Released 2026-11-13 · Compare with 1.0.0-rc.6
Features
- Decision spans now name their turn and purpose, and record per-question results as
gen_ai.decision.resultswhen the trace records outputs.
decision <model>spans carryagent.turn.idandagent.decision.purpose(model-selection,tool-approval,agent-routing, orcustom), so Agent Runs can place a model-selection decision in a turn that is still running. Results follow the proposed OpenTelemetry GenAI decision conventions: choices, predicates, scores with their levels, and refusals, each keyed by the question's position. They are output content, recorded only when outputs are and dropped for any destination that declines outputs.AgentRun.decisionaccepts an optionalpurpose, and completing a decision withoutputrecords it as the results. (#6389)
Bug Fixes
- Update eve's bundled dependencies, including the AI SDK, Vercel SDKs, Chat SDK, and MCP, to their latest releases, and OpenTelemetry to its latest 2.x releases.
eve now requiresai7.0.172 or later.
YAML frontmatter, eval data files, and OpenAPI specs now parse with js-yaml 5 and YAML 1.2 values. An unquoted date such as2024-01-02loads as a string instead of aDate, and a0bbinary literal such as0b101loads as a string instead of a number. The YAML 1.1 tags!!timestamp,!!binary,!!set,!!omap, and!!pairsare no longer recognized, and a mapping with a non-scalar key, such as? [a], throws.<<merge keys still work.
MCP connections over Streamable HTTP now fail when theserver/discoverrequest fails before the server answers, instead of retrying withinitialize. SetprotocolVersionDiscovery: falsefor a server that can't handle that request. Servers that answerserver/discoverwith a JSON-RPC error or an HTTP error status other than401still useinitialize, and servers that reject the protocol version with a supported legacy version still negotiate it. (#6378)