Skip to content

Roadmap

Near Term

  • Keep mock-backed CLI, GUI, API, Agent, and tests green.
  • Add a Docker train profile so real training dependencies stay separate from the default CI/test image.
  • Capture final GUI screenshots and CLI demo images from the themed desktop app.
  • Keep MkDocs, CI, Docker, package, and desktop build workflows green.

Model and Data

  • Publish only reviewed model metadata and weights with clear licenses and model cards.
  • Add richer dataset statistics and review workflows.
  • Expand evaluation and model comparison reports after real data is approved.
  • Add a documented path for users to import local, non-redistributed screenshots into data/raw.

Desktop and API

  • Improve desktop packaging across Linux, Windows, and macOS.
  • Keep API lightweight and local-first.
  • Avoid online inference demos through GitHub Pages.

Agent

  • Keep the default provider rule-based and API-key-free.
  • Add optional LLM providers only behind allowlisted tools and explicit user confirmation for expensive operations.
  • Add richer natural-language parsing for dataset paths, training configs, and model registry operations without granting shell access.