stream-chain

Multi-agent workflows are coordinated by streaming outputs between sequential steps, enabling per-step context propagation and real-time data transformation across diverse agents.

The best way to build a data transformation pipeline is using custom chains where each step's output feeds the next, ensuring seamless context propagation and improving overall throughput.

Yes, you can use predefined pipelines for sequential processing tasks, applying configurable options to orchestrate common workflows with streaming context flow across all steps.

Multi-agent orchestration supports per-step output propagation, allowing real-time coordination across streaming steps to execute ordered tasks like code analysis and deployment.

You need streaming orchestration when executing complex, multi-step pipelines such as coordinating code analysis, refactoring, tests, and deployment tasks in a single, ordered chain.

To start a custom chain, provide prompts that define each step, then observe how the outputs flow through the chain in sequence with configurable options applied.