MLFlow Release Finishing Touches (#2359)

* add missed mkdocs link

* Fix MLflow integration issues with TruLens scorers

1. Fix AttributeError when instrumenting LiteLLM endpoint

   The _instrument_class method was failing with:
   AttributeError: 'CallTypes' object has no attribute '__name__'

   This occurred because the method tried to wrap non-callable attributes
   (like enum values) that happened to have the same name as the target
   method. Added a callable() check to skip non-callable attributes.

2. Suppress noisy third-party library warnings

   Added warning filters to suppress:
   - pkg_resources is deprecated warning from munch library
   - python-dotenv could not parse statement warnings/logs

   These warnings were cluttering output when using TruLens scorers
   via MLflow GenAI evaluation framework.

.... Generated with [Cortex Code](https://docs.snowflake.com/user-guide/snowflake-cortex/cortex-agents)

Co-Authored-By: Cortex Code <noreply@snowflake.com>

* Add MLflow + TruLens scorers example notebook

New example notebook demonstrating how to use TruLens feedback functions
as first-class scorers in MLflow GenAI evaluation framework (MLflow 3.10+).

Covers:
- RAG evaluation scorers (Groundedness, ContextRelevance, AnswerRelevance)
- Output scorers (Coherence)
- Agent trace scorers (ToolSelection, ToolCalling)
- Batch evaluation with mlflow.genai.evaluate
- Threshold configuration and multi-provider support

.... Generated with [Cortex Code](https://docs.snowflake.com/user-guide/snowflake-cortex/cortex-agents)

Co-Authored-By: Cortex Code <noreply@snowflake.com>

* Update MLflow + TruLens scorers example notebook

Removed redundant Best Practices section to keep notebook focused on
demonstrating scorer usage.

.... Generated with [Cortex Code](https://docs.snowflake.com/user-guide/snowflake-cortex/cortex-agents)

Co-Authored-By: Cortex Code <noreply@snowflake.com>

* Add agent evaluation examples to MLflow integration docs

Added new "Agent Evaluation" section covering:
- Batch agent evaluation with predict_fn
- Evaluating individual agent traces
- Note clarifying when to use agent vs RAG scorers

.... Generated with [Cortex Code](https://docs.snowflake.com/user-guide/snowflake-cortex/cortex-agents)

Co-Authored-By: Cortex Code <noreply@snowflake.com>

* Rename 'Running Feedback Functions' to 'Running Metrics'

---------

Co-authored-by: sfc-gh-jreini <sfc-gh-jreini@users.noreply.github.com>
Co-authored-by: Cortex Code <noreply@snowflake.com>

Author: sfc-gh-jreini

docs/component_guides/evaluation/mlflow.md

@@ -176,6 +176,86 @@ scorer = Groundedness(model="openai:/gpt-4o")
 feedback = scorer(trace=trace)
 ```
 
+## Agent Evaluation
+
+Agent GPA scorers evaluate tool selection and execution in agentic workflows. These scorers require traces since they inspect tool call spans.
+
+### Batch Agent Evaluation
+
+Use `predict_fn` with `mlflow.genai.evaluate` to trace and evaluate agent runs:
+
+```python
+import mlflow
+from mlflow.genai.scorers.trulens import (
+    Groundedness,
+    ToolSelection,
+    ToolCalling,
+    Coherence,
+)
+
+mlflow.openai.autolog()
+
+
+def run_agent(inputs: dict) -> str:
+    response = openai.chat.completions.create(
+        model="gpt-4o",
+        messages=[{"role": "user", "content": inputs["user_query"]}],
+        tools=[...],  # your tool definitions
+    )
+    # ... handle tool calls and return result
+
+
+agent_queries = [
+    "What's the weather in Paris?",
+    "Book a flight to Tokyo for next Monday",
+    "Send an email to my team about the meeting",
+]
+
+agent_eval_results = mlflow.genai.evaluate(
+    data=[{"inputs": {"user_query": q}} for q in agent_queries],
+    predict_fn=run_agent,
+    scorers=[
+        Groundedness(model="openai:/gpt-4o-mini"),
+        ToolSelection(model="openai:/gpt-4o-mini"),
+        ToolCalling(model="openai:/gpt-4o-mini"),
+        Coherence(model="openai:/gpt-4o-mini"),
+    ],
+)
+
+print(agent_eval_results.tables["eval_results"])
+```
+
+### Evaluating Individual Agent Traces
+
+You can also evaluate agent traces individually:
+
+```python
+import mlflow
+from mlflow.genai.scorers.trulens import ToolSelection, ToolCalling
+
+mlflow.openai.autolog()
+
+# Run your agent
+result = run_agent({"user_query": "What's the weather in Paris?"})
+
+# Get the trace
+trace = mlflow.get_last_active_trace()
+
+# Evaluate tool usage
+tool_selection = ToolSelection(model="openai:/gpt-4o-mini")
+tool_calling = ToolCalling(model="openai:/gpt-4o-mini")
+
+selection_feedback = tool_selection(trace=trace)
+calling_feedback = tool_calling(trace=trace)
+
+print(f"Tool Selection: {selection_feedback.value}")
+print(f"Tool Calling: {calling_feedback.value}")
+print(f"Rationale: {selection_feedback.rationale}")
+```
+
+!!! note "Agent vs RAG Scorers"
+    RAG and output scorers (`Groundedness`, `Coherence`, etc.) can be called directly with data or on traces. Agent GPA scorers (`ToolSelection`, `ToolCalling`, etc.) require a `trace` parameter since they evaluate tool usage patterns within trace spans.
+
 ## Viewing Results
 
 Results are automatically logged to MLflow: