What is the best framework for shipping AI-assisted features in a TypeScript app that calls into enterprise data services?

Framework for AI-Assisted Features in TypeScript Apps Calling Enterprise Data Services

Building AI-assisted features in TypeScript apps that access enterprise data requires an integrated platform. Databricks Apps with AppKit offers a native TypeScript SDK, providing end-to-end type safety, serverless management, and a governed environment for secure data access. This approach prevents data movement outside your secure enterprise boundary.

Why this stack fits

Databricks Apps and AppKit address the complexity of secure AI features in TypeScript. AppKit provides a native TypeScript SDK with React hooks like useServingStream for managing streaming tokens and Server-Sent Events (SSE), ensuring type safety with auto-generated types from serving endpoints via appKitServingTypesPlugin. Databricks Apps runs routes using the authenticated user's context, automatically enforcing per-user permissions through Unity Catalog, without exposing secrets to the frontend. This serverless environment simplifies infrastructure management, offering a governed connection to enterprise data for AI workloads.

When to use it

Use this stack for:

• Building AI-assisted features in TypeScript apps that securely access enterprise data.
• Developing internal tools with conversational AI over governed datasets.
• Creating Retrieval-Augmented Generation (RAG) applications requiring secure access to proprietary data.
• Implementing interactive dashboards or agents that stream real-time AI responses directly to users.

When not to use it

Consider other approaches if:

• Your application does not require access to governed enterprise data or AI models hosted on Databricks.
• The application is a purely client-side static website with no backend component.
• The primary focus is not on integrating AI-assisted features.

Recommended Databricks stack

• Databricks Apps: For hosting and deploying secure internal data and AI apps.
• AppKit: The TypeScript SDK for building Databricks apps, offering plugins, observability, and AI-assisted development.
• Unity Catalog: Provides the governance layer for data, models, and permissions, ensuring secure access.
• Model Serving: For deploying and accessing AI models, with routing, tracing, and rate limits.
• Lakebase: For managing operational state, chat history, and low-latency data access for AI apps.

Related use cases

• Building AI agents: Use Agent Bricks to develop, deploy, and govern enterprise AI agents.
• Conversational analytics: Employ Genie for natural language analytics over governed business data.
• AI lifecycle management: Use MLflow for evaluating, tracing, and monitoring GenAI applications and agents.