A structured lesson workspace with readable content, hands-on examples, and a clean path to completion.
Our agent needs access to conference details. We will build a lightweight retrieval workflow that grabs a markdown file from GitHub and answers questions based on its contents. This is a sub-workflow, not a full RAG pipeline.
Build a simple knowledge retrieval workflow
This video requires an access code. Enter your code below to unlock it instantly.
Don't have a code? Contact us to get access.
Start with an 'Execute Workflow' trigger expecting a string input named 'question'. Connect a GitHub node to list and download the target markdown file. Use the 'Extract from Text File' node to convert the binary payload into plain text.
Feed that text into a standard Gemini 'Message a Model' node. The context window on Gemini 2.5 Flash supports up to 1 million tokens, making it trivial to pass entire documents without complex chunking strategies for small datasets.
{"systemMessage": "Answer the user question using ONLY the provided document context.","userMessage": "={{ $json.question }}","context": "={{ $json.extracted_text }}"}
Use a Set node to format the final output into a clean 'response' field. The heavy lifting is done. This workflow takes a question, fetches context, and returns a deterministic answer string.
Finish the lesson once you have worked through the material. This awards ★ 30 XP.