Automated Research Report Pipeline

Problem/Feature Description

A market research firm wants to automate their analyst workflow in a Quarkus backend. Currently, analysts manually perform two steps for every research request: first they gather and summarize raw data from a topic, then they write an executive summary suitable for business stakeholders. These are distinct cognitive tasks — the summarization specialist works differently from the executive-writing specialist — so the team wants them modeled as separate, independently testable agents.

The two agents must run in sequence: the output of the data summarizer feeds directly into the executive writer as its input. A single REST endpoint (POST /research/report) accepts a research topic string and returns the final executive report. The overall pipeline should be implemented using Quarkus agentic workflow primitives so the orchestration logic is declarative and the individual agents remain focused on their own responsibilities.

Output Specification

Produce a complete Maven project with:

• pom.xml — all required dependencies including the agentic extension
• src/main/java/.../SummarizerAgent.java — agent interface for data summarization
• src/main/java/.../WriterAgent.java — agent interface for executive writing
• src/main/java/.../ResearchPipeline.java — the orchestrating workflow component that composes the two agents
• src/main/java/.../ResearchResource.java — REST endpoint
• src/main/resources/application.properties — model provider configuration

Write a ARCHITECTURE.md describing how data flows between the agents and how the final result is extracted from the workflow.