agents

This release introduces significant enhancements to the Factory control plane, including durable workspaces, SSH command affordances, and improved terminal streaming. It also includes various bug fixes and performance optimizations across the platform.

This release introduces new features like in-process Phoenix MCP toolset for agents, retrieval relevance evaluator, and AI Query for trace filter DSL. It also includes several bug fixes and dependency updates.

This release introduces new features for agent tracing, batch annotation configuration, and playground error display. It also includes several bug fixes across agents, PXI, UI, and database components.

This release introduces new features for agents, including approval-gated annotation config tools and span-coding. The UI has been enhanced with a new Metrics tab for dataset pages and refined experiment metric charts. Several UI bug fixes improve the user experience, such as persistent filters on project tables and resizable trace columns.

This release introduces the migration of PXI evals to a pytest plugin and includes bug fixes for agent tool calls and cost tracking by updating model token prices.

Version 16.0.0 introduces powerful Sandboxed Code Evaluators for flexible evaluation strategy composition and enables native web search/fetch for agents. This release contains breaking changes requiring migration.

Version 15.11.1 primarily addresses a bug fix related to PXI agent system prompt instrumentation on LLM spans.

This patch release fixes two bugs related to agent span persistence and PXI cache reporting in the application, alongside documentation cleanup.

This release introduces a local-network-access permission for Chromium, renames the agents module to test agents (a breaking change), adds a cwd fallback for test config, and fixes a VS Code mcp workspaceFolder reference.