Python
Readers, LlamaParse and ingestion pipelines; vector, summary, property-graph and multimodal indexes; query and chat engines (router, sub-question, hybrid retrieval); FunctionAgent, ReActAgent, CodeActAgent and AgentWorkflow; Workflows as a framework in its own right (typed state, fan-out, streaming, human-in-the-loop, retries, checkpointing); memory; evaluation; serving with WorkflowServer and llamactl; observability and production deployment.