# mlflow
pip install git+https://github.com/mlflow/mlflow.git@refs/pull/19920/merge
# mlflow-skinny
pip install git+https://github.com/mlflow/mlflow.git@refs/pull/19920/merge#subdirectory=libs/skinny

%sh curl -LsSf https://raw.githubusercontent.com/mlflow/mlflow/HEAD/dev/install-skinny.sh | sh -s pull/19920/merge

Related Issues/PRs

What changes are proposed in this pull request?

Support distributed tracing.

Motivation:
Currently, MLflow can only create a trace from single process/service. However, multi-agent system is becoming more popular, and users can deploy multiple agents as separate services and use them within a single product. For example, they can have "SupervisorAgent", "SQLAgent", "PythonAgent", "WebSearchAgent", deployed into different model serving endpoints, and route requests from chatbot frontend based on user's question. In this case, it is ideal for customers that they can see all those agent actions for a single user request in a single trace. However, since MLflow can only create trace within a scope of process, they can only create separate trace per agent. Debugging issue is painful because they need to find and stitch those traces.

In this PR,
it supports building a trace that spans over multiple services. Note that OpenTelemetry already have a mechanism for that, which is called "context propagation" and use a protocol called W3C TraceContext by default. So that this PR is built on top of it.

APIs:

def get_tracing_context_headers_for_http_request():
    """
    Get the http request headers that hold information of the tracing context.
    The trace context is serialized as the traceparent header which is defined
    in the W3C TraceContext specification.
    For details, you can refer to
    https://opentelemetry.io/docs/concepts/context-propagation/
    and
    https://www.w3.org/TR/trace-context/#traceparent-header

    Returns:
        The http request headers that hold information of the tracing context.
    """

def set_tracing_context_from_http_request_headers(headers):
    """
    Extract the trace context from the http request headers,
    and return the context manager to set the extracted trace context as the
    current trace context.
    The trace context must be serialized as the traceparent header which is defined
    in the W3C TraceContext specification, please see
    :py:func:`mlflow.tracing.distributed.get_tracing_context_headers_for_http_request`
    for how to get the http request headers.

    Args:
        headers: The http request headers to extract the trace context from.
    """

How is this PR tested?

• Existing unit/integration tests
• New unit/integration tests
• Manual tests

Does this PR require documentation update?

• No. You can skip the rest of this section.
• Yes. I've updated:
  • Examples
  • API references
  • Instructions

Release Notes

Is this a user-facing change?

• No. You can skip the rest of this section.
• Yes. Give a description of this change to be included in the release notes for MLflow users.

Support distributed tracing

What component(s), interfaces, languages, and integrations does this PR affect?

Components

• area/tracking: Tracking Service, tracking client APIs, autologging
• area/models: MLmodel format, model serialization/deserialization, flavors
• area/model-registry: Model Registry service, APIs, and the fluent client calls for Model Registry
• area/scoring: MLflow Model server, model deployment tools, Spark UDFs
• area/evaluation: MLflow model evaluation features, evaluation metrics, and evaluation workflows
• area/gateway: MLflow AI Gateway client APIs, server, and third-party integrations
• area/prompts: MLflow prompt engineering features, prompt templates, and prompt management
• area/tracing: MLflow Tracing features, tracing APIs, and LLM tracing functionality
• area/projects: MLproject format, project running backends
• area/uiux: Front-end, user experience, plotting, JavaScript, JavaScript dev server
• area/build: Build and test infrastructure for MLflow
• area/docs: MLflow documentation pages

How should the PR be classified in the release notes? Choose one:

• rn/none - No description will be included. The PR will be mentioned only by the PR number in the "Small Bugfixes and Documentation Updates" section
• rn/breaking-change - The PR will be mentioned in the "Breaking Changes" section
• rn/feature - A new user-facing feature worth mentioning in the release notes
• rn/bug-fix - A user-facing bug fix worth mentioning in the release notes
• rn/documentation - A user-facing documentation change worth mentioning in the release notes

Should this PR be included in the next patch release?

Yes should be selected for bug fixes, documentation updates, and other small changes. No should be selected for new features and larger changes. If you're unsure about the release classification of this PR, leave this unchecked to let the maintainers decide.

• Yes (this PR will be cherry-picked and included in the next patch release)
• No (this PR will be included in the next minor release)