Sequential

What this lesson demonstrates

Sequential orchestration wires agents into a pipeline: each agent runs in turn and passes its output to the next. It fits any “each step builds on the last” job — draft then review, extract then summarize, translate then proof-read. You hand a list of participants to SequentialBuilder, call .build(), and run it once with a prompt.

The lesson chains a copywriter (drafts a marketing sentence) into a reviewer (critiques the draft), showing how order in the list is the execution order.

One real excerpt

The builder lives in agent_framework.orchestrations; participants run strictly in list order:

from agent_framework.orchestrations import SequentialBuilder

return SequentialBuilder(participants=[writer, reviewer]).build()

# ...
events = await workflow.run("Write a tagline for a budget-friendly eBike.")
final: AgentResponse = events.get_outputs()[0]   # ONLY the last agent's messages
for msg in final.messages:
    print(f"[{msg.author_name or 'assistant'}]\n{msg.text}\n")

The gotcha

By default each agent sees the full prior conversation (inputs + all responses). Pass chain_only_agent_responses=True so each agent consumes only the previous agent’s reply — handy for translation or progressive-refinement chains. Also, events.get_outputs()[0] is an AgentResponse holding only the last agent’s messages, not the whole chat; use msg.author_name to tell which agent produced each message.

How it maps to Azure AI Foundry

Both participants are FoundryChatClient + AzureCliCredential agents; SequentialBuilder is client-agnostic, so the same pipeline works with any provider. Construction is offline — only .run() calls the model, once per participant in order.

Run it

uv run tutorial/03-workflows/orchestrations/01_sequential.py

Needs Foundry credentials (az login). You should see a ===== Final Response ===== block with each participant’s messages labelled by name.

Next: Concurrent