Skip to content

FlowgentraAI

Build AI agent workflows with graphs.

FlowgentraAI is a Python library for building LLM-powered agent workflows using a graph-based architecture. It's powered by a high-performance Rust engine via PyO3, giving you Python's ease of use with Rust's speed.

Why FlowgentraAI?

  • Graph-first: Model your agent logic as nodes and edges -- easy to understand, test, and debug
  • Rust-powered: Core engine written in Rust for maximum performance
  • Multi-provider LLM support: OpenAI, Anthropic, Mistral, Groq, Ollama, HuggingFace, Azure
  • Built-in RAG: Vector stores, embeddings, text splitting, retrieval pipelines
  • Human-in-the-loop: Interrupt and resume graph execution with checkpointing
  • Multi-agent: Supervisor pattern for orchestrating multiple agent graphs
  • Observability: Execution tracing, visualization (DOT, Mermaid)

Quick Example

from flowgentra_ai import StateGraphBuilder, SharedState, END

def process(state):
    state["output"] = state["input"].upper()
    return state

builder = StateGraphBuilder()
builder.add_node("process", process)
builder.set_entry_point("process")
builder.add_edge("process", END)
graph = builder.compile()

result = graph.invoke(SharedState({"input": "hello"}))
print(result["output"])  # "HELLO"

Get Started

  • :material-download: Installation -- Install FlowgentraAI
  • :material-rocket-launch: Quick Start -- Build your first graph workflow
  • :material-book-open-variant: Guides -- In-depth tutorials
  • :material-api: API Reference -- Full API documentation