Build Python data apps with RAG and Dash

This video walks through major updates to both Dash Enterprise and open-source Dash, focusing on performance, observability, and ease of building AI-powered data applications. It also includes a detailed walkthrough of building a production-ready retrieval-augmented generation (RAG) app using Dash. Here’s what you’ll learn within the video:

Dash Enterprise now supports virtual machines (VMs) that can be started, stopped, and parked to better control resource usage. Additionally, a new live monitoring console offers real-time visibility into VM status and running services.

• Start/stop/park VMs for better cost and performance control
• Live monitoring of VM usage and services
• Identify bottlenecks and free up resources easily
• Observability now built into the platform
• Expanded data access + credential storage

Support for new cloud data sources and custom credential management makes it easier to connect Dash apps to enterprise data.

• Native support for Azure Data Lake and AWS Redshift
• Custom data source credential storage for flexibility
• Securely connect to any external database or API
• Simplified authentication for non-standard sources

Next, we explore Plotly graphing library and Dash framework updates for more speed and flexibility. Several major updates improve performance and compatibility on the open-source side of Dash.

• Universal DataFrame support: pandas, polars, and others
• Enhanced typed arrays in JavaScript boost performance
• Full Jupyter Notebook and JupyterLab widget integration
• Rewritten static image generation backend

Plotly AI Chat is now integrated into Dash Enterprise to assist with data app development with AI. It processes the structure and statistics of your data and returns executable pandas + Plotly code that fits into your Dash apps. It supports a range of use cases, including documents processed with vector embeddings and LLMs.

• Uses OpenAI's GPT models via metadata-aware prompts
• Drag-and-drop documents into apps for vector embedding
• Ask high-level questions; get back working Python code
• Flexible enough to handle non-tabular sources like PDFs

We then showcase a live example of using Dash for RAG apps. The demo walks through a Dash app built to analyze a poem using a RAG model. The same pattern could be used for clinical trial documents, financial filings, or research papers.

• Uses vector embeddings + LLM (via LangChain) to query documents
• Full frontend customizability with Dash components
• Backend integrates cleanly with OpenAI, Anthropic, Mistral, etc.
• Easy prompt tweaking and agent management

Dash Enterprise simplifies deploying and scaling AI apps. A single command deploys the app, and environment variables are easy to manage. Stateless architecture allows effortless horizontal scaling.

• One-line deployment to production
• Smooth integration of environment configs
• Add containers and replicas easily to handle more users
• No special setup needed for AI workflows

The latest major release of Dash brings performance upgrades, better dev tooling, and a cleaner codebase.

• New DevTools UI
• React 18 is now the default
• Better prop typing for component authors
• Faster rendering with large component trees

Watch the video to follow along with the demos.