Getting started

This guide takes you from zero to a working Zingle workspace with your first AI-generated data model shipped as a pull request. Expect about 10 minutes for the full setup.

---

Prerequisites

Before you begin, make sure you have:

Requirement	Details
Zingle account	Workspace access with sign-in credentials
GitHub repository	A repo that contains (or will contain) your dbt-style models
GitHub personal access token	With repo scope for PR creation
Snowflake account	Account identifier, credentials, role, warehouse, database, and schema
Airflow instance (optional)	Base URL and credentials for orchestration

---

Step 1 — Connect your integrations

1. Connect a GitHub repository

   Navigate to Connections in the sidebar. Click Connect Repository and provide your repository URL or org/repo name.

   Zingle validates the repository exists, checks the default branch, and begins scanning for models, tests, macros, and semantic files.

   Once connected, you'll see sync status (pending, in progress, success) and high-level stats for your repo.

2. Connect Snowflake

   In the same Connections page, scroll to the Snowflake section and click Connect Snowflake.

   Enter your account identifier, username, password, role, warehouse, database, and schema. Zingle runs a validation query with a 30-second timeout to confirm connectivity.

   tip

   Copy your account identifier from Snowflake → Admin → Account Details. Do not include .snowflakecomputing.com.

3. Connect Airflow (optional)

   If you use Airflow for orchestration, click Connect Airflow and provide the base URL and credentials.

   Once connected, you can wire models to Airflow DAGs and deep-link from Zingle directly into the Airflow UI.

---

Step 2 — Configure your workspace

Open Settings from the sidebar. You'll find three configuration areas:

1. YAML config — workspace-level dbt or project configuration (materializations, model paths, etc.)
2. Modeling guidelines — naming conventions, layer rules (e.g., "no joins in raw/bronze"), performance practices
3. Custom instructions — fine-tune how the AI assistant writes SQL, PR titles, docstrings, and handles PII

These settings directly influence how Zingle's AI generates schemas, SQL, and semantic definitions. Take a few minutes to configure them upfront — it pays dividends in output quality.

---

Step 3 — Create your first data pipeline

1. Open the AI workspace

   Navigate to Data Pipelines in the sidebar and click Create.

   This opens the AI modeling workspace, which includes:

   • A lineage canvas showing Bronze → Silver → Intermediate → Gold → Semantic layers
   • A bottom pane with tabs for Schema Details, SQL Query, Tests, Schedule, and PRs
   • The Unified Chat Assistant on the right side

2. Describe your model

   In the chat assistant, describe the table you want to build. You can:

   • Write a natural language description (e.g., "I need a customer lifetime value table that joins orders with customer demographics")
   • Paste existing SQL for Zingle to refactor
   • Upload a notebook file

   Zingle analyzes your input and proposes schemas for intermediate (int_) and gold layer tables.

3. Review schemas and SQL

   Use the lineage canvas to select each proposed table. In the bottom pane:

   • Review the Schema Details tab for column definitions
   • Review the SQL Query tab with syntax highlighting and diff view
   • Accept or request changes for each table

4. Configure tests and schedule

   For each table, configure:

   • Data quality tests — not-null, unique, accepted values, and custom tests
   • Compute engine — choose between Snowflake sizes or managed DuckDB
   • Schedule — set execution frequency and timing

5. Raise a pull request

   Once all quality gates pass (schemas accepted, SQL accepted, tests configured, compute set, schedule defined), the Review changes and raise PR button becomes active.

   Click it to review a summary of all changes, then create the PR. Zingle opens a branch, commits SQL files, YAML configs, test definitions, and semantic layer changes to your GitHub repo.

---

What happens next

Your pull request is now in GitHub, ready for code review by your team. The PR includes:

• SQL model files for each layer (intermediate, gold)
• dbt-compatible YAML configuration
• Test definitions
• Semantic layer YAML (if applicable)
• Compute engine and schedule metadata

Once merged, your pipeline is in production — governed, tested, and version-controlled.

---

Recommended reading

Topic	Link
Deep dive into Data Pipelines	Data Pipelines
Explore data with natural language	Playground
Set up Snowflake	Snowflake Integration
Configure workspace rules	Settings
Common issues	Troubleshooting