Petal Flow Overview

Petal Flow

Petal Flow is a graph runtime for building AI agent workflows. Chain LLM calls, tools, routers, and transformations into directed graphs that are inspectable, testable, and production-ready.

Why Petal Flow

Graph-first

Model workflows as nodes and edges with explicit entry points.

Runtime control

Built-in runtime events, debugging hooks, and step controllers.

LLM-native

LLM nodes, routers, tools, and guards built for AI operations.

Composable

Mix built-in nodes with custom Go functions.

What You Build with Petal Flow

Customer support routers that triage and escalate tickets
RAG workflows that retrieve, synthesize, and cite sources
Data enrichment pipelines with validation and transformation stages
Human-in-the-loop review systems with approval gates
Observable agent workflows with event-driven monitoring

How It Fits with Iris

Petal Flow and Iris are designed to work together. Use Iris providers inside Petal Flow nodes to power LLM calls while keeping routing and orchestration inside a graph.

// Use Iris providers inside Petal Flow nodes
llmNode := petalflow.NewLLMNode("summarize", llmConfig{
    Provider: openai.NewFromKeystore(),
    Model:    "gpt-4o",
    Prompt:   "Summarize the following: {{.input}}",
})

This keeps your AI stack modular and observable. Petal Flow handles the workflow orchestration while Iris handles the LLM interactions.

Quick Links

Getting Started Build your first workflow graph.

Concepts Learn nodes, edges, envelopes, and runtimes.

CLI Reference Command-line interface for running workflows.

Guides Deep dives into graph design and patterns.

API Reference Complete API documentation.

Examples Production-ready workflow examples.