Tinker420 tokens · 183 words · 1083 characters420 of 160 tokens, counted by this site’s tokenizer, live.
Chapter 7 of 7
Workbench: fit the prompt
Larkspur’s assistant builds each prompt from parts: instructions, the customer’s question, document pieces, examples. The team has a token budget per request and must choose what fits.
This chapter: Which parts of Larkspur’s prompt fit in the budget, and which matter most?
You will learn
- How a prompt’s token count adds up from its parts
- How to fit a prompt into a budget, counting live
- Why you count with the tokenizer you will be billed by
Words you’ll meet (2)
- Token budget
- The most tokens a team allows for one request, set for cost or for the window.
- Prompt
- Everything sent to the model in one request.
The picture on the left is live. Each step below asks you to do one thing in it; the story continues here as you go.
Step 1 of 4
Larkspur’s assistant builds each prompt from parts: rules, tone guide, worked examples, earlier chat, refund policy, question. The team has a budget of 160 tokens per request, set to keep the cost down. The picture counts the prompt live, with this site’s tokenizer. Before you start: