near count
- Last Updated: June 23, 2026
- 1 minute read
- Semaphore
- Documentation
The near count metadata allows us to specify how many tokens can be skipped across the entire context.
This only counts tokens, that is, words, or, if punctuation is included, then it will also include punctuation marks in that count (as well as the words). The count, even if the context is across sentence or paragraphs, is still courning tokens.