Datus-ai/Datus-agent

What it solves

Datus is an open-source data engineering agent designed to bridge the gap between data warehouses, catalogs, semantic layers, and BI tools. It automates the tedious parts of the modern data stack, such as writing and validating SQL, constructing semantic models and metrics, and generating data pipelines, reports, and dashboards, while allowing teams to maintain ownership of their context.

How it works

At its core, Datus uses a context engine that gathers database schemas, reference SQL, and business rules. It employs a system of subagents—which package specific tools and rules for a business domain—and skills to execute tasks. The agent connects to various systems (LLMs, warehouses, schedulers, and BI tools) via adapters and plugins. Every correction and run is fed back into the context engine, meaning the agent becomes more accurate over time as it learns from user feedback.

Who it’s for

• Data Engineers: Who use the CLI to explore data, build assets, and manage the context engine.
• Data Analysts: Who use the chat interface (web, Slack, Feishu, or VS Code) to ask business questions and generate reports.
• Developers: Who want to integrate data services into other applications via REST API or MCP (Model Context Protocol).

Highlights

• Automated Semantic Modeling: Generates OSI semantic models and metric definitions from database schemas and SQL history without manual YAML writing.
• Dosi Execution Engine: Compiles semantic models into SQL for over 13 database dialects.
• Contextual Learning: A context engine that stores metadata and corrections in LanceDB, SQLite, or PostgreSQL to improve future outputs.
• Multi-Interface Access: Available via CLI, Web Chatbot, REST API, MCP server, and VS Code extension.
• Enterprise Governance: Includes statement-level SQL authorization, tiered permissions, and OS-level sandboxing for bash commands.