Tutorials
The guides cover the prompt you type into a chat box. These cover the prompt that lives in your code — the instructions behind an assistant, the wording that makes a model return clean JSON, the rules that keep an answer grounded in your data. Each one is prompt-first, with small code snippets where they make it concrete.
Prompting is not just for chat. Each tutorial points at a real, runnable example app — a fraud-triage service, an investigation agent, and a dispute-resolution RAG pipeline — so you can see the same patterns in production code, in Python and Java.
Writing a system prompt
The persistent instructions behind every response in an app. What belongs in the system prompt versus the user turn, and the failures that come from mixing them up.
Read the tutorial →Getting reliable structured output
Prompting a model to return JSON your code can parse on the first try — every time. Schemas, examples, and the rule for missing fields that keeps a pipeline honest.
Read the tutorial →Grounding answers on your own data (RAG)
The prompt half of retrieval-augmented generation: making the model answer from the context you retrieved, cite it, and say “not found” instead of inventing an answer.
Read the tutorial →Prompting agents and tool use
Instructing a model on when to call a tool versus answer directly, how to describe a tool so it is used correctly, and how to keep an agent from looping or guessing.
Read the tutorial →