Unit 3 — Agentic RAG¶

Overview¶

Agentic RAG (Retrieval-Augmented Generation) goes beyond naive RAG pipelines by giving the agent control over when and how to retrieve information. Instead of always retrieving before generating, the agent decides:

Whether retrieval is necessary at all
Which knowledge source to query
How many retrieval steps to perform
Whether to re-rank, filter, or synthesize results

Naive RAG vs. Agentic RAG¶

graph LR
  subgraph Naive RAG
    Q1[Query] --> R1[Retrieve] --> G1[Generate]
  end
  subgraph Agentic RAG
    Q2[Query] --> A2[Agent Reason]
    A2 -->|retrieve| R2[Vector Store]
    A2 -->|search| W2[Web]
    A2 -->|compute| C2[Code]
    R2 & W2 & C2 --> A2
    A2 -->|done| G2[Final Answer]
  end

Key techniques¶

Technique	Description
Router	Classifies the query to select the right retriever
Query rewriting	Reformulates the query before retrieval
Sub-question decomposition	Breaks complex questions into simpler sub-queries
Re-ranking	Re-scores retrieved chunks before synthesis
Iterative retrieval	Retrieves multiple times, using each answer to guide the next query

Notebooks¶

Notebook	Description
`notebooks/unit3/01_basic_rag_agent.ipynb`	RAG tool inside a smolagents agent
`notebooks/unit3/02_multi_source_rag.ipynb`	Agent that queries multiple knowledge bases

Notes¶

Add your unit 3 notes and experiments here.