🦋 New version release

[brentrager]

This PR was opened by the Changesets release GitHub action. When you're ready to do a release, you can merge this and the packages will be published to npm automatically. If you're not ready to do a release yet, that's fine, whenever you add more changesets to main, this PR will be updated.

Releases

@smooai/smooth-operator@1.54.2

Patch Changes

• 3da620b: Emit usage provenance, cost provenance, and the gateway response id on the turn span.

  Closes a hole in the previous release. That change gated gen_ai.usage.input_tokens on prompt_tokens > 0, which worked only against the streaming estimator — the one that hardcodes prompt tokens to zero. Core has a second estimator on the non-streaming path that derives prompt tokens from the request JSON length, so it produces a plausible non-zero count. The old gate published that invented number as a measurement; a test restoring it prints the invented 372.

  Now reads AgentEvent::Completed.usage_estimated (core 1.10) instead of inferring provenance from the value. A flag carries the fact; a heuristic guesses at it, and this is the second bug today caused by inferring provenance from a plausible-looking number.

  Adds:

  • gen_ai.usage.cost_source — gateway or estimated, set alongside cost_usd. Without it a billable figure is indistinguishable from a guess against a local price table, which matters on a metered SKU.
  • gen_ai.response.id — the gateway's chatcmpl-…, previously discarded at deserialization on all four LLM paths. It joins LiteLLM_SpendLogs.request_id, whose row carries the gateway's authoritative dollars and real token counts — so it matters most exactly when the counts on the span are absent.

  Requires core 1.10; the wire protocol is unchanged (the flags are telemetry-only, pinned by a key-count assertion).

@smooai/smooth-operator-web-chat-example@0.0.96

Patch Changes

• Updated dependencies [3da620b]
  • @smooai/smooth-operator@1.54.2