Agentic RAG (RAG as an Agent Loop)

What Problem It Solves

Traditional RAG is often “one retrieve → one generate”. Agentic RAG lets the model decide:

• when to retrieve
• what to retrieve
• when evidence is sufficient
• when to stop and answer

Core Flow (ReAct + Retrieval Tool + Evidence Ledger)

flowchart TD
  Q["Question"] --> D["Decide next action"]
  D -->|search| S["Retrieve"]
  S --> L["Update evidence ledger"]
  L --> D
  D -->|final| A["Answer + cite doc_ids"]

Evolution Path

• Built on: ReAct + Retrieval Loop ideas
• Frequently combined with: CoVe (verify claims), Memory (store insights)

Repo Reference

• Code: src/agent_patterns_lab/patterns/agentic_rag.py
• Example: examples/41_agentic_rag.py
• Tests: tests/test_agentic_rag.py