record_request

record_request(
  self,
  model_name: str,
  input_toks: int,
  output_toks: int,
  provider

Parameters

Name	Type	Description
`model_name`*	`str`	The model that served this request. Combined with `provider` to form the per-model key. Pass an empty string to skip the per-model breakdown for this request.
`input_toks`*	`int`	Input tokens for this request.
`output_toks`*	`int`	Output tokens for this request.
`provider`	`str`	Default:`''`
`cost_usd`	`float \| None`	Default:`None`
`kind`	`UsageKind`	Default:`'assistant'`
`cache_read_tokens`	`int`	Default:`0`
`cache_write_tokens`	`int`	Default:`0`

Accumulate usage for one completed LLM request.

Updates session totals plus the per-model and per-type breakdowns.

Provider that served the model (e.g. openai).

Combined with model_name to form the per-model key, so the same model served by different providers is tracked separately.

Estimated request cost, or None when no estimate exists.

Missing estimates leave monetary totals unchanged.

Request class used for /cost type breakdowns.

Input tokens served from provider caches.

Input tokens written to provider caches.