Multi-Agent Workflow Examples

Custom workflow examples implementing patterns from multiagents/custom.md, multiagents/code/rag-agent-workflow.md, and multiagents/code/sql-agent-workflow.md.

Packages

Package	Description	Config
`ragagent`	RAG workflow: rewrite → retrieve → agent	`workflow.rag.enabled=true`
`sqlagent`	SQL agent with list_tables, get_schema, run_query tools	`workflow.sql.enabled=true`

RAG Agent

Flow: Query → Rewrite → Retrieve → Prepare → Agent → Response

Rewrite: LLM rewrites the query for better retrieval
Retrieve: Vector similarity search (deterministic)
Prepare: Formats context and question into a prompt
Agent: ReactAgent with context; can use get_latest_news tool

Requires: DashScope API key, EmbeddingModel (from spring-ai-alibaba-starter-dashscope).

SQL Agent

Flow: Question → Agent (list_tables → get_schema → run_query) → Answer

Uses H2 in-memory with Chinook-like schema
Agent uses three tools: sql_db_list_tables, sql_db_schema, sql_db_query
Only SELECT queries allowed (no DML)

Run

bash

# RAG agent
mvn spring-boot:run -Dspring-boot.run.arguments="--workflow.rag.enabled=true --workflow.runner.enabled=true"

# SQL agent
mvn spring-boot:run -Dspring-boot.run.arguments="--workflow.sql.enabled=true --workflow.runner.enabled=true"

Set AI_DASHSCOPE_API_KEY environment variable.