Record Token Usage on the Current Span

Convenience wrapper that records input and output token counts on the currently active span from the trace context.

Usage

record_tokens(input_tokens, output_tokens, model = NULL)

Arguments

input_tokens

Number of input tokens.

output_tokens

Number of output tokens.

model

Optional model name string.

Value

Invisible NULL.

Examples

# Record tokens on the active span inside a trace
with_trace("token-demo", {
  with_span("llm-step", type = "llm", {
    record_tokens(500, 200, model = "gpt-4o")
    current_span()$input_tokens
  })
})
#> [1] 500