Technique
Prompt Chaining
Decomposing a task into a sequence of model calls, each consuming the prior output.
Definition
Prompt chaining pipes outputs from one LLM call into the prompt of the next. It is more reliable than asking a single prompt to do everything, easier to debug, and the basis for most LangChain/LlamaIndex 'chain' abstractions and agentic workflows.
Common use cases
- Workflow orchestration
- Document pipelines
- Quality control