multi-agent-rag-orchestration

name	multi-agent-rag-orchestration
title	Multi-Agent RAG Orchestration
description	Coordinate planner and specialist agents to decompose, retrieve, and synthesize complex RAG answers.
allowed-tools	["Read","Grep","Glob","Bash"]
category	rag-agents
tags	["multi-agent","orchestration","planner","decomposition"]

Overview

Multi-agent orchestration splits a complex question across specialized agents, such as a planner, one or more retrievers, and a synthesizer. Each agent owns a narrow job, and a coordinator sequences them and merges the results into a single grounded answer.

Problem Statement

Some questions require several independent lookups, cross-source reasoning, or distinct skills that a single prompt handles poorly. Cramming everything into one agent produces long, unfocused prompts and answers that quietly drop parts of the question.

Key Concepts

• Planner: Decomposes the query into ordered or parallel subtasks.
• Specialist agents: Narrowly scoped workers, such as per-source retrievers.
• Coordinator: Sequences agents, handles failures, and enforces limits.
• Synthesizer: Merges partial results into one grounded, cited answer.

Implementation Guide

Step 1: Decompose the Query

Have a planner break the question into subtasks that can be handled independently, with explicit dependencies where order matters.

Why: Clear decomposition prevents subtasks from overlapping or dropping parts of the original question.

Step 2: Assign Narrow Specialist Agents

Give each subtask to an agent with a single responsibility and only the tools it needs.

Why: Narrow scope keeps prompts short, decisions debuggable, and behavior predictable.

Step 3: Synthesize and Attribute

Merge partial answers in a synthesis step that resolves conflicts and preserves source citations.

Why: A dedicated synthesis step avoids contradictory fragments and keeps the final answer traceable.

For a worked pattern, see the Multi-Agent RAG example and the LangGraph multi-agent guide.

When to Use This Skill

• Questions need several independent lookups or cross-source reasoning
• Subtasks require distinct skills or tools that one agent handles poorly
• You need parallelism across retrievers to cut wall-clock latency

When NOT to Use This Skill

• Single-agent routing already answers the query well
• The added coordination cost and latency are not justified
• Subtasks are tightly coupled and cannot be cleanly separated

Related Skills

• Agentic RAG Routing - Try routing before orchestration
• Explainable Retrieval with Citations - Attribution during synthesis
• Graph RAG - Cross-entity reasoning
• RAG Evaluation Frameworks - Evaluate the whole system

Metrics & Success Criteria

• Decomposition Quality: Subtasks fully cover the original question
• Answer Completeness: Multi-part questions are answered in full
• Latency: Parallel agents keep end-to-end time acceptable
• Traceability: Each answer segment maps back to a source