feat(agno): instrument agno evals

examples/agno/README.md

@@ -9,6 +9,8 @@ Each script calls `braintrust.auto_instrument()` before importing `agno`, so all
 | `async_simple_agent_stream.py` | one agent, async + streamed |
 | `team_agent.py` | research + advisor team, sync |
 | `async_team_agent.py` | research + advisor team, async + streamed |
+| `accuracy_eval.py` | `AccuracyEval` over one agent, scored in Braintrust |
+| `eval_suite.py` | an `agno.eval` suite; each `Case` becomes an experiment row |
 
 ## Run
 

examples/agno/accuracy_eval.py

@@ -0,0 +1,33 @@
+import braintrust
+
+
+braintrust.auto_instrument()
+
+# An eval run logs to whatever is current. Swap init_logger for
+# braintrust.init(project=..., experiment=...) to score it as an experiment row instead.
+braintrust.init_logger(project="agno-evals-project")
+
+from agno.agent import Agent
+from agno.eval.accuracy import AccuracyEval
+from agno.models.openai import OpenAIChat
+from agno.tools.yfinance import YFinanceTools
+
+
+agent = Agent(
+    name="Stock Price Agent",
+    model=OpenAIChat(id="gpt-4o-mini"),
+    tools=[YFinanceTools()],
+    instructions="You are a stock price agent. Answer with the ticker and nothing else.",
+)
+
+evaluation = AccuracyEval(
+    name="Ticker Lookup",
+    model=OpenAIChat(id="gpt-4o-mini"),
+    agent=agent,
+    input="Which ticker does Figma trade under?",
+    expected_output="FIG",
+    num_iterations=2,
+)
+
+result = evaluation.run(print_summary=True)
+print(f"average score: {result.avg_score}/10")

examples/agno/eval_suite.py

@@ -0,0 +1,51 @@
+# agno.eval lazy-imports its submodules through a module-level __getattr__, which
+# static analysis cannot see.
+# pylint: disable=no-name-in-module
+
+import sys
+
+import braintrust
+
+
+braintrust.auto_instrument()
+
+# A suite run opens a Braintrust experiment of its own, so each Case lands as a
+# scored experiment row. Pass eval_experiments=False to setup_agno() (or set
+# BRAINTRUST_AGNO_EVAL_EXPERIMENTS=false) to keep suite runs in logs instead.
+braintrust.init_logger(project="agno-evals-project")
+
+from agno.agent import Agent
+from agno.eval import Case, cli
+from agno.models.openai import OpenAIChat
+from agno.tools.yfinance import YFinanceTools
+
+
+agent = Agent(
+    id="stock-agent",
+    name="Stock Price Agent",
+    model=OpenAIChat(id="gpt-4o-mini"),
+    tools=[YFinanceTools()],
+    instructions="Use your tools for any market data question.",
+)
+
+CASES = (
+    Case(
+        name="looks_up_current_price",
+        agent=agent,
+        input="What is the current price of FIG?",
+        tags=("smoke",),
+        criteria="Reports a current share price for Figma.",
+        expected_tool_calls=("get_current_stock_price",),
+    ),
+    Case(
+        name="explains_pe_ratio",
+        agent=agent,
+        input="Explain the P/E ratio in one sentence.",
+        criteria="Explains that the P/E ratio compares share price to earnings per share.",
+    ),
+)
+
+if __name__ == "__main__":
+    # python eval_suite.py --tag smoke     # run a tagged subset
+    # python eval_suite.py --list          # list cases without running them
+    sys.exit(cli(CASES))

py/src/braintrust/integrations/agno/__init__.py

@@ -2,13 +2,19 @@
 
 import logging
 
-from braintrust.logger import NOOP_SPAN, current_span, init_logger
+from braintrust.logger import NOOP_SPAN, current_experiment, current_span, init_logger
 
+from .eval_experiments import configure as _configure_eval_experiments
 from .integration import AgnoIntegration
 from .patchers import (
+    wrap_accuracy_eval,
     wrap_agent,
+    wrap_agent_as_judge_eval,
+    wrap_eval_suite,
     wrap_function_call,
     wrap_model,
+    wrap_performance_eval,
+    wrap_reliability_eval,
     wrap_team,
     wrap_workflow,
 )
@@ -19,9 +25,14 @@
 __all__ = [
     "AgnoIntegration",
     "setup_agno",
+    "wrap_accuracy_eval",
     "wrap_agent",
+    "wrap_agent_as_judge_eval",
+    "wrap_eval_suite",
     "wrap_function_call",
     "wrap_model",
+    "wrap_performance_eval",
+    "wrap_reliability_eval",
     "wrap_team",
     "wrap_workflow",
 ]
@@ -31,20 +42,30 @@ def setup_agno(
     api_key: str | None = None,
     project_id: str | None = None,
     project_name: str | None = None,
+    eval_experiments: bool | None = None,
 ) -> bool:
     """
-    Setup Braintrust integration with Agno. Will automatically patch Agno agents, models, and function calls for tracing.
+    Setup Braintrust integration with Agno. Will automatically patch Agno agents, models,
+    function calls, and evals (``agno.eval``) for tracing.
 
     Args:
         api_key: Braintrust API key (optional, can use env var BRAINTRUST_API_KEY)
         project_id: Braintrust project ID (optional)
-        project_name: Braintrust project name (optional, can use env var BRAINTRUST_PROJECT)
+        project_name: Braintrust project name (optional; defaults to the Global project)
+        eval_experiments: Whether an eval suite run should open a Braintrust experiment,
+            so its cases land as experiment rows rather than logs. Defaults to the
+            BRAINTRUST_AGNO_EVAL_EXPERIMENTS env var, which itself defaults to true.
+            Individual evals (AccuracyEval and friends) always log to whatever is
+            current, so pass eval_experiments=False to keep suite runs in logs too.
 
     Returns:
         True if setup was successful, False otherwise
     """
-    span = current_span()
-    if span == NOOP_SPAN:
+    _configure_eval_experiments(eval_experiments)
+
+    # An experiment opened by the caller is the destination for eval rows, so don't
+    # install a logger that would only shadow it for non-eval tracing.
+    if current_span() == NOOP_SPAN and current_experiment() is None:
         init_logger(project=project_name, api_key=api_key, project_id=project_id)
 
     return AgnoIntegration.setup()