NewYour coding agent can read the release notes before it upgrades.Set up the MCP server →
PyPI · #3171 most downloaded on PyPI
AI Observability and Evaluation
Last release today
23 Sep 2026
Ships on a steady schedule
a new release about every 8 days
Nearly every release is documented
notes for 60 of the last 60 stable releases
19 versions withdrawn
withdrawn after publishing
4 years old
697 releases · first in 2023
One column per quarter.
Display tool call and tool result Ids in span details
update Gemini models for playground
route user to forgot-password page in welcome email url
allow project name as identifier in REST path for projects endpoints
add PHOENIX_ALLOWED_ORIGINS env to phoenix server
add aiohttp to container for azure-identity
add REST endpoints to list or create prompt version tags
Allow no working dir configured w/ postgres
UI: use icon only buttons in span details when view is small
perf: add a toggle to treat orphans as root
### Features * ui: use skeleton loader
environment variable for inserting admin users at startup
add delete experiment to action menu
Add resize capability to Span, Trace, and Session table
make azure api key optional for playground
### Bug Fixes * ui: heading size
components: Add react-aria Tabs components
be more careful about metadata structure
add none as option for tool choice for anthropic 0.49.0
add model validator for template type
add anthropic thinking config param for python client
Implement quick filtering support for metadata cells in trace table
Improve performance on project tables
perf: show + n more spans in traces table
graphql: query to get number of spans for each trace
performance: limit depth (defaults to 3) on descendant spans query
admin: introspection into DB usage
client: typedoc docs gen for phoenix-client
don't flex the trace tree toolbar
suppress token icons in trace tree when count is zero
performance: don't fetch new traces when the traces slideover is visible
Allow configuration of postgres username/password/db/host using standard env vars
### Bug Fixes * also reject on template_type (#6468) (02a9668) ### Documentation * Update README.md
Add previous_version resolver on PromptVersion
previous_version resolver on PromptVersion (#5933) (d7dfe70)/prompts/{id}/versions (#6098) (87ecece)strip leading and trailing whitespace characters from values of string environment variables
prompts: trim leading and trailing whitespace from mustache template variables
allow prepared_statement_cache_size for asyncpg
dab09f1 : Add filter expressions to getTraces and listSessions (requires Phoenix server >= 20.12.0). Deprecate the individual trace error and latency…
filter expressions to getTraces and listSessions (requires Phoenix server >= 20.12.0). Deprecate the individual trace error and latency parameters while retaining their behavior and support for server >= 20.8.0.2b73847 : Accept metadata in promptVersion() so prompt versions can carry arbitrary JSON metadata.
metadata in promptVersion() so prompt versions can carry arbitrary JSON metadata.playground: if base url is provided but not api key, use fake api key for openai client
playground: disable retry and auth refresh if auth is disabled
Replace corepack with pinned pnpm version in Dockerfile
c839254 : Add error , minLatencyMs , and maxLatencyMs filters to getTraces , matching the query parameters on GET /v1/projects/{id}/traces (requires P
error, minLatencyMs, and maxLatencyMs filters to getTraces, matching the query parameters on GET /v1/projects/{id}/traces (requires Phoenix server >= 20.8.0).graphql: add default values for TimeRange input
upgrading wrapt to 1.17.2 to fix cross-platform issues
incorporate headers from environment for the trace span exporter when running experiments
Allow experiment table resize for large inputs
bf047af : Add Meta (Muse Spark models) as a built-in OpenAI-compatible model provider ( META ).
META).reduce query memory usage in DatasetExampleRevisionsDataLoader
58b7017 : Add Z.ai (GLM models) as a built-in OpenAI-compatible model provider ( ZAI ).
ZAI).allow import of LLMRelationalEvaluator
1cdff14 : Expose cumulative prompt, completion, and total token counts on the high-level Session objects returned by getSession and listSessions .
Session objects returned by getSession and listSessions.37916d7 : Add setProjectRetentionPolicy to the projects entry point for assigning an existing retention policy by GlobalID or resetting a project to t
setProjectRetentionPolicy to the projects entry point for assigning an existing retention policy by GlobalID or resetting a project to the default policy.transferTraces to the traces subpath for moving traces between Phoenix projects.getCurrentUser helper through the @arizeai/phoenix-client/users entrypoint.openai peer dependency range to ^6.10.0 || ^7.0.0 so apps already on the OpenAI SDK v7 can install the client without a peer conflict.playground: Support anyOf json schema
Your coding agent can read these notes before it upgrades. Set up the MCP server →