Companion videos

RAG beyond chunks: graph RAG, agentic RAG, long-context RAG — companion videos

The article steps past chunk-and-embed RAG into the three families that have actually pushed quality forward — knowledge-graph retrieval, agentic retrieval loops, and the "just stuff a million tokens" approach. These two videos give you a working demo of two of those families side by side, plus the Microsoft-style GraphRAG architecture in its native form, so the comparisons in the article land with concrete reference points.

Primary pick

38:48

Introducing RAG 2.0: Agentic RAG + Knowledge Graphs (FREE Template)

Cole Medin

A working agentic-RAG-plus-knowledge-graph build, with the agent deciding when to do vector search, when to hit Neo4j, and when to do both. It's the cleanest demonstration on YouTube of the "agent as the retrieval planner" pattern the article describes, in code you can actually pull down and run.

What you should get from this: Compare agentic retrieval and graph retrieval in a concrete implementation.

Watch or know first: Comfortable with basic RAG, graph databases and running a template locally.

AI Expert note: Treat template dependencies, model choices and orchestration code as version-sensitive. The useful idea is retrieval planning; production use still needs permissions, evals and failure handling.

Open video page

Also worth watching

15:58

Graph RAG: Improving RAG with Knowledge Graphs

Prompt Engineering

A focused walkthrough of Microsoft's GraphRAG — entity extraction, community summaries, query-focused summarization — set up on a local machine with cost notes. Watch it for the graph-RAG section of the article specifically; the cost discussion is the part most write-ups skip.

What you should get from this: Understand the Microsoft-style GraphRAG flow: entity extraction, communities, summaries and query-focused synthesis.

Watch or know first: Know why plain vector search can miss relationship-heavy questions.

AI Expert note: Setup commands and costs can drift, but the architecture is useful. Recalculate current token costs and validate whether graph construction is justified before adopting this pattern.

Open video page