Skip to content

0.1.0 - First public release

Choose a tag to compare

View all tags
@rodrigo-arenas rodrigo-arenas released this 14 Jul 15:29
· 24 commits to main since this release
0.1.0
d5ef120

Ciaren is now available as an open-source, Python-first visual workflow platform for data and machine learning pipelines.

This release includes:

  • Visual workflow editor
  • Python execution backend
  • Extensible plugin architecture
  • Example project and getting-started documentation

Start here: Getting Started
Try the example: Demo Project & Tutorials
Report a problem: GitHub Issues

This is an early release. The most valuable feedback is whether you can install it, complete the first workflow, and understand where Ciaren fits in your stack.

What's in 0.1.0

First public release of Ciaren — a local-first visual builder for data and ML workflows that exports clean, readable pandas/polars Python. This is alpha software: APIs, data models, and generated code may still change before a stable 1.0.

Visual builder

  • 80 nodes across 9 categories — inputs, cleaning, columns, reshape, analytics, data quality, charts, machine learning, and outputs — including 8 chart nodes (bar, line, area, scatter, pie, histogram, box plot, correlation heatmap) that render instantly from a compact run artifact and support one-click PNG export.
  • File, SQL, and cloud-storage I/O (S3, GCS, Azure Blob, Snowflake) plus a built-in, read-only REST API connector (auth, custom headers, pagination, SSRF-guarded).
  • Per-node data preview, undo/redo, duplicate flows, and copy/paste nodes with their connecting edges.
  • Import options with auto-detection — CSV/TSV dialect, encoding, and multi-sheet Excel workbooks are detected on upload, with an explicit override always available.

Execution and code export

  • Dual-engine execution — polars (default) or pandas, selectable per run.
  • Exported Python reads like hand-written code (keyword assign(), chainable .loc[lambda _d: …] filters, fused method chains, reused dataframe variables on straight-line flows) and is verified equivalent to the in-app run by an automated pandas/polars, eager/lazy equivalence harness.
  • Cancel a running run from the UI or API; optional failure webhook notifications when a run fails or a schedule auto-disables.

Data quality and Machine Learning

  • Data contract nodes: assert not-null, unique, value range, row count, and arbitrary expressions.
  • Built-in ML: split, feature engineering, train, predict, and evaluate nodes, with MLflow tracking and a Models page. pip install ciaren alone gives a working ML palette (scikit-learn, MLflow, and joblib are core dependencies); XGBoost and LightGBM are available via the ml extra.

Scheduling and automation

  • Built-in cron scheduler with retries, catch-up, and auto-disable after repeated failures.
  • REST, CLI, and webhook-based ways to trigger flows outside the UI.

Plugin platform

  • A versioned, contract-first Plugin API/SDK (backend/app/plugin_api/, Apache-2.0) — plugins can contribute transformation nodes, ML model types, connectors, and schema-driven config forms without forking the core.
  • Signed .ciarenplugin packaging, local and entry-point discovery, and permission-gated model loading.
  • ciaren-client — a Python SDK for scripting against the server (flows, runs, datasets, plugins).
  • Two documented example plugins (hello-world node, MLP classifier) covering both extension paths.

Documentation and project

  • Full documentation site: getting started, guides, transformation reference, recipes, plugin authoring, and API reference.
  • Contributing guide, Developer Certificate of Origin (DCO) policy, issue and PR templates, and public GitHub Discussions.
  • Dual licensing: Ciaren Core is AGPL-3.0-only; the public Plugin API/SDK is Apache-2.0 so plugin authors can choose their own license.
  • Docker and Docker Compose support for single-container local/self-hosted use.

Security

  • pip-audit and npm audit run on every push to main and fail CI on new high/critical advisories in pinned dependencies.

Known limitations

  • Ciaren is alpha software; recommended for controlled internal workflows while the project matures, not yet for critical production jobs.
  • Not a distributed or streaming data engine — batch-style local and self-hosted workflows are the product center.
  • Has not yet completed a formal independent third-party security audit.

Full Changelog: https://github.com/ciaren-labs/Ciaren/blob/0.1.0/CHANGELOG.md

Assets 2
Loading