ADD Single Agent with Spring AI Alibaba

This project implements the assignment option Qwen3-Max + Single-agent with Spring AI Alibaba.

The application runs one architecture design agent that:

• uses only the prior knowledge in src/main/resources/knowledge/prior-knowledge.md;
• follows the required four ADD 3.0 iterations;
• performs draft, self-reflection, and revision for each iteration;
• writes a timestamped JSONL conversation log;
• writes the final ADD output in Markdown.

Requirements

• Alibaba Cloud Bailian/DashScope API key
• Docker Desktop, for Docker development
• JDK 17+ and Maven 3.8+, only for local non-Docker runs

Docker Development

Use Docker to avoid installing Java or Maven on the host machine.

cp .env.example .env
# edit .env and set AI_DASHSCOPE_API_KEY
docker compose run --rm dev

Inside the container:

mvn test
mvn spring-boot:run

Or run the assignment agent directly from the host:

docker compose run --rm agent

Maven dependencies are stored in the Docker named volume maven-cache, not in the project directory.

Local Run

export AI_DASHSCOPE_API_KEY=your_api_key
export AI_MODEL=qwen3-max
mvn spring-boot:run

Generated files:

• logs/conversation_log.jsonl
• outputs/add_results.md

The log stores timestamps, prompts, responses, latency, and estimated token counts. For the final report, use DashScope/Bailian billing or usage records if exact token consumption is required.

If the model identifier in your DashScope workspace differs from qwen3-max, keep the assignment report model name as Qwen3-Max and set AI_MODEL to the exact model id available in the workspace.

Assignment Mapping

• AI paradigm: Single-agent
• LLM: Qwen3-Max
• Framework: Spring AI Alibaba DashScope chat model
• Source code deliverable: this Maven project
• Conversation log deliverable: logs/conversation_log.jsonl
• Report input: outputs/add_results.md