and AI assistants for data platforms (Snowflake, Databricks, BigQuery, dbt, etc.) produce accurate, trustworthy answers.

How to Use the Zingle AI Extension

Follow these four simple steps to generate documentation from your source codebase:

1. Open the Application Service Repository

Open the application’s source code repository that contains the services and business logic writing directly to your database tables.

2. Open the Zingle Extension

Inside VS Code/Cursor, open the Zingle AI sidebar extension.

3. Generate Documentation for a Table

Select database type and table name, then click Generate Documentation.

4. View the Generated Documentation

Select your table to see automatically generated, business-ready documentation.

Why this matters

Schemas rarely capture real business meaning. That’s why AI answers are often ambiguous.

Example

• Column: fare
• Typical doc: “The fare amount.”
• Result: Confusion (Does it include surge? Tips? Tolls? What about cancelled rides?)

With Zingle AI

• Column: fare
• Generated doc:
  “Final computed fare after ride completion. Calculated by the fare engine using distance, duration, base fare, surge multipliers, and tolls. Excludes tips and cancellation fees. Null value when ride not completed.”
• Result: Clear logic → trustworthy AI answers

Example

• Column: status
• Typical doc: “Current ride status.”
• Result: Confusion, what’s the difference between NEW and UPCOMING?

With Zingle AI:

• Column: status
• Generated doc: “Current state of the ride in its lifecycle. It is updated during ride state transitions. Exception values include NEW (immediate rides ready to be started by driver), UPCOMING (scheduled rides waiting for start time), INPROGRESS (ride currently active with driver en route or at destination), COMPLETED (ride finished successfully), and CANCELLED (ride terminated before completion by driver, dashboard, or system).”
• Result: Ambiguity Clarification → AI understand business processes

Example

• Column: END_OTP
• Typical doc: “One-time password used to end ride.”
• Result: Confusion ( When is it null? Which ride categories require it?)

With Zingle AI:

• Column: END_OTP
• Generated doc: “One-time password that is required to complete a ride. It is generated at ride start when trip category requires OTP verification. Null value means trip category (rental, inter-city, delivery rides) doesn’t require OTP verification, ride has No Otp enabled, or ride hasn’t started yet.”
• Result: Business Process Understanding → AI understands Business Intent Of Your Questions

Example

• Column: TOLL_CONFIDENCE
• Typical doc: “...”
• Result: Less used column, what’s the meaning of it?

With Zingle AI:

• Column: TOLL_CONFIDENCE
• Generated doc: "Indicates the accuracy of toll charge estimation for a ride route. It is populated when toll charges are calculated during ride completion processing. The value is derived from toll calculation algorithms that assess route accuracy and toll detection reliability. Null value means toll calculation is not performed, route does not include toll roads etc."
• Result: Understanding of logics buried in your codebases → Improve every part of your product

Zingle AI is designed to work with your application’s codebase (services, business logic) — not data pipeline repos.

Why extract from your source codebase?

For database tables created and used by your backend services (internal software), your application’s source code contains the true business meaning of your tables and columns — how values are generated, updated, and tied to real business processes.

Your data pipelines (dbt, Spark, Airflow, etc.) transform and propagate data. AI chatbots built into your data platforms already have access to those pipeline definitions, but they lack access to your application codebase, where the real logic is defined.

When you use Zingle AI to extract context from the source system codebase and add it to your warehouse’s first layer of tables (often called ‘raw’ or ‘base’ tables), this context naturally flows downstream. 90%+ of columns in downstream layers are propagated from upstream. Once AI understands the meaning of your warehouse’s first layer of tables, it also interprets downstream columns correctly

💡 One-time effort, lifetime value
Extracting context from your codebase is largely a one-time process. Once added at the source, it will help AI understand source and downstream models.

How to add this context to your data platform

👉 Visit our Zingle Docs guide to learn how to add this context to your data platform.

Examples

📖 See real samples of Zingle-generated context 🤖 See How AI chatbot works with vs without Zingle Context

What Zingle AI does

Zingle scans your application codebase and generates one Excel file for every table.
Each file contains:

• Table-level description – what the table represents in your business process
• Column-level documentation – meaning, calculation logic, enum codes explained

👉 You then upload these Excel files into your data platform (Snowflake, Databricks, BigQuery, dbt, etc.) by following our Zingle Docs guide.

Result: AI copilots inside Snowflake, Databricks, dbt, BigQuery start giving clear, accurate, business-aware answers.

Requirements

• VS Code with an internet connection
• Node.js ≥ 18 installed
• Windows users: enable WSL

Privacy & Security

• Your code stays local — source and generated docs remain on your machine
• No external training — your code is never used to train models
• Local processing — documentation generation runs locally

Support

• Slack: (Recommended) Join the community — https://join.slack.com/t/zingleai/shared_invite/zt-3c3rp1vpg-6BXpK2eTdwUNogugSDcM~Q
• WhatsApp: Instant help — https://wa.me/919131266517?text=VS-Code%20Extension
• Live Demo: Book a 30-min call — https://calendly.com/founder-zingle/30min
• Email: support@getzingle.com

Try it on one table in your codebase — you’ll see the difference immediately.” ⭐ leave a review:
https://marketplace.visualstudio.com/items?itemName=ZingleAI.zingleai-codebase2docs&ssr=false#review-details

License

MIT