flowagents 0.1.1

A zero-code AI workflow orchestration framework with state-driven agents

FlowAgents

State-driven AI agent framework for task automation with human-in-the-loop.

Build agents that collect information, validate input, and execute tasks with optional approval.

Perfect For

• Booking & Reservations - Restaurants, flights, hotels, appointments
• Request & Approval - Leave requests, expense reports, access requests
• Form Wizards - Multi-step data collection with validation
• Business Workflows - Order processing, customer service, IT ticketing

Installation

pip install flowagents              # Core
pip install flowagents[openai]      # + OpenAI
pip install flowagents[anthropic]   # + Claude
pip install flowagents[all]         # All providers

Quick Example

import asyncio
from flowagents import flowagent, StandardAgent, InputField, AgentStatus
from flowagents import Orchestrator, OpenAIClient

@flowagent(triggers=["book", "reservation"])
class BookingAgent(StandardAgent):
    guests = InputField("How many guests?")
    date = InputField("What date?")

    async def on_running(self, msg):
        return self.make_result(
            status=AgentStatus.COMPLETED,
            raw_message=f"Booked for {self.guests} on {self.date}!"
        )

async def main():
    llm = OpenAIClient(api_key="sk-xxx", model="gpt-4o-mini")
    orchestrator = Orchestrator(llm_client=llm)
    await orchestrator.initialize()

    result = await orchestrator.handle_message("user_1", "I want to book a table")
    print(result.raw_message)  # "How many guests?"

    result = await orchestrator.handle_message("user_1", "4")
    print(result.raw_message)  # "What date?"

    result = await orchestrator.handle_message("user_1", "Friday")
    print(result.raw_message)  # "Booked for 4 on Friday!"

asyncio.run(main())

Key Features

• State Machine - INITIALIZING → WAITING_FOR_INPUT → RUNNING → COMPLETED
• Field Collection - Declarative InputField with validation
• Approval Workflow - Human confirmation before sensitive actions
• Multi-LLM - OpenAI, Claude, Azure, Gemini, Ollama, DashScope
• Orchestrator - Route messages to the right agent
• Multi-tenant - Built-in isolation for SaaS

Documentation

• Getting Started - Build your first agent
• Full Documentation - Complete guide
• Examples - Sample code

Guides: Agents | Tools | LLM Providers | Orchestrator | Workflow

Contributing

See CONTRIBUTING.md for guidelines.

License

MIT