HTML widgets represent a fundamental shift in how you can customize and extend the OpenBB Workspace. Instead of being limited to our pre-built widget types, this new architecture allows institutional clients to integrate fully interactive, sandboxed tools using any web technology they prefer.
We are effectively shipping two distinct capabilities for OpenBB Workspace users:Enterprise/On-Prem: Full JavaScript execution enabled, allowing for complete interactive capabilities (buttons, forms, real-time data).
Cloud (pro.openbb.co): HTML and CSS only, focused on secure, high-fidelity visualization without script execution.
Unify your fragmented stack without “porting” code
In most financial firms, the workflow is fragmented. An analyst might have a terminal open for market data, a web browser for internal risk models, a Jupyter notebook for backtesting, and a separate window for AI chat. This constant context switching kills productivity and fractures insights.
HTML widgets solve this by allowing you to merge your proprietary, in-house tools directly into the OpenBB Workspace.
Financial professionals already have tools they’ve built. Quant teams have sophisticated D3 risk visualizations; developers maintain internal web components; data scientists have Plotly dashboards. Until now, bringing those assets into a unified workspace meant rebuilding them from scratch to fit a vendor's proprietary system.
By treating your internal applications as widgets, they live side-by-side with our advanced pre-built visualization tools. More importantly, they exist within the same context as the OpenBB Copilot or any other AI agent your firm would like to integrate. This means an analyst can view your firm’s custom risk model, compare it against market data, and ask an AI agent to synthesize findings across both, all in a single pane of glass.
We are removing the barrier between "buy" and "build." You no longer have to choose between a vendor platform and your own custom stack; you get the efficiency of a unified workspace without sacrificing the specificity of your internal tools.
Two Deployment Models
Enterprise/On-prem
For clients running their own instances, we enable the full suite of capabilities. This allows your developers to build widgets that act as complete operational tools, submitting orders, updating internal databases, or running real-time simulations.
Cloud
On pro.openbb.co, we disable JavaScript execution. This might sound restrictive, but it remains incredibly powerful. You can still:
- Export visualizations from Jupyter notebooks as static HTML.
- Build complex layouts that our standard grid system doesn't support.
- Apply custom CSS for branded reports and dashboards that match your firm's identity (with white labeled solutions available).
Two approaches to integrate with AI
Getting an LLM to "see" what is inside a custom HTML widget is difficult because raw HTML is often too noisy for AI to parse effectively. We have designed two specific integration patterns to solve this.
See this in action with a custom crypto order book widget below. Notice how the Copilot seamlessly interprets the market depth data to answer natural language questions.
1. The Raw Endpoint (recommended)
This approach is similar to the mechanism we allow for Plotly charts.
You implement a raw=true parameter on your widget's backend endpoint. When the AI is invoked, we don't scrape the HTML; instead, we hit your endpoint with that parameter, and you return a structured JSON response.
This is the best solution to get clean separation of concerns, one view for the human (HTML), one for the machine (JSON).
2. Direct HTML Parsing
Alternatively, if your widget doesn't implement a raw endpoint, the AI will parse the HTML structure directly.
While this works, structured endpoints provide cleaner results. LLMs can struggle with nested HTML, so we recommend implementing the raw endpoint when accuracy is critical for your use case.
Real-World Scenarios
We are already seeing how this flexibility changes workflows:
Quant Teams: A team that spent months building a D3.js risk analysis tool was told they would need to rebuild it to use our platform. With HTML widgets, they simply dropped their existing files into the widget, and it worked immediately.
Data Science: Teams using Plotly Dash for internal tools can now export specific components as standalone HTML and embed them into shared dashboards.
Custom Branding: A firm needed reporting dashboards that strictly adhered to their brand fonts and colors. HTML widgets gave them the freedom to apply custom CSS and achieve pixel-perfect control over every visual element.
HTML widgets are about removing friction. Instead of forcing you to build our way; we're providing the foundation for you to keep yours.
Want to learn more about running custom widgets in a secure enterprise environment? Contact our team to discuss deployment options or schedule a demo.