# What changed in @arizeai/openinference-vercel

> 5 changes in @arizeai/openinference-vercel 3 from its official release notes, upgrade guides and advisories: 2 new APIs and options. Each one cited. Sources checked 2026-10-21. HTML: https://unstale.sh/npm/@arizeai/openinference-vercel/

To see which of these apply to a repository, run `npx @unstale/cli init` in it.

## @arizeai/openinference-vercel 3

### New APIs and options

- A new callback for OpenTelemetry's enrichSpan option classifies the outer decide operation as a CHAIN and avoids duplicating model identity and usage. (3.3.0) — [release notes](https://github.com/Arize-ai/openinference/releases/tag/%40arizeai/openinference-vercel%403.3.0)
- AI SDK v7 decision model calls are recorded as DECISION spans with decision.* model, provider, system, and token count attributes. (3.3.0) — [release notes](https://github.com/Arize-ai/openinference/releases/tag/%40arizeai/openinference-vercel%403.3.0)

### Behavior and default changes

- GenAI model and usage attributes are removed from decision spans after mapping, preventing GenAI converters from re-labeling the decision model as an LLM. (3.3.0) — [release notes](https://github.com/Arize-ai/openinference/releases/tag/%40arizeai/openinference-vercel%403.3.0)
- A session.id already on a span takes precedence over the value derived from gen_ai.conversation.id. (3.2.3) — [release notes](https://github.com/Arize-ai/openinference/releases/tag/%40arizeai/openinference-vercel%403.2.3)
- agent.step, agent.action, and agent.approval spans are classified as CHAIN instead of LLM; agent.action spans for subagent or remote-agent calls remain AGENT. (3.2.3) — [release notes](https://github.com/Arize-ai/openinference/releases/tag/%40arizeai/openinference-vercel%403.2.3)