·
42 commits
to main
since this release
Immutable
release. Only release title and notes can be modified.
First public release of Data Workers Community Edition — 10 autonomous AI agents for data engineering via MCP.
Highlights
- 10 MCP agents: pipelines, incidents, catalog, schema, quality, governance, observability, orchestration, ML/models, and connectors
- 160+ MCP tools exposed via the Model Context Protocol for LLM-driven data engineering
- 49 connectors: 14 catalog connectors + 35 enterprise connectors (Snowflake, Databricks, BigQuery, dbt, Airflow, and more)
- 9 infrastructure backends with in-memory stubs that auto-upgrade to real adapters (Redis, Neo4j, Postgres, etc.) when env vars are set
- 3,000+ passing tests across 149+ test files
- Apache Iceberg native: REST Catalog connector, Polaris connector, MERGE INTO generation, schema evolution from snapshots
- Hybrid search engine: vector + BM25 + graph with Reciprocal Rank Fusion reranking in the catalog agent
- ML lifecycle: experiment tracking, model registry, drift detection, SHAP explainability, A/B testing
- Enterprise-ready foundation: OAuth 2.1 middleware, SHA-256 hash-chain audit trail, PII scanning, license-tier tool gating
Quick Start
```bash
npm install dw-claw
```
See the README for full setup instructions, architecture overview, and agent documentation.
Known Issues
See MCP Startup Bug Report for 6 documented root causes and fixes related to local MCP agent startup.
Built with the community in mind. Contributions welcome — see CONTRIBUTING.md.