Generative AI on Microsoft Azure by Adrin Gonzlez Snchez;Jaime De Mora;Jorge Garca Ximnez;

Author:Adrin Gonzlez Snchez;Jaime De Mora;Jorge Garca Ximnez;
Language: eng
Format: epub
Publisher: O'Reilly Media
Published: 2026-04-16T00:00:00+00:00

Figure 6-5. Workflow pattern example

Concurrent Pattern

In many real-world tasks, the workload naturally decomposes into pieces that can be processed in parallel rather than strictly in sequence in a scenario that the concurrent pattern addresses. In the concurrent pattern, multiple agents execute simultaneously (or near-simultaneously), each working on a separate subtask, and then the results are reconciled by a merge or aggregator agent (see Figure 6-6). For example, when processing a complex document, you could have a language agent finishing summarization, another checking factual consistency, a third doing taxonomy tagging, all in parallel; once they complete, an aggregator agent combines results, resolves conflicts or duplicates, and produces the final output. The advantage is speed and utilization of specialized agents, but you must carefully design for partial failures (one agent fails while others succeed), and you must ensure that the aggregator logic handles inconsistencies, duplicates, or missing outputs reliably.

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.