🦋 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@1.53.1

Patch Changes

• e0077e4: Stop exporting fabricated token counts on the turn span, and adopt the cross-language cost_unavailable attribute.

  gen_ai.usage.input_tokens was published whenever either count was non-zero. That || was the whole bug: when the gateway drops the usage chunk, smooai-smooth-operator-core's collect_stream fabricates the struct — prompt_tokens hardcoded to 0 and completion_tokens estimated as content.len() / 4 — so the fabricated struct always had a non-zero completion count, and the || published input_tokens = 0 beside it. LiteLLM drops that chunk for smooth-* aliases, so this was the common path, not an edge case: no streamed turn has ever exported a measured token count.

  The estimated output count looked plausible precisely because it is derived from the reply text — an estimate computed from the output cannot help but track the output. Only the zeroed input half looked obviously wrong.

  Now gated on prompt_tokens > 0 alone: both counts are exported, or neither. Absent is honest; 0 is a lie. The underlying fabrication still needs fixing in core.

  Also adds smooai.gen_ai.cost_unavailable = "unpriced", set instead of gen_ai.usage.cost_usd when no cost could be established — the same attribute name and value the TypeScript emitters use, so a consumer never has to special-case per engine.

@smooai/smooth-operator-web-chat-example@0.0.93

Patch Changes

• Updated dependencies [e0077e4]
  • @smooai/smooth-operator@1.53.1