stream-chain

Multi-agent workflows for code refactoring are automated by chaining sequential prompts where one agent's output feeds the next, enabling tasks like identifying code smells, planning changes, implementing refactors, and verifying results.

Prompt chaining for data transformation pipelines is the process of executing sequential, multi-agent workflows where the output of one agent serves as the input for the next, utilizing streaming JSON for agent coordination.

Custom prompt chains can be defined and executed to create tailored workflows, allowing you to sequence prompts for specific data transformation and multi-agent automation tasks beyond predefined pipelines.

Multi-agent pipeline orchestration supports predefined pipelines for common tasks like code analysis, testing, and optimization, providing battle-tested workflows that can be executed immediately without custom setup.

The best way to coordinate agents in a sequential pipeline is using streaming JSON to pass data between agents, ensuring the output of one step directly serves as the structured input for the next.

Multi-agent prompt chaining should be avoided for simple, single-step tasks that do not require sequential data transformation or when your workflow lacks a clear input-output dependency chain between distinct processing stages.