ControlFlow
What is ControlFlow?
ControlFlow is a Python framework for building agentic AI workflows.
An agentic workflow is a process that delegates at least some of its work to an LLM agent. An agent is an autonomous entity that is invoked repeatedly to make decisions and perform complex tasks. To learn more, see the AI glossary.
ControlFlow provides a structured, developer-focused framework for defining workflows and delegating work to LLMs, without sacrificing control or transparency:
- Create discrete, observable tasks for an AI to solve.
- Assign one or more specialized AI agents to each task.
- Combine tasks into a flow to orchestrate more complex behaviors.
This task-centric approach allows you to harness the power of AI for complex workflows while maintaining fine-grained control. By defining clear objectives and constraints for each task, you can balance AI autonomy with precise oversight, letting you build sophisticated AI-powered applications with confidence.
Quickstart
Here’s a simple but complete ControlFlow script that writes a poem:
The run()
function is the main entry point for ControlFlow. This single line of code creates a task, assigns it to an agent, and immediately executes it, returning the result. You can completely customize those behaviors by learning more about tasks, agents, and flows.
Key features
Let’s explore some of ControlFlow’s key features:
Structured results
ControlFlow tasks can return more than just text, including any structured data type supported by Pydantic:
You can also output a list of strings or choose from a list of predefined options:
Custom tools
Provide any Python function as a tool for agents to use:
Multi-agent collaboration
Assign multiple agents to a task to enable collaboration:
User interaction
Quickly give agents the ability to chat with users:
Flows
Use flows to create complex workflows by running all tasks with a shared context and message history:
import controlflow as cf
@cf.flow
def create_story():
# get the topic from the user
topic = cf.run(
"Ask the user to provide a topic for a short story", interactive=True
)
# choose a genre
genre_selector = cf.Agent(
name="GenreSelector",
instructions="You are an expert at selecting appropriate genres based on prompts.",
)
genre = genre_selector.run(
"Select a genre for a short story",
result_type=["Science Fiction", "Fantasy", "Mystery"],
context=dict(topic=topic),
)
# choose a setting based on the genre
if genre == "Science Fiction":
setting = cf.run("Describe a distant planet in a binary star system")
elif genre == "Fantasy":
setting = cf.run("Create a magical floating city in the clouds")
else: # Mystery
setting = cf.run("Design an isolated mansion with secret passages")
# create a writer agent
writer = cf.Agent(
name="StoryWriter", instructions=f"You are an expert {genre} writer."
)
# create characters
characters = writer.run(
f"Create three unique characters suitable for a the provided genre, setting, and topic.",
context=dict(genre=genre, setting=setting, topic=topic),
)
# write the story
story = writer.run(
f"Write a short story using the provided genre, setting, topic, and characters.",
context=dict(genre=genre, setting=setting, topic=topic, characters=characters),
)
return dict(
topic=topic,
genre=genre,
setting=setting,
characters=characters,
story=story,
)
result = create_story()
print(result)
Why ControlFlow?
- 🔗 Seamless Integration: Blend AI capabilities with your existing Python codebase effortlessly.
- 🎛️ Fine-grained Control: Balance automation with oversight, maintaining control over your AI workflows.
- 📈 Scalability: From simple scripts to complex applications, ControlFlow grows with your needs.
- 🔍 Transparency: Gain insights into your AI’s decision-making process with built-in observability.
- 🚀 Rapid Prototyping: Quickly experiment with AI-powered features in your applications.
- 🤝 Productivity: Focus on your application logic while ControlFlow handles the intricacies of AI orchestration.
By providing a structured yet flexible approach to AI development, ControlFlow empowers you to create robust, intelligent applications with confidence.
Next Steps
- Install ControlFlow
- Explore the Core Concepts
- Browse Patterns for common use cases
- Check out the API Reference