ai.internal.uncached_input_tokens
Input tokens billed at the full rate, for a provider whose `total` counts the cached share (and, where it reports them, the cache writes) too: null when it reported no total, never below zero.
Reference version
Signature
function ai.internal.uncached_input_tokens(
total: int | null,
cached: int | null,
cache_write: int | null = …
) -> int | null throws never
Input tokens billed at the full rate, for a provider whose total counts
the cached share (and, where it reports them, the cache writes) too: null
when it reported no total, never below zero.
Source:<builtin>/ai/ns_internal/helpers.bamlbytes 7147–7518