Chris Parmer — home

Publish Anything, Methodology, Refreshed App Design, Skills UI

Big release Plotly Studio release today, including a few features that we've been thinking about and circling around for a few months now.

Publish Anything, Publish Everything

We started Plotly Technologies because we believed it should be vastly easier to share data, graphs, and analytics (screenshots begone!).

This week's release of Plotly Studio adds a "Publish" button to every step, regardless of its "type". Studio's steps generate Python code that renders data as tables, stat cards, graphs, apps, or markdown. Previously, only the full blown Dash app steps could be published.

Now every step's artifacts can be shared with a click of a button - no fussing with deployment scripts, it just works.

Publishing is nicely integrated into Plotly Cloud, our managed service at cloud.plotly.com, and Dash Enterprise, our self-hosted platform built for internal deployments at F500 companies.

Screenshot of Plotly Studio and a browser tab that shows the published table from a Studio session in the web browser.
Sharing a table from a Plotly Studio session to Plotly Cloud or Dash Enterprise with a click of a button, and the link to the artifact in the browser on the right.
Screenshot of a graph in Studio that is published, showing the 'Publish' popover that appears for each artifact.
Share multiple artifacts in a single session - here's a map of those buoy's we listed above.
Screenshot of a dash app rendered in Plotly Studio that is also published
A single session can evolve from a table to a single chart to a full-blown interactive data app - "from exploration to application". And you can publish at every step along the way.
A few steps of a Plotly Studio session and each step has a publish button
Sessions work through analytics problems one step at a time in an agentic loop. Unlike other AI tools, we're designing Studio so that the data is front and center so you can trace back and verify of step of the process. By making each step publishable, you can now easily verify this work collaboratively by sharing any step you might want second eyes on.
A writeup in Plotly Studio
This is a "write up" in Plotly Studio and - like every step - it's independently publishable. Write ups are Studio's version of a report. What makes them unique is that they are computational, meaning that every single number in the report was computed not inferred. This report is backed by code - it's a Python function that returns a markdown string with templated numbers - so you can verify and track the provenance of every number (unlike practically every other hallucination-prone AI tool out there).

Methodology

In this release, every step includes a "Methodology" section that breaks down the analytics approach of the step: the data source, the columns, the formulas, the assumptions, etc.

Trust and verification are among the most important problems to solve in AI-assisted analytics. Methodology is one of many features we've designed to address these problems.

I wrote about this at length in a previous post: https://chris-parmer.com/methodology/

Here are a few screenshots of what Methodology looks like in today's release.

Screenshot of the methodology step in Studio
Methodology surfaces all of the important details and assumptions about the analysis every step of the way
Another screenshot of methodology summary in studio
Methodology breaks down the approach and the assumptions so you can verify the approach and steer accordingly.

Refreshed App Design

We tweaked the look and feel of the default Dash apps in Studio and they're looking very clean. Many AI website builders make apps that look all the same these days (purple gradients, etc) that are steered more towards marketing pages than rigorous data apps. We've been designing our apps to have a high data-to-ink ratio and let what's important shine: the data visualization.

A clean, modern Dash app that shows a map of SF buoys on the left and time series charts on the right showing wave height and wave speed on the right.
Very clean, modern app design with a high data-to-ink ratio that puts the data front-and-center.

Skills UI

Skills are slash commands - shortcuts to run predefined prompts. We ship 9 skills out-of-the-box for common data workflows and allow you to write your own.

We updated the Skills UI in this release with some nice quality of life features.

Skills are still available in the header:

Skills UI showing 9 skills
9 skills that are included in Plotly Studio by default.

But now they also appear in your prompt with nice typeahead and search:

Skills typeahead
Nice typeahead search for Skills in the latest Plotly Studio release.

When you include a skill in your prompt, we show a nice little badge that indicates that Studio recognized it. This is such a nice touch to the UI - a lot of slash command UIs out there (I'm looking at you, Slack) really leave you hanging when you write out a command, giving you no indication that it's a "recognized" command until your press enter.

We also display the underlying skill content when you hover over the skill, so you can see the exact prompt. A lot of Skill UIs keep the skill content a secret and we think that's misguided; skills are meant to be examined, copied, forked, and remixed. The best part about seeing the skill content right up front is that you can always just copy and paste the content of the skill into your prompt and edit it directly; there's no need to dig into the internals.

Skills UI showing the description of the skill.
Hover over the skill in the prompt and you can see the full skill text.

That's a wrap for this week

Give these features a try and let us know what you think. You can get started for free here: https://plotly.com/studio.