dataenginex 0.4.0

DataEngineX — open-source, self-hosted, local-first Data + ML + AI workbench library

dataenginex

The Python library that powers DEX Studio — an open-source, self-hosted, local-first Data + ML + AI workbench. Use the library directly when you want code, not a UI.

Pre-1.0 status. 0.4.0 is honest about that. See the scope reset CHANGELOG for the rationale.

Install

pip install dataenginex                    # lean base — DuckDB, structlog, Pydantic, Click, pyarrow

Optional integrations — install only what you need:

pip install 'dataenginex[postgres]'        # asyncpg-backed lineage, persistence
pip install 'dataenginex[qdrant]'          # Qdrant vector store backend
pip install 'dataenginex[queue]'           # ARQ async job queue (pulls redis)
pip install 'dataenginex[cloud]'           # S3, GCS, BigQuery storage backends
pip install 'dataenginex[ml]'              # scikit-learn, xgboost, sentence-transformers
pip install 'dataenginex[tracking]'        # MLflow integration
pip install 'dataenginex[data]'            # PySpark, databricks-cli

LiteLLM: install separately — it pins python-dotenv==1.0.1 which conflicts with our >=1.2.2:

pip install 'litellm>=1.83.3' --no-deps

Quick start

from pathlib import Path
from dataenginex.engine import DexEngine

# Load config and initialize all backends
engine = DexEngine(Path("dex.yaml"))

# Data — run pipelines defined in dex.yaml
engine.run_pipeline("clean_users")

# ML — train, register, predict
models = engine.model_registry.list_models()
result = engine.model_registry.predict("churn_model", features)

# AI — chat with an agent over your data
response = engine.agents["assistant"].chat("summarise the latest pipeline run")

# Persistence — query DuckDB-backed history
runs = engine.store.list_pipeline_runs(limit=10)

Smaller surfaces — use only what you need:

from dataenginex.config import load_config
cfg = load_config("dex.yaml")

from dataenginex.core.interfaces import BaseConnector
from dataenginex.core.registry import BackendRegistry

from dataenginex.ml import ModelRegistry
from dataenginex.ai.llm import get_llm_provider
from dataenginex.ai.vectorstore import VectorStoreBackend

Submodules

Module	Description
dataenginex.engine	DexEngine — single entry point; loads config, inits store, wires backends
dataenginex.store	DexStore — DuckDB-backed persistence (.dex/store.duckdb)
dataenginex.config	dex.yaml schema, loader, env-var resolution
dataenginex.core	Exceptions, Base* ABCs, BackendRegistry
dataenginex.cli	dex CLI (validate, version, init)
dataenginex.data	Connectors (CSV, Parquet, DuckDB, HTTP, …), pipeline runner, schema registry
dataenginex.ml	Classical ML — training, model registry, serving, drift
dataenginex.ai	LLM providers, agents, RAG, vector store, memory, observability
dataenginex.orchestration	Scheduler, background workers
dataenginex.middleware	structlog config, Prometheus metrics
dataenginex.lakehouse	Storage backends, catalog, partitioning
dataenginex.warehouse	Transforms, lineage tracking
dataenginex.secops	PII detection, masking, audit logging
dataenginex.api	Pydantic response models (no HTTP server bundled)
dataenginex.plugins	Entry-point plugin discovery

Want the UI?

dataenginex is the engine. The web UI lives in a separate repo:

git clone https://github.com/TheDataEngineX/dex-studio && cd dex-studio
docker compose up         # open http://localhost:7860

DEX Studio imports dataenginex directly — no separate API server.

Links

• Source: github.com/TheDataEngineX/dex
• Docs: docs.thedataenginex.org
• Roadmap: docs/docs/roadmap/DESIGN-2026.md
• ADRs: docs/adr/
• Issues: github.com/TheDataEngineX/dex/issues
• License: MIT