AKJUS · pull · Jun 3, 2025 · Jun 3, 2025
diff --git a/sdks/opik_optimizer/scripts/dspy_mipro_hotpot.py b/sdks/opik_optimizer/scripts/dspy_mipro_hotpot.py
@@ -0,0 +1,126 @@
+"""
+In this example, we show how to use DSPy's MIPROv2 with:
+
+* Opik's metrics
+* Opik's datasets
+* Opik's optimization infrastructure
+"""
+
+import random
+import os
+
+import opik
+from opik.evaluation.metrics import Equals
+from opik_optimizer.datasets import hotpot_300
+from opik.integrations.dspy.callback import OpikCallback
+from opik_optimizer import (
+    MetricConfig,
+    TaskConfig,
+    from_dataset_field,
+    from_llm_response_text,
+)
+from opik_optimizer.mipro_optimizer.utils import (
+    opik_metric_to_dspy,
+    create_dspy_training_set,
+)
+import dspy
+
+# Setup cache on disk:
+import litellm
+from litellm.caching import Cache
+
+# Using disk cache for LLM calls
+disk_cache_dir = os.path.expanduser("~/.litellm_cache")
+litellm.cache = Cache(type="disk", disk_cache_dir=disk_cache_dir)
+
+# To log all data, use this:
+from opik_optimizer.mipro_optimizer import MIPROv2
+
+# Else:
+# from dspy.teleprompt import MIPROv2
+
+# First, we set the Opik callback for all dspy calls:
+project_name = "DSPy-MIPROv2"
+opik_callback = OpikCallback(project_name=project_name, log_graph=True)
+dspy.settings.configure(
+    lm=dspy.LM(model="openai/gpt-4o-mini"),
+    callbacks=[opik_callback],
+)
+
+
+# Define our tools:
+def search_wikipedia(query: str) -> list[str]:
+    """
+    This agent is used to search wikipedia. It can retrieve additional details
+    about a topic.
+    """
+    results = dspy.ColBERTv2(url="http://20.102.90.50:2017/wiki17_abstracts")(
+        query, k=3
+    )
+    return [x["text"] for x in results]
+
+
+# This are useful methods of logging optimization data:
+metric_config = MetricConfig(
+    metric=Equals(project_name=project_name),
+    inputs={
+        "output": from_llm_response_text(),
+        "reference": from_dataset_field(name="answer"),
+    },
+)
+
+task_config = TaskConfig(
+    instruction_prompt="Answer the question",
+    input_dataset_fields=["question"],
+    output_dataset_field="answer",
+    tools=[search_wikipedia],
+)
+
+dataset = hotpot_300()
+trainset = create_dspy_training_set(dataset.get_items(), "question", 50)
+metric_function = opik_metric_to_dspy(metric_config.metric, "answer")
+
+experiment_config = {
+    "optimizer": "DSPy-MIPROv2",
+    "tools": [search_wikipedia.__name__],
+    "metric": metric_config.metric.name,
+    "num_threads": 10,
+    "num_candidates": 5,
+    "num_trials": 5,
+    "dataset": dataset.name,
+}
+
+opik_client = opik.Opik()
+optimization = opik_client.create_optimization(
+    dataset_name=dataset.name,
+    objective_name=metric_config.metric.name,
+    metadata={"optimizer": "DSPy-MIPROv2"},
+)
+
+optimizer = MIPROv2(
+    metric=metric_function,
+    auto=None,
+    num_candidates=experiment_config["num_candidates"],
+    num_threads=10,
+    verbose=False,
+    seed=42,
+    # Add these if using Opik's MIPROv2:
+    opik_dataset=dataset,
+    opik_metric_config=metric_config,
+    opik_prompt_task_config=task_config,
+    opik_project_name=project_name,
+    opik_optimization_id=optimization.id,
+    experiment_config=experiment_config,
+)
+
+program = dspy.ReAct("question -> answer", [search_wikipedia])
+
+results = optimizer.compile(
+    student=program,
+    trainset=trainset,
+    provide_traceback=True,
+    requires_permission_to_run=False,
+    num_trials=experiment_config["num_trials"],
+)
+
+optimization.update(status="completed")
diff --git a/sdks/opik_optimizer/src/opik_optimizer/mipro_optimizer/__init__.py b/sdks/opik_optimizer/src/opik_optimizer/mipro_optimizer/__init__.py
@@ -1 +1 @@
-from .mipro_optimizer import MiproOptimizer
+from .mipro_optimizer import MiproOptimizer, MIPROv2
Original file line number	Diff line number	Diff line change
		@@ -1 +1 @@
		from .mipro_optimizer import MiproOptimizer
		from .mipro_optimizer import MiproOptimizer, MIPROv2