det (Data Engineering Tool)

det is a lightweight, local-first SQL ETL orchestration framework and data transformation workspace powered by DuckDB 2.0.

det offers data engineers the familiar ergonomics of declarative SQL modeling—automatic dependency resolution, {{ ref() }} and {{ source() }} lineage, and schema testing—combined with an embedded vectorized analytical engine and a built-in zero-dependency Web IDE.

🚀 Key Highlights & Capabilities

• ⚡ Embedded Vectorized Execution (DuckDB 2.0 & Arrow):
  • High-throughput analytical SQL execution directly in-process.
  • Zero cloud serialization overhead, SIMD parallelism, and multi-threaded query execution for local and on-premises workloads.
• 🔌 Unified Multi-Tier Catalog:
  • Relational Databases: PostgreSQL, MySQL, SQLite, Microsoft SQL Server / Fabric SQL endpoints.
  • Open Lakehouse Formats: Delta Lake (read-only source scanner with time-travel snapshot querying), Apache Iceberg, Microsoft Fabric OneLake (abfss://).
  • Object Storage & Flat Files: AWS S3, Cloudflare R2, Parquet, CSV, JSON, and Excel (.xlsx).
• 🧩 Declarative SQL Modeling & DAG Resolution:
  • {{ ref('model_name') }}: Declare model lineage and automatically generate topological execution DAGs.
  • {{ source('connection', 'target') }}: Read directly from any catalog connection using vectorized scanners.
  • {{ env_var('VAR_NAME', 'default_value') }}: Substitute environment tokens at compilation time.
• 🛡️ Resilient Data Quality & Row Quarantine:
  • Declarative assertions: not_null, unique, accepted_values, expression, relationships (referential integrity), row_count, and custom_sql.
  • Non-blocking row quarantine: Isolate violating data rows into persistent quarantine sinks for audit and remediation without crashing downstream pipeline runs.
• 📊 Granular Monitoring & Pluggable Telemetry:
  • Track individual model execution stages: status, start timestamps, millisecond durations, rows processed, and error diagnostics.
  • Swappable storage sinks: Local DuckDB (.det/telemetry.duckdb), append-only flat-file JSONL (.det/telemetry/), or external database audit tables.
  • One-click run export to JSON for external log ingestion and auditing.
  • Built-in performance guard capping log queries to keep UI navigation responsive.
• 📚 Self-Contained Interactive Documentation:
  • Automatic column and schema introspection with live documentation editing.
  • Interactive Mermaid DAG lineage diagrams.
  • Static self-contained HTML portal export via det docs generate.
• 💻 Dual Interface:
  • Full headless CLI and Python SDK for headless automation and CI/CD.
  • Built-in, zero-dependency browser Web IDE (det gui).

🛠️ CLI Quickstart

1. Initialize a Workspace

det init my_pipeline
cd my_pipeline

2. Verify Catalog Connections

det catalog list
det catalog test

3. Execute Pipeline Models

det run

4. Run Data Quality Assertions

det test

5. Generate and Serve Documentation

det docs generate

6. Launch the Local Web IDE

det gui --port 8080

Open http://localhost:8080 to explore the visual DAG canvas, models, documentation portal, and execution telemetry.

🐍 Python SDK Example

from pathlib import Path
from det import Workspace

# Open active workspace
ws = Workspace(Path("."))

# Test catalog connections
health = ws.test_catalog()
print("Catalog Health:", health)

# Execute topological pipeline DAG
report = ws.run()
print(f"Executed {report.total_models} models in {report.duration_ms:.1f}ms")

# Run data quality tests with automatic quarantine routing
test_suite = ws.test()
print(f"Passed: {test_suite.passed_count}, Failed: {test_suite.failed_count}")

# Generate interactive HTML documentation
docs_path = ws.generate_docs()
print("Documentation portal generated at:", docs_path)

🔮 Roadmap & Upcoming Features

We are actively developing the following enhancements:

• ✍️ Intelligent SQL Editor Autocompletion & Language Server (LSP):
  • Context-Aware {{ ref() }} Completion: Instant autocomplete suggestions for all model names defined across active pipelines.
  • Catalog Source Suggestion: Live autocompletion for {{ source('connection', 'table') }} referencing catalog items and mounted views.
  • Column-Level Suggestions: Schema-aware column autocompletion introspected from upstream models and catalog connections.
  • In-Browser Syntax Validation: Real-time SQL syntax highlighting, linting, and error hints within the Web IDE.
• 📦 Pluggable Extension Ecosystem: Dynamic third-party connector packaging and community plugins.
• ⚡ Streaming Ingestion Adapters: Incremental real-time micro-batch processing with DuckDB append-only appenders.

🧪 Running Tests

det includes a comprehensive automated test suite covering catalog mounting, DAG parsing, topological execution, quarantine routing, documentation extraction, and telemetry sinks:

pytest

108 passing unit and integration tests.

📦 Building Distribution Packages

python -m build
python -m twine check dist/*

📄 License & Attribution

Maintained by Saurabh Rai (saurabhr2@gmail.com) and det-Labs.

Licensed under the MIT License.