Python agent release notesRSS

Notes

This release of the Python agent adds additional model information to LangChain Chain LLM events, and fixes a bug in LangGraph streaming instrumentation.

Install the agent using easy_install/pip/distribute via the Python Package Index, using conda via the Conda-Forge Package Index, or download it directly from the New Relic download site.

Improved Features

• Add additional model information to LangChain Chain LLM events

  • Add request.model and response.model attributes to LLM events produced by LangChain Chains.
  • This information is pulled on a best effort basis and may not always be available.

Bugs

• Fix TypeError in LangGraph streaming instrumentation

  • Previously, when using a subset of streaming APIs in LangGraph, the agent would encounter a TypeError due to the use of weakref.WeakMethod by LangGraph. This has been fixed, and the agent should now work correctly with those LangGraph streaming methods.

Support statement

We recommend updating to the latest agent version as soon as it's available. If you can't upgrade to the latest version, update your agents to a version no more than 90 days old. Read more about keeping agents up to date.

See the New Relic Python agent EOL policy for information about agent releases and support dates.

Notes

This release of the Python agent adds support for multiple new Redis commands, fixes a bug in LangGraph instrumentation, and improves handling and logging of harvest limit settings.

Install the agent using easy_install/pip/distribute via the Python Package Index, using conda via the Conda-Forge Package Index, or download it directly from the New Relic download site.

Improved Features

• Add instrumentation for new Redis v8.8.0 commands

  • Add instrumentation for the following new Redis commands from redis-py v8.8.0:
    • Misc commands: INCREX, XNACK
    • Array commands: ARCOUNT, ARDEL, ARDELRANGE, ARGET, ARGETRANGE, ARGREP, ARINFO, ARINSERT, ARLASTITEMS, ARLEN, ARMGET, ARMSET, ARNEXT, AROP, ARRING, ARSCAN, ARSEEK, ARSET

Bugs

• Fix crash in LangGraph instrumentation

  • Previously, when using LangGraph instrumentation, the agent would crash due to an issue with context propagation to jobs in AsyncBackgroundExecutor. This has been fixed, and the agent should now work correctly with LangGraph.

• Improved handling and logging of harvest limit settings

  • The agent now more accurately reports the list of harvest methods and limits to the Environment tab in New Relic One.
  • The agent now has more consistent debug logging on startup for harvest limit settings.

Support statement

We recommend updating to the latest agent version as soon as it's available. If you can't upgrade to the latest version, update your agents to a version no more than 90 days old. Read more about keeping agents up to date.

See the New Relic Python agent EOL policy for information about agent releases and support dates.

Notes

This release of the Python agent fixes a context propagation issue in LangChain and LangGraph, an OpenAI chat completion recording issue, and an issue when using OpenAI(https://pypi.org/project/openai) streaming through LiteLLM as a proxy.

Install the agent using easy_install/pip/distribute via the Python Package Index, using conda via the Conda-Forge Package Index, or download it directly from the New Relic download site.

Bugs

• Fix context propagation in LangChain and LangGraph

  • Adds additional transaction context propagation to LangChain and LangGraph to support a wider range of application structures.

• Fix an OpenAI chat completion recording issue

  • OpenAI reintroduced logic that closes the streaming response even if the consumer hasn't read all data. Previously, the chat completion data was only recorded when reaching the stream's end which resulted in chat completion data not being recorded at the closing of the stream. Logic has now been added to capture the chat completion data even if the stream has not been fully read.

• Fix an issue when using OpenAI streaming through LiteLLM as a proxy

  • Previously when using the OpenAI SDK connected through a LiteLLM proxy server it was possible to lose reporting of LLM response data due to a missing finish reason. This has been fixed, and the response should now always be recorded.

Support statement

We recommend updating to the latest agent version as soon as it's available. If you can't upgrade to the latest version, update your agents to a version no more than 90 days old. Read more about keeping agents up to date.

See the New Relic Python agent EOL policy for information about agent releases and support dates.

Notes

This release of the Python agent removes support for Python lambda function in the config file, excludes the newrelic header by default, adds support for exclude and include settings for Hybrid Agent traces, anthropic instrumentation, gemini response streaming, improves caching of IS_PYPY in gc_data.py for performance, and adds http.statusCode attribute in transaction events. This release also resends modules during reconnect ensuring they are recorded, guards garbage collection metrics against re-entrant calls, fixes task propagation in async generators and starlette context propagation in StreamingResponse.

Install the agent using easy_install/pip/distribute via the Python Package Index, using conda via the Conda-Forge Package Index, or download it directly from the New Relic download site.

Removed features

• Remove support for Python lambda functions in the config file

  • Previously, the agent allowed specifying a Python lambda function to dynamically name transactions, background task, database trace, profile trace, external trace url and method, function trace, generator trace, and memcache trace. Since support for this feature exposes a security vulnerability in the agent this feature is being removed.

Deprecations

• The response.status attribute in WebTransactions is considered deprecated and will be removed in the upcoming 14.0.0 release. It is being replaced by http.statusCode. We recommend replacing all references to response.status with http.statusCode.

New features

• Exclude the newrelic header by default

  • The newrelic distributed tracing header is now no longer sent by default. Only the W3C traceparent and tracestate headers will be sent.

• Add exclude and include settings for Hybrid Agent traces

  • opentelemetry.traces.include for the configuration file or NEW_RELIC_OPENTELEMETRY_TRACES_INCLUDE to be used as an environment variable
  • opentelemetry.traces.exclude for the configuration file or NEW_RELIC_OPENTELEMETRY_TRACES_EXCLUDE to be used as an environment variable

  The Hybrid Agent's default behavior will be different as well. Now, only libraries and tracers that are supported by OpenTelemetry but not already supported by New Relic will utilize OpenTelemetry's instrumentation. Otherwise, the agent will default to the New Relic instrumentation hooks.

  More information about these settings can be found in the Python Agent configuration page as well as the OpenTelemetry API support page.

• Add Anthropic streaming instrumentation

  • Adds instrumentation support for anthropic.

• Improve caching of IS_PYPY in gc_data.py for performance

• Now checks and caches IS_PYPY once when the module is loaded.

• Add Gemini response streaming

  • Add response streaming support to Gemini through the google-genai package.

• Add http.statusCode attribute to transaction events

  • To conform with the rest of the language agents, the Python agent now sends the http status to the WebTransaction attribute http.statusCode. The Python agent will continue to send the attribute to response.status as well, but the response.status attribute will be removed in the future.

Bugs

• Resend modules during reconnect

  • The agent previously only sent the module list periodically after the first connect cycle. If the agent reconnected to the collector, this data would not be resubmitted, so the UI would no longer show the module data. This has now been fixed.

• Guard garbage collection metrics against re-entrant calls

  • Adds a flag to prevent re-entrant calls to _GCDataSource.record_gc() to prevent infinite recursion on Python 3.13+.

• Enable task propagation in async generators

  • Fixes an issue where AsyncGeneratorProxy would not properly handle transaction context propagation if spread over multiple tasks.

• Fix Starlette context propagation in StreamingResponse

  • Adds trace context propagation to starlette StreamingResponse for both synchronous and asynchronous generators.

Support statement

We recommend updating to the latest agent version as soon as it's available. If you can't upgrade to the latest version, update your agents to a version no more than 90 days old. Read more about keeping agents up to date.

See the New Relic Python agent EOL policy for information about agent releases and support dates.

Notes

This release of the Python agent adds support for time to first token and fixes an incompatible Flask add_url_rule function signature.

Install the agent using easy_install/pip/distribute via the Python Package Index, using conda via the Conda-Forge Package Index, or download it directly from the New Relic download site.

New features

• Added time to first token for LLM response streams:

  • The agent records a new attribute,time_to_first_token on LlmChatCompletionSummary events for OpenAI and Bedrock response streams. This attribute measures the duration between the initial request and the first token received in the response.

Bugs fixes

• Resolved Flask add_url_rule function signature mismatch

  • Updated the instrumentation for Flask's to align its add_url_rule function signature with the native Flask implementation. This fix prevents potential application crashes caused by the previous signature mismatch.

Support statement

We recommend updating to the latest agent version as soon as it's available. If you can't upgrade to the latest version, update your agents to a version no more than 90 days old. Read more about keeping agents up to date.

See the New Relic Python agent EOL policy for information about agent releases and support dates.

Notes

This release of the Python agent removes support for Cross Application Tracing (CAT), Python 3.8, and the bundled SSL certificates. It also updates all vendored packages to their latest versions, and adds support for free-threaded CPython, Windows operating systems, and improved instrumentation for Microsoft AG2 (Autogen). Additionally, this release includes a new optional extra for installing SSL certificates through certifi. This release also fixes a bug in Strands Agents instrumentation which caused duplicate exceptions to be reported.

Install the agent using easy_install/pip/distribute via the Python Package Index, using conda via the Conda-Forge Package Index, or download it directly from the New Relic download site.

Removed features

• Cross Application Tracing (CAT) has been removed in favor of Distributed Tracing (DT).

  • Cross Application Tracing (CAT) was deprecated in agent version 7.0.0.166, and has been removed in this release. If you are still using Cross Application Tracing (CAT), we recommend migrating to Distributed Tracing (DT) as soon as possible to avoid disruption.

• The bundled SSL certificates have been removed.

  • If you require certificates for your environment, you can either install them via your operating system's package manager or install the agent with the optional newrelic[certificates] extra.

• Python 3.8 support has been removed.

  • Python 3.8 reached End-of-Life on 2024-10-07 and is no longer supported by the agent. Please consider upgrading your application to a supported version of Python in order to continue to receive updates to the Python Agent.

Security

• All internal packages updated to their latest versions.

  • urllib3 updated to v2.6.3
    • This fixes all open CVEs on urllib3 as of the date of this release.
  • opentelemetry_proto updated to v1.40.0
  • wrapt updated to v2.1.2
  • asgiref updated to v3.11.1

New features

• New newrelic[certificates] extra for installing SSL certificates through certifi.

  • The agent now provides an optional newrelic[certificates] extra which can be used to install the bundled SSL certificates. This is intended for users who require certificates but are unable to install them via their operating system's package manager.

• Free-Threaded CPython support

  • Starting with Python 3.13, CPython releases builds of Python with the Global Interpreter Lock (GIL) disabled, allowing for true multi-threading.
  • This release adds support for this behavior, allowing the agent to safely run in a free-threaded environment.
  • This release also includes pre-built binaries for free-threaded CPython on Linux and Windows, allowing for easy installation in these environments.

• Windows Support

  • This release includes pre-built binaries for CPython 3.13 or higher on Windows.
  • Note that Python 3.12 or lower is not officially supported on Windows due to large variances in the resolution of timers. It's still possible to install and use the agent from source, but users may experience issues with inaccurate timing data and should upgrade to Python 3.13 or higher for the best experience.

• Improved instrumentation for Microsoft AG2 (Autogen)

  • Added support for streaming responses through the asynchronous generator method on_messages_stream().
  • Instrumentation now includes a subcomponent attribute on spans that represent tool or agent calls. This is used to provide more enhanced entity mapping in the New Relic UI.

Bugs

• Duplicate exceptions in Strands Agents

  • After an update to Strands Agents which encapsulates some exceptions in a new EventLoopException type, duplicate errors could be reported. This has been fixed and the underlying error should only be reported once.

Support statement

We recommend updating to the latest agent version as soon as it's available. If you can't upgrade to the latest version, update your agents to a version no more than 90 days old. Read more about keeping agents up to date.

See the New Relic Python agent EOL policy for information about agent releases and support dates.

Notes

This release of the Python agent adds trace/span support for Hybrid Agent, New Relic's integration with OpenTelemetry. It also adds entity GUIDs to Agent Control health check files, subcomponent attributes to agentic AI instrumentation, support for new Redis client methods, fixes a bug where there may be duplicate ImportHookFinder instances and a bug in OracleDB Cursor.callproc() instrumentation.

Install the agent using easy_install/pip/distribute via the Python Package Index, using conda via the Conda-Forge Package Index, or download it directly from the New Relic download site.

Features

• Add Hybrid Agent tracing support

  • Hybrid Agent is designed to bridge the gap for teams looking to adopt OpenTelemetry standards while maintaining backwards compatibility with their existing New Relic dashboards. Hybrid Agent offers a "best-of-both-worlds" experience by combining New Relic’s deep visibility with OpenTelemetry API compatibility. It provides an observability solution with integrated support for the following OpenTelemetry APIs:
    • Tracing
    • Standalone library and native framework instrumentation
  • Key enhancements like Span Links and Events on Spans are now available, ensuring seamless interoperability in mixed-mode environments.

• Introduces a new Trace ID Ratio based sampler

  • A new Trace ID Ratio based sampler. This sampling algorithm is based off the sampler of the same name from OpenTelemetry, and can be configured to sample a set percentage of transactions.

• Add subcomponent attributes to agentic AI instrumentation

  • Instrumentation for LangChain, MCP, and Strands now includes a subcomponent attribute on spans that represent tool or agent calls. This is used to provide more enhanced entity mapping in the New Relic UI.

• Add entity GUID to Agent Control health check files

  • When the agent is started within an Agent Control environment, a health check file is created at the configured file location for every agent process. This file now includes the GUID of the entity related to the agent when available.

• Add instrumentation for new Redis methods

  • Add instrumentation for the following methods new methods on Redis clients:
    • hotkeys_get
    • hotkeys_reset
    • hotkeys_start
    • hotkeys_stop
    • vrange
    • xcfgset

Bugs

• Fix duplicate ImportHookFinder instances

  • Previously it was possible for duplicates of the import hook finder to be added to sys.meta_path.

• Fix incorrect signature on OracleDB Cursor.callproc()

  • Previously attempting to use keyword arguments with Cursor.callproc() would cause crashes as the argument names were incorrect in our instrumentation. This has been fixed, and the additional argument keyword_parameters is now also supported.

Support statement

We recommend updating to the latest agent version as soon as it's available. If you can't upgrade to the latest version, update your agents to a version no more than 90 days old. Read more about keeping agents up to date.

See the New Relic Python agent EOL policy for information about agent releases and support dates.

Notes

This release of the Python agent adds support for LangGraph agents.

Install the agent using easy_install/pip/distribute via the Python Package Index, using conda via the Conda-Forge Package Index, or download it directly from the New Relic download site.

Features

• Add support for LangGraph agents

  • Adds instrumentation for LangGraph agent invocations via create_agent() in LangChain. This includes agent calls to invoke(), ainvoke(), stream(), and astream().

Support statement

We recommend updating to the latest agent version as soon as it's available. If you can't upgrade to the latest version, update your agents to a version no more than 90 days old. Read more about keeping agents up to date.

See the New Relic Python agent EOL policy for information about agent releases and support dates.

Notes

This release of the Python agent adds new ways of passing errors to notice_error(), adds support for Strands-Agent synchronous tool error capturing, adds support for gpt-5 and newer OpenAI LLM models, and fixes a crash in Azure Functions' utilization logic.

Install the agent using easy_install/pip/distribute via the Python Package Index, using conda via the Conda-Forge Package Index, or download it directly from the New Relic download site.

Features

• Add support for exception instances as arguments to notice_error()

  • notice_error() can now accept a caught exception instance (such as one from a try/except block) for the error argument, rather than an exception tuple from sys.exc_info().

  • This is in addition to and does not replace the existing behavior. Previous usage of notice_error() with zero arguments or with an exception tuple should still function correctly.

• Add support for gpt-5 and newer OpenAI LLM models

  • Add updates to OpenAI instrumentation to support newer attribute names used with later OpenAI models

• Add support for Strands-Agent synchronous tool error capturing

  • Synchronous Python functions used as tools in Strands-Agents now have proper error capturing, and capture traces within tools such as database lookups.

Bug fixes

• Guard against crashes in Azure Functions utilization

  • Previously, some users experienced edge cases where Azure Functions utilization detection would crash when attempting to match regular expressions. The crash has been fixed, and should no longer be possible in the future.

Support statement

We recommend updating to the latest agent version as soon as it's available. If you can't upgrade to the latest version, update your agents to a version no more than 90 days old. Read more about keeping agents up to date.

See the New Relic Python agent EOL policy for information about agent releases and support dates.