Open Source

Build synthetic data pipelines
with composable blocks

dataset

→

PromptBuilder

→

LLMChat

→

TagParser

→

Filter

→

enriched

Chain LLM, parsing, transform, filtering, and agent blocks in any order. Each block does one thing well.

Define multi-step pipelines in YAML. Portable, reproducible, and version-controlled by design.

BlockRegistry and FlowRegistry find and catalog all available components automatically. Zero boilerplate.

100+ LLM providers through LiteLLM with async execution. Built for throughput from the ground up.

Build synthetic data pipelineswith composable blocks