Skip to content

v1.0.0

Choose a tag to compare

View all tags
@philippjfr philippjfr released this 28 Jan 14:05
· 160 commits to main since this release
v1.0.0
3991574

Lumen AI 1.0 is a major milestone that focuses on stability, clarity, and long-term extensibility. This release significantly reworks both the user interface and the core execution model, while improving performance and real-world applicability.

UI & UX

  • Complete UI rewrite on panel-material-ui
    • Cleaner, more consistent visual language
    • Better layout primitives for complex analytical outputs
    • Reduced visual noise while preserving inspectability
  • Improved default presentation
    • Key artifacts like SQL and Vega-Lite specs remain inspectable
    • Internal reasoning and chain-of-thought are hidden by default

Core Architecture

  • New context-driven API
    • Removed global mutable memory
    • Agents and tools now communicate via explicit, typed context
    • Inputs and outputs defined using Pydantic models
  • Typed, auditable execution model
    • LLM-generated plans are executed as structured tasks
    • Deterministic and LLM-driven steps can be mixed safely

Reporting & Execution

  • New task execution architecture

    • Foundation for exportable, repeatable reports
    • Enables multi-step workflows that may or may not invoke an LLM
    • Basis for upcoming reporting and application features

Data & Model Support

  • Expanded database support
    • Broader SQL coverage via SQLAlchemy sources
    • Improved handling of production databases and warehouses
  • Improved LLM integration
    • Support for more providers and configurations
    • Better defaults favoring low-latency, reliable models
  • More robust behavior
    • Improved performance, stability, and error handling

Lumen 1.0 is optimized for fast, structured insights without writing code, while establishing the foundation for turning those explorations into reusable reports and data applications in future releases.

Assets 2
Loading