Weights Studio UI

Weights Studio is the visual frontend for WeightsLab experiments. It ships inside the Python package. Running weightslab start serves the bundled SPA and proxies gRPC-Web to your training backend, all from one Python process.

Architecture

Runtime path:

1. Browser (served from weightslab start)

2. weightslab start — pure-Python HTTP server that:

   • Serves the pre-built Weights Studio SPA (vendored in weightslab/ui/static/)

   • Translates gRPC-Web (browser) to raw gRPC (backend) via an embedded proxy

3. WeightsLab Python gRPC service (started by wl.serve())

Quick start

1. Install WeightsLab:

   pip install weightslab
   

2. In your training script, start the backend:

   import weightslab as wl
   wl.serve(serving_grpc=True)
   # ... training loop ...
   wl.keep_serving()
   

3. In another terminal, start the UI:

   weightslab start
   

4. Open the URL printed by weightslab start in your browser.

The UI auto-discovers the backend on localhost:50051 (default). Pass --backend-port to override:

weightslab start --backend-port 50052

To suppress auto-opening the browser:

weightslab start --no-browser

Sections

Everything the studio puts on screen, what it is for, and how to drive it — grouped into four parts. Each page below covers its own subsections; see the page itself for the details.

• Landing page
  • Report Generation
  • Integrated Notebooks
• Agent
  • Agent Window
  • Commands
  • Setting it up
  • History behavior
• Left panel
  • Runs Management
  • Hyperparameters modification in-training
  • Painting mode for tag
  • Metadata Sorting / Hist. Generation
  • Data actions
• Main area
  • Plots Board
  • Data Board
• More
  • Setup and access
  • Operating it

• Landing page — the pre-experiment surface: agent chat, local Jupyter, Colab quickstarts, report generation, and the Integrated Notebooks.

• Agent — the docked chat bar and Agent Window: commands, /loop jobs, setup, and history behavior.

• Left panel — run management (training controls, evaluation, mode, auto-refresh), in-training hyperparameter edits, tag painter mode, metadata sorting/histograms, and data actions (save, export).

• Main area — the Plots Board (search, merged curves, error bands, right-click actions, resource monitoring) and the Data Board (grid/list modes, quick filters, selection, tagging, the detail modal).

• More — More to know.