Arize Phoenix

Version 17.21.0 introduces several UI enhancements, including a global search palette and session statistics panel, alongside backend improvements for PXI tracing and session cleanup.

This release introduces user-selected metric charts in the UI and includes several bug fixes related to agent error recovery, token prices, and UI stability.

This release introduces new application features, including trace tree search and a consolidated account menu in the side navigation.

Version 17.18.0 introduces UI consistency improvements, specifically regarding disclosure arrows in the application.

This release introduces new UI features, including shift-click range selection for tables and recent searches in the span filter typeahead, alongside a bug fix to restrict range selection to same-depth rows.

This release introduces significant feature enhancements across the API, CLI, and UI, including new authentication discovery, experiment tagging, and improved data visibility in tables. Several bug fixes address issues related to annotation handling and UI synchronization.

Phoenix version 2.12.0 introduces new REST endpoints for managing annotation configurations and dataset labels, alongside a fix for the async client timeout policy.

Version 17.15.0 introduces support for Claude Sonnet 5 in the playground and includes several bug fixes related to cost calculation, Bedrock tool handling, and file path rendering on Windows. It also updates the internal client dependency.

This release introduces an annotation summary feature in the UI and fixes an issue related to preserving generator inputs during dataset creation and example addition.

This release introduces an annotation summary feature in the UI and fixes a bug related to Anthropic prompt caching in agents.

Phoenix version 17.13.0 introduces a new CLI TUI, a pytest plugin for evaluation CI, and updates built-in model token prices. Several minor bug fixes and dependency updates were also included.

This release introduces significant enhancements to agent capabilities, including session context and experiment editing, alongside new features like a direct server agent endpoint and a pytest plugin for CI.

Phoenix version 17.12.0 introduces several enhancements for agents, including user ID attachment and UI fixes, alongside new features like streaming subagent progress and a CI eval testing API for the JavaScript client.

This release introduces several new features including enhanced agent interaction and improved trace visualization. A critical bug fix addresses dependency conflicts by capping pydantic-ai-slim below version 2.0.

This release introduces the feature to display trace-level annotations within the trace header and updates documentation regarding Go OpenInference integrations.

Version 17.9.0 introduces significant enhancements to agents, including server-side bash capabilities and new configuration options. It also brings improvements to prompt/dataset management and UI display elements.

Patch release 17.8.1 primarily addresses a bug related to route path resolution within the prometheus middleware when using FastAPI 0.137.

Version 17.8.0 introduces an Agent GraphQL skill and improves theme handling for background colors. A bug fix standardized icons across the application.

This release introduces new token detail metrics charts and shareable trace time range URLs. It also includes a bug fix for trace ID encoding in UI routes and adds session context for agents.

Phoenix 17.6.0 introduces new agent capabilities, enhanced metrics visualization with time series annotations, and improved UI controls for time range navigation.

This release introduces new features for agent configuration and UI time range selection, alongside numerous bug fixes including dependency updates and improvements to error handling and agent behavior.

Version 2.9.0 introduces new playground tools for repetitions and experiment recording, along with an invokable skill menu for agents. A key bug fix ensures correct experiment example resolution by node GlobalID.

Version 17.4.0 introduces new features for agents, dataset evaluation management in PXI, and improved time range selection in the UI, alongside several bug fixes related to token pricing and GraphQL performance.

This release introduces significant enhancements to the playground and agent capabilities, including new PXI tools for experiment recording, dataset management, and support for the claude-fable-5 model. Several bug fixes address UI alignment and server claim extraction issues.

Version 2.8.0 introduces significant features around PXI playground tooling, agent capabilities like span annotation and native web search, and new sandboxing/code evaluation features. A minor bug fix addresses handling of top_k parameters for Google providers.

Version 17.2.0 introduces the PXI route info tool and includes bug fixes related to chat history scoping and prompt table refreshing.

This release introduces several new features for the PXI agent and playground, including dataset loading and LLM-evaluator authoring, alongside a fix to prevent initialization failures from halting server startup.

This release introduces several new features including model switching in the PXI playground, a web search toggle for agent chat, and code-evaluator authoring. Documentation for Go LLM provider integration was also added.

Version 16.5.0 introduces significant enhancements to agent capabilities, including new chat controls and prompt tooling, alongside several bug fixes for error handling and UI rendering.

This release introduces significant enhancements to the PXI agent experience, including Claude Opus 4.8 support and dynamic quick actions, alongside UI improvements like styled code blocks and better chart handling.

This release introduces new UI features like drag-to-zoom on charts and enhances the PXI agent capabilities by adding skills support and expanding the settings model selector.

This release introduces improved agent accessibility via modal overlays and fixes an issue related to token count confinement during ingestion.

Version 16.1.0 introduces several UI and agent enhancements, including a new eval dataset skill and floating assistant panel in PXI, alongside bug fixes for token counting and model 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 15.12.0 introduces new features for agent tool metadata and UI interaction, alongside several bug fixes addressing scrolling, routing, playground configuration, and run visualization.

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

This release introduces generative UI rendering capabilities within agent chats and includes several bug fixes related to server configuration and build processes.

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

Version 15.10.0 introduces several UI and agent enhancements, including a trace feedback toolbar and curated agent menus, alongside updates to tracing attribute standards and bug fixes for metadata layout and token prices.

This release introduces significant updates to agent functionality, including new API endpoints and playground tools, alongside improvements to client-side tracing and REST API payload details. Notably, the v1 /chat route has been removed.

Version 15.9.0 introduces feature enhancements for agent tool span tracking and resolves bugs related to agent consent and GraphQL node handling.

This release introduces new controls for Anthropic and Google in the playground, improves agent tracing by populating project sessions, and includes several bug fixes related to token prices and security.

Version 15.7.0 introduces user feedback annotations, session tagging improvements, and updates the UI toast behavior. This release also removes the deprecated v1 agent /chat route.

This release introduces new agent session summary endpoints and playground preference settings. It also fixes an issue related to interrupted tool calls.

Patch release 15.5.1 primarily addresses a bug fix related to preserving built-in evaluators during dataset evaluator deletion.

This release introduces experimental support for /chat-v2 with tools and improves frontend REST call typing. It also addresses dependency vulnerabilities by updating litellm.

Version 15.4.0 introduces significant enhancements to agents, adds filter-based DELETE endpoints via the API, and includes token counts in REST API payloads. A bug related to trace selection in the session view was also resolved.

Version 15.3.0 introduces feature enhancements like removing user instructions from PXI and adding an empty state shader for chat, alongside dependency updates and documentation improvements.

Version 3.1.0 introduces server-side prompt assembly and caching for agents, alongside improvements to prompt handling in evals and a fix for model capability listings.

This release primarily updates documentation by adding LinkedIn social links alongside existing X/Bluesky links.

This release introduces TanStack AI tracing integration and moves PXI prompt assembly server-side for agents. It also includes several bug fixes, notably addressing an SSRF vulnerability in the agent chat endpoint.

Version 15.1.0 introduces new features across API, CLI, and phoenix-client, including annotation querying, auth profile management, and TS trace annotations, alongside several bug fixes.

This release introduces dataset upsert functionality and adds an API endpoint for session notes. It also includes a breaking change related to dataset upsert implementation.

Version 15.0.0 introduces dataset upsert capabilities and extends database prompt types for invocation parameters.

Version 14.17.0 introduces enhanced agent configuration flexibility and a new API endpoint for session notes, alongside several bug fixes including updated token prices and context leak prevention.

Version 14.16.0 introduces new agent context advertising and server-side GraphQL field redaction. It also includes several bug fixes related to Relay query ownership and token price updates.

Version 14.15.0 introduces new UX toggles for sessions and adds a new skills audit feature. A critical bug fix addressed an SSRF vulnerability by requiring authentication for custom provider credential tests.

This release introduces a new REST endpoint for trace notes within the Phoenix CLI, enhancing command-line interaction capabilities.

Version 14.14.0 introduces trace note support to the px command-line interface.

This release introduces a dedicated span notes column in the UI and adds a trace note REST endpoint for the Phoenix CLI. It also updates built-in model token prices.

This release introduces new exports within the phoenix-otel package, specifically re-exporting openinference helpers, decorators, and semconv.

This release introduces new tracing visualization features, including trace annotations and span aggregation, alongside several application and data handling bug fixes.

This release introduces new features for agent session evaluation and secrets management, alongside a critical dependency update migrating `authlib.jose` to `joserfc`.

Version 2.4.0 introduces the ability to pass trace_id to experiment evaluators and adds type-aware attribute filtering to the /v1/spans REST API endpoint.

Version 14.10.0 introduces token count display for chat sessions and resolves dependency pinning and configuration alignment issues.

Patch release v14.9.1 primarily addresses a database-related bug fix concerning Azure AsyncEngine handling.

Version 14.9.0 introduces new PXI controls, adds Claude Opus 4.7 support, and includes several UI and API enhancements alongside minor bug fixes.

Version 14.8.0 introduces new database connectivity features, specifically enabling Azure managed identity authentication for PostgreSQL.

Version 14.7.0 introduces session pagination and trace annotation dataloader support, alongside UI improvements for the projects page refresh cycle.

Phoenix version 14.6.0 introduces an evaluation harness for agents and updates the application layout for session turns. This release also includes CI improvements and better tracking of evaluation failures.

Release 14.5.0 introduces new features for chat metadata, CLI annotations, and agent controls, alongside a fix for malformed pxi skill YAML files.

Version 14.4.0 introduces a new capability registry for agent tooling and resolves an issue by unpinning the graphql-core dependency.

Version 14.3.1 primarily focuses on a security enhancement by hardening the authentication mechanism during dataset uploads.

Version 14.3.0 introduces an assistant agent settings page and updates internal dependency handling for strawberry-graphql compatibility.

This patch release removes WebSocket support for GraphQL subscriptions, migrating to HTTP multipart for better compatibility and stability.

Version 14.2.0 introduces a new API endpoint for accessing projects by name and includes documentation updates.

Version 14.1.1 primarily addresses a bug related to connection exhaustion by implementing dataloader for session annotations.

Phoenix version 14.1.0 primarily focuses on bug fixes, including improvements to OpenAPI examples, agent error surfacing, and environment variable handling for SQL schema.

This patch release deprecates the evals 1.0 API and removes the associated legacy experiments module.

Version 3.0.0 introduces a major overhaul to the evals system by deprecating and removing the legacy evals 1.0 API and the associated experiments module.

Version 14.0.0 introduces major architectural changes, including a subcommand-first CLI structure and the removal of legacy evaluation and client plumbing. Key features include enhanced agent tooling, PostgreSQL read replica support, and ephemeral experiment handling.

This release introduces a new utility for converting ATiF traces into trajectory format.

Version 13.23.0 introduces bundled TypeScript SDK documentation in npm packages and resolves a bug with the Playground button in the prompts table.

Patch release 13.22.2 primarily addresses a database query optimization within the ProjectHasTracesDataLoader.