Changelog

New Features

Upload files during a session: You can now upload files at any point during a session, not just at the start. This makes it easier to bring in new data mid-analysis without restarting.

Sample sessions: When you first open Studio, you can now explore an example session while the backend loads in the background.

Fixes

• Fixed a scrolling issue where submitted answers would briefly disappear. The "scroll to bottom" button is also no longer hidden behind the input field.

• Fixed a Windows-specific issue where background processes continued running after an app update, which could prevent the app from restarting successfully.

API keys and a CLI are now generally available in Plotly Cloud. To get started, the best first stop is our documentation. You can access the API keys configuration menu in your team settings.

API keys are available in the Pro plan and above on Plotly Cloud.

Together, these features make it possible to publish apps to Plotly Cloud from CI/CD pipelines, like Github Actions. You can authenticate your usage of the Plotly Cloud CLI by logging in with your Plotly Cloud account, or, in a CI/CD pipeline, using the —api-key flag:

plotly --api-key "your_api_key_here" app publish --name my-app

We’ve also added examples to our documentation of how to configure a deployment pipeline in various CI/CD systems.

If you’re building Dash apps with AI agents, like Claude Code, the new CLI makes it easier to bring projects to production. AI agents can leverage the for both build and publish steps. A typical agent workflow looks like this:

1. Ask the agent to create a Dash app: Create a Dash app that shows a scatter plot of the iris dataset (px.data.iris()) with dropdowns to select the x and y columns.

   The agent writes the code and can run it locally:

   plotly app run

2. Once you're satisfied, ask the agent to publish: Publish this to Plotly Cloud as iris-explorer.

   The agent runs:

   plotly app publish --name iris-explorer

You can learn more about recommendations for using agents with the CLI in our documentation.

Fixes

• Fixed several issues where AI-generated components would fail to load, appear empty, or show as failures. Components that have non-fatal issues (like large figures or empty data) now save and display correctly instead of blocking the session.

• Fixed text encoding errors on Windows systems using non-UTF-8 locales that could prevent components from rendering.

• Fixed an issue where AI output cut off mid-generation could leave a component in a broken state. Studio now recovers gracefully and surfaces clear errors rather than silently failing.

• Fixed installation failures on Mac ARM (Apple Silicon) systems. Snowflake connector issues that were causing agentic sessions to fail with unclear error messages are also resolved.

• Improved automatic recovery during upgrades for users with broken or partially installed setups.

New Features

In-product Feedback (Beta): After completing an agentic session, you can now rate your experience and leave comments directly in the product. Feedback prompts appear when a session reaches a finished state or after Python code runs.

Improvements

Faster Agentic Session Startup: Agentic sessions now start faster and use fewer system resources. Sessions share a single Python environment instead of spinning up a new one each time, so you spend less time waiting and your machine stays more responsive throughout your work.

Fixes

• Fixed an issue on Windows where authentication tokens would go stale between sessions, forcing unexpected re-logins.

• Deployment error messages are now easier to read. Technical details and authorization tokens are hidden, and overly long messages are trimmed to the relevant information.

• Fixed a font rendering issue in agentic mode text areas on Windows where inputs displayed in monospace instead of the system font.

Introducing the Agentic Workflow in Plotly Studio (Beta)

Go from a question to a production-ready data app through natural conversation. The new agentic workflow in Plotly Studio is here, and it changes how you work with data.

Plotly Studio's new agentic workflow is an AI-powered environment where you can explore data, answer questions, and build shareable apps using a single universal chat interface. The agent works autonomously, writing and executing Python code behind the scenes so you stay focused on the analysis.

More flexibility. More power.

The classic Studio workflow is great for building BI dashboards, but the agentic workflow goes beyond. Now you can run ad-hoc queries, produce executive-level reports, and develop fully interactive Dash-powered data apps in a single, seamless session.

The agent can traverse your data source freely, enriching its own context across many steps to answer broad or ambiguous questions. Apps can pull from multiple queries and even disparate data sources. Tab layouts, custom styling, and interactive controls are all on the table.

Agentic workflow overview

Upload files or connect data sources. Drop in a file or connect to an external data source to get started. The agent reads, queries, and surfaces insights automatically.

Ask questions in plain language. Start a chat session and dig into your data. Ask follow-up questions without ever leaving the conversation.

Visualizations and tables are front and center. Request a Plotly chart or table and the agent writes, runs, and returns the result in a first class display.

Analysis summaries included. Every result comes with written findings and suggestions on where to take the analysis next.

Build reports and interactive data apps. Turn your session into a polished report or a fully interactive Dash app with live charts and user controls.

Publish to Plotly Cloud or Dash Enterprise. The streamlined publishing workflow you know is still here. Share your app with your team in seconds.

Agentic analytics: the next frontier

The goal for Plotly Studio has always been to be the best AI-driven tool for working with data. The agentic workflow pairs a conversational interface for casual users with deep customization for power users, bridging the gap between automated insights and bespoke data application development.

The agentic workflow is currently in Beta. Try it in Plotly Studio today and let us know what you think. Your feedback shapes what comes next.

Full documentation can be found at docs.plotly.com/studio/agentic, including a Prompting Guide to get the most out of the agent.

Fixes

Authentication Reliability: We resolved an issue where the login process occasionally required a second attempt, and we’ve made the login flow a more responsive and smoother process.

Model Response Stability: We corrected an issue for Enterprise plan users where certain high-capacity language models would return empty responses. By adjusting minimum token limits, the system now ensures consistent output generation across all supported models.

Improvements

Enhanced Response Streaming (Dash Enterprise)

We have updated the Dash Enterprise LLM connection to support response streaming. This improvement provides a more responsive experience and reduces the likelihood of gateway timeout errors, particularly for enterprise environments with strict network configurations.

Runtime Performance And Connectivity

The internal runtime has been updated to improve task scheduling and network error handling. These technical updates ensure more reliable background processing and better stability when managing complex network requests.

Fixes

macOS Dock Bug

An issue has been resolved where temporary browser processes would cause the Chrome application icon to appear unexpectedly in the macOS dock. This occurred primarily during the background generation of visualization thumbnails and has been corrected to ensure a cleaner desktop experience.