Skip to main content

Module 16: Static Orchestration - Linear and Parallel Edges

Theory​

In ADK 2.0, orchestration is no longer about choosing between a "Sequential" or a "Parallel" agent. Instead, you design the geometry of your graph using Edges.

A Workflow is a deterministic engine that executes nodes based on these connections. By arranging your edges, you can create linear pipelines, concurrent branches, or a mix of both.

1. Sequential Flow (Linear Edges)​

When you define edges in a chain, the ADK executes them one after another. Each node waits for the previous one to finish and receives its output as input.

from google.adk import Workflow

# A -> B -> C
my_pipeline = Workflow(
    name="SequentialSystem",
    edges=[
        ("START", researcher),
        (researcher, writer),
        (writer, editor)
    ]
)

2. Parallel Flow (Fan-Out)​

When multiple edges originate from the same source (like "START" or a specific node), the ADK triggers all target nodes simultaneously.

# A and B start at the same time
parallel_run = Workflow(
    name="ParallelSystem",
    edges=[
        ("START", task_a),
        ("START", task_b)
    ]
)

3. Synchronization (The JoinNode)​

When you have parallel branches, you often need to "wait" for all of them to finish before moving to a final step. This is called Fan-In, and it is handled by the JoinNode.

The JoinNode acts as a barrier: it will not fire its outgoing edge until every incoming edge has completed.

from google.adk.workflow import JoinNode

syncer = JoinNode(name="sync_point")

# Fan-out to A and B, then converge at 'syncer' to run C
complex_graph = Workflow(
    name="HybridSystem",
    edges=[
        ("START", task_a, syncer),
        ("START", task_b, syncer),
        (syncer, final_task)
    ]
)

Why use Static Orchestration?​

  1. Performance: Parallel execution reduces total latency to the duration of the slowest branch.
  2. Predictability: The path is fixed and code-defined; there is no "hallucination" risk in the routing.
  3. Transparency: The Dev UI Graph View gives you a 1:1 visual map of your business process.

Key Takeaways​

  • Edges define the flow: Chain them for sequential, branch them for parallel.
  • JoinNode is the mandatory synchronization point for parallel branches.
  • Data Flow: ADK 2.0 handles passing results between nodes automatically, ensuring a seamless pipeline.