Thinking in LionAGI¶

You're in Step 2 of the Learning Path

Coming from the Quick Start? Great! Now let's understand the mental model that makes LionAGI different from other frameworks.

The key to successfully using LionAGI is understanding its fundamental paradigm shift: from conversations to graphs, from sequential to parallel, from rigid to flexible.

Core Ideas¶

Why LionAGI?¶

The technical differences that matter.

Branches as Agents ¶

Why we call them branches, not agents.

Graphs Over Chains ¶

Parallel execution with dependency management.

The Builder Pattern ¶

Constructing workflows programmatically.

The Paradigm Shift¶

From Conversations to Graphs¶

Old way: Agents have conversations

agent1: "I found X"
agent2: "Based on X, I think Y"
agent3: "Given X and Y, we should Z"

LionAGI way: Agents execute operations

x = await agent1.analyze()
y = await agent2.evaluate(x)
z = await agent3.synthesize([x, y])

From Sequential to Parallel¶

Old way: Wait for each step

step1()  # 2 sec
step2()  # 2 sec  
step3()  # 2 sec
# Total: 6 seconds

LionAGI way: Run simultaneously

await gather(step1(), step2(), step3())
# Total: 2 seconds

From Rigid to Flexible¶

Old way: Fixed agent roles

researcher = Agent(role="researcher")
# Always a researcher, can't adapt

LionAGI way: Dynamic capabilities

branch = Branch(system=context_specific_prompt)
# Adapts to the task at hand

Ready for Core Concepts?

Now that you understand LionAGI's mental model, dive into the practical mechanics:

Next: Core Concepts - Learn Sessions, Branches, Operations, and Memory
Or jump to: Patterns if you want to see these concepts in action