Guides
How to build real things with the catalogue — each guide works through a problem end to end with code you can run.
Chunking strategies for retrieval: what actually changes results
Chunk size gets the attention and boundaries do the work. Why fixed-size splitting underperforms, how to split on structure instead, and what to attach to every chunk.
How to build link previews that don't break
Reading og:title takes an hour. Handling the web that doesn't have it takes a week. The fallback chains, redirect handling and caching strategy that make unfurling actually work.
How to control LLM API costs before the invoice arrives
Most teams discover their inference bill after they've spent it. The four places cost actually accumulates, how to measure each one, and the guard clauses that stop a runaway.
How to generate QR codes that actually scan when printed
Most QR codes that fail in the field fail for three avoidable reasons: raster output, too little quiet zone, and poor contrast. What to do instead, with runnable examples.
How to give an AI agent real tools with MCP
Model Context Protocol lets an agent call an API as a tool without you writing an integration. What MCP actually is, how to connect one, and why tool descriptions decide whether it works.
How to cut your email bounce rate before you send anything
Bounces do not just waste a message — they lower the sender reputation that decides whether your next one arrives. A practical method for catching bad addresses at the point of capture.
How to prepare web content for a RAG pipeline without wasting tokens
Raw HTML is mostly boilerplate, and every byte of it is billed. A practical method for extracting, measuring and chunking web content before it reaches a model.