Starstruk Logo MarkStarstruk

2026.4.1

hkchengrex/MMAudio2026.4.1Apr 24, 2026by GabrielBianconi

AI Summary

This release introduces async observability writes to reduce latency, adds support for vLLM's reasoning field and TypeScript evaluators, and enhances UI capabilities with aggregated variant usage data and cost tracking in traces.

Key Highlights

  • Gateway defaults to async observability writes to reduce tail latency
  • Added TypeScript evaluators for inference evaluations
  • Added support for vLLM's new reasoning field
  • Added aggregated variant usage data (tokens, cost) to the UI
  • Added inference cost data to exported OpenTelemetry traces

Breaking Changes

  • Gateway defaults to async observability writes (configurable back to sync)

New Features

  • TypeScript evaluators
  • vLLM reasoning field support
  • Aggregated variant usage data in UI
  • Inference cost data in OpenTelemetry traces
  • Config option to include content in exported traces

Full Release Notes

> [!CAUTION]
> **Breaking Changes**
>
> - The gateway now defaults to async observability writes to reduce tail latency: inferences are sent to the client before they are persisted in the database. To restore the previous behavior, set `observability.async_writes = false`. **[[docs]](https://www.tensorzero.com/docs/gateway/configuration-reference)**

> [!WARNING]
> **Deprecations**
>
> - Removed the TensorZero Autopilot "Sessions" page from the UI. We recently added a TensorZero MCP that integrates nicely with coding agents, and we'll re-introduce advanced TensorZero Autopilot workflows in a platform-agnostic format soon.

**Bug Fixes**

- Return HTTP code 429 for rate limiting errors.
- Fixed a bug affecting ClickHouse database names with hyphens. (thanks @ianliuy!)

**New Features**

- Added TypeScript evaluators (for inference evaluations).
- Added support for vLLM's new `reasoning` field.
- Added aggregated variant usage data (tokens, cost, etc.) to the UI.
- Added inference cost data to exported OpenTelemetry traces. (thanks @kimsehwan96!)
- Added `export.otlp.traces.include_content` (default false) configuration field to include inference content (e.g. prompts, messages) in exported OpenTelemetry GenAI traces.

_& multiple under-the-hood and UI improvements_