Skip to content

Opik 2.2.119 moves playground experiments to the server

Updated Opik

Summary

Regular-dataset playground experiments can run and stop server-side, with fixes for results, retries and tool-call tracing.

Tool
Opik
Category
Agent infra
Date
Pricing
Not stated in the sources we cited.

Opik published 2.2.119 on October 23, 2026.

Server-side playground runs

The release adds running and stopping regular-dataset playground experiments on the server. Related fixes preserve a run's results on screen and settle a stop when the server has completed it. A backend fix processes a redelivered experiment item once.

The stated scope is regular datasets. These notes do not establish that every experiment type moved to the same execution path, or provide a benchmark for faster evaluation.

Tracing and evaluation fixes

The SDK now reports how many scores an aggregate left out. The TypeScript SDK also counts Gemini thinking and tool-use prompt tokens, and its OpenAI integration keeps streamed parallel tool calls apart.

Those details matter when reviewing experiment results and traces, but they do not prove that the evaluated application itself became more accurate. Teams depending on existing run and stop behavior should review the official changes before upgrading.

See the Opik hub.

More from Opik

View all

Questions

What is Opik?

According to its docs, Opik is an open-source platform that helps you understand what your LLM application is doing, measure how well it works, and improve it.

Is Opik open source?

Yes. The docs say Opik is open source, and the GitHub API reports the repository license as Apache-2.0.

Does Opik have an MCP server?

The docs say you can install the Opik MCP server to drive your workspace from tools such as Claude Code, Cursor, VS Code Copilot, Codex or opencode.

Filed on Agent Radar. Claims in this note are limited to the sources above.