Send usage: { include: true } on every chat completion request (an
OpenRouter-specific extension - requires Usage Accounting enabled on
the OpenRouter account; other OpenAI-compatible providers generally
ignore unknown top-level request fields). When the response includes
OpenRouter's own usage.cost, forward it to Langfuse as
costDetails.total, so cost is the real per-request figure OpenRouter
billed rather than a guess from Langfuse's model price list (which
has no entry for proxied/aggregated OpenRouter model names anyway).
Verified: full suite (84/84) passing, tsc/eslint clean, docker
compose build succeeds. Per Langfuse's OpenRouter integration guide:
https://langfuse.com/integrations/gateways/openrouter
OpenAI's chat completions response includes prompt_tokens_details.
cached_tokens when prompt caching kicks in - previously dropped
entirely. Now extracted in LlmClient and forwarded to Langfuse as the
flat cache_read_input_tokens usage key.
Used the flat key rather than OpenAI's nested prompt_tokens_details
shape: the installed @langfuse/tracing SDK's usageDetails type is
Record<string, number> | OpenAiUsage (a 3-field camelCase shape, no
_details nesting) even though the wire schema in @langfuse/core
supports nesting - a gap in the SDK's public types, not something
worth fighting with a cast.
Verified: full suite (82/82) passing, tsc/eslint clean.
