ADK Workflows & Sub-Agents

• Key points are in the visual diagram above.

Monolithic agents hallucinate on long tasks. Workflows decompose work into focused sub-agents with clear inputs/outputs, improving reliability and testability.

ADK Workflows are a film production crew: director (orchestrator) assigns scenes to specialists (sub-agents) who work in sequence or parallel.

Scenario

Research workflow: ParallelAgent runs `web_search_agent` and `internal_kb_agent` → SequentialAgent passes merged context to `synthesis_agent` → `citation_checker_agent`.

Solution

In Google ADK, apply ADK Workflows & Sub-Agents to this scenario: Research workflow: ParallelAgent runs `web_search_agent` and `internal_kb_agent` → SequentialAgent passes merged context to `synthesis_agent` → `citation_checker_agent`. Identify the inputs, run the technique, validate the output, and note one thing you would monitor in production.

Do this before moving to the next module — reading alone is not enough.

Open the Code Walkthrough below and run it locally. Change one parameter related to ADK Workflows & Sub-Agents (e.g. model, temperature, top_k, or tool name), observe the difference in output, and write 2–3 sentences explaining what changed.

Highlighted lines show where ADK Workflows & Sub-Agents happens in the code.

1from google.adk.agents import SequentialAgent, ParallelAgent, LlmAgent  # import dependencies2
3research = ParallelAgent(agents=[web_agent, kb_agent], name="gather")4pipeline = SequentialAgent(agents=[research, synthesis_agent], name="research_flow")

• Treating ADK Workflows & Sub-Agents as a black box without evaluation
• Ignoring cost and latency in production
• Skipping error handling for google adk workflows

Quick recap — the most important points from this module.