Common LLM Architectures

With Vellum Workflows’ extensive Nodes and orchestration features, you can build a wide variety of LLM application architectures. Below you’ll find common patterns that serve as excellent starting points for your own applications.

For more extensive example Workflows, click here to view our examples section.

Build a Retrieval Augmented Generation system to enhance LLM responses with relevant data.

Escalation to a Human

Automatically route sensitive or complex messages to human operators.

Prompt Retry Logic

Implement error handling with automatic retries for non-deterministic failures.

PDF Content Summarization

Extract and summarize the contents of PDF documents.

Fallback Models

Implement dynamic model selection with fallback logic to handle errors and optimize cost/performance.