The HATEFUL Stack

The HATEFUL stack is a full-stack architecture designed for rapid development of AI-augmented web applications. On the frontend, we use HTMX to keep our markup dynamic and responsive without requiring a heavy JavaScript framework for every interaction, while Angular handles more complex single-page application logic where needed. TypeScript underpins both layers, ensuring type safety across our entire codebase and reducing runtime errors in production.

The backend leverages Express as our lightweight Node.js server for handling HTTP requests, API routing, and session management, while FastAPI runs as a dedicated microservice for data-heavy operations and machine learning integrations. We chose this dual-runtime approach because Express excels at serving our web tier and coordinating requests, while FastAPI's async capabilities and automatic OpenAPI documentation make it ideal for our compute-intensive workloads. We manage our Python dependencies and virtual environments with uv, whose Rust-based performance ensures our builds and deployments stay fast as we scale.

LangChain sits at the core of our intelligence layer, orchestrating interactions with large language models and managing the chains of operations that power our AI features. When a user initiates an action through HTMX or Angular, the Express server routes it appropriately—lightweight requests stay local, while anything requiring AI reasoning gets marshaled through FastAPI to LangChain, which handles the model calls, memory management, and tool integration. This separation of concerns lets us scale each component independently while keeping deployment straightforward.