GraphRAG in Production

What you'll learn

• ✓Prepare real PDF and HTML documents for indexing: parsing pitfalls, cleaning, and a token diet that cuts cost
• ✓Run a production-size GraphRAG index and read its wall time, token usage and dollar cost
• ✓Evaluate a knowledge graph: structure metrics, LLM-judged extraction quality, and answer quality against a vector baseline
• ✓Build a multi-source graph and trace an answer back to its source document and type
• ✓Build a GraphRAG agent with function calling and compare it with a simple router
• ✓Decide, with numbers, when GraphRAG is worth its cost and when it is not

About this course

A tutorial corpus never taught anyone what production feels like. This course runs GraphRAG on real public documents — US congressional records, bills and committee reports about artificial intelligence legislation, all public domain — and answers the question that decides whether you ship it: is this graph actually any good?

Session 1 is the document work nobody demos: downloading from govinfo, parsing PDFs and HTML, headers and reprinted text, a token diet that cut our corpus by a quarter, and a cost estimate before you spend anything. Then a production index, with settings, entity types and prompt tuning, and its real wall time and bill. Session 2 evaluates it: graph-structure metrics, entity-extraction quality judged by an LLM against the source text, automatically generated evaluation questions, and answer quality for local search, global search and a plain vector baseline. Session 3 builds multi-source graphs with provenance so an answer can be traced back to the source type it came from. Session 4 puts a function-calling agent in front of the graph and compares it with a simple router.

Every number in the course comes from a run you can repeat — including the uncomfortable one: on simple factual questions the vector baseline outscored both graph retrievals at a thirtieth of the cost. That is taught as the first decision rule, not hidden.

**Licensing.** The corpus is deliberately unencumbered: Congressional Record extracts, House bills and committee reports from govinfo.gov are works of the United States Government and carry no copyright under 17 U.S.C. § 105, so you may redistribute, train on and build commercially on `data/congress/` freely. `Resources/LICENSES.md` records this and covers Microsoft GraphRAG (MIT) and the optional local Ollama models.

Curriculum5 sections · 19 lectures · 2h 8m

Read alongside the course