WORTH TESTING · 27 OCTOBER 2026
Google's free Graph RAG agent codelab (reposted as a 'course')
This 54-minute video is a reupload of Google Cloud's free Graph RAG codelab, presented by Annie, a Developer Relations Engineer at Google. Credit goes to Google: the 'course' framing (and the '$750,000 skill' hype) is the poster's (@ryouxbt), not Google's. It builds a small agent that stores people and skills in a Spanner graph database, adds meaning-based and hybrid search in SQL, and wraps it all in Google's open-source ADK. Verdict: a genuinely useful hands-on intro to Graph RAG on a toy dataset; intermediate to advanced, coding needed (terminal, Python, SQL), and you'll need a Google Cloud project. Go straight to the codelab: goo.gle
Try this
- Wrap an agent as a tool that another agent calls only when needed, instead of building deep hierarchies. 02:32
- Claim the $5 Google Cloud workshop credit linked from the codelab (check whether a card is needed). 06:18
- Use Cloud Shell Editor (VS Code in the browser) to skip local setup problems. 07:08
- Install the Python dependencies with uv sync. 14:24
- Run a graph query in Spanner Studio to see your data as an interactive map of nodes and links. 18:58
- Generate embeddings inside SQL with a Spanner 'virtual model' that points at Vertex AI. 25:10
- Add an embedding column to an existing table and fill it with ML.PREDICT. 29:30
- Do semantic search by ordering results on COSINE_DISTANCE(). 31:00
- Combine keyword and semantic search with Reciprocal Rank Fusion, so 'medic' also finds 'Medical Training'. 40:05
- Use adk web for a local chat UI that traces tool calls and timings before you build a front end. 44:18
Timestamps point to the moment in the source video.
Cost
Free codelab; needs a Google Cloud project with billing (the codelab offers a $5 workshop credit).
How we knowSteps to try, not a claim that we independently tested every result. Follow the linked sources and check current prices. Our methods & sources ↗