From d5a2a4b4fbf762a42471459eaa29ba0a5bd1f586 Mon Sep 17 00:00:00 2001
From: Anthony Naddeo <anthony@whylabs.ai>
Date: Wed, 10 Apr 2024 10:55:46 -0700
Subject: [PATCH 1/2] Add response.similarity.context

We had this for the prompt before and now we have it for the response.
---
 langkit/metrics/library.py                    |  7 ++++++
 .../metrics/test_input_context_similarity.py  | 22 +++++++++++++++++++
 2 files changed, 29 insertions(+)

diff --git a/langkit/metrics/library.py b/langkit/metrics/library.py
index c7f1a5e..77a7bc4 100644
--- a/langkit/metrics/library.py
+++ b/langkit/metrics/library.py
@@ -39,6 +39,7 @@ def all(prompt: bool = True, response: bool = True) -> MetricCreator:
                 response_refusal_similarity_metric,
                 response_presidio_pii_metric,
                 lib.response.toxicity(),
+                lib.response.similarity.context(),
                 lib.response.topics.medicine(),
             ]
 
@@ -512,6 +513,12 @@ def refusal(onnx: bool = True) -> MetricCreator:
 
                 return partial(response_refusal_similarity_metric, onnx=onnx)
 
+            @staticmethod
+            def context(onnx: bool = True) -> MetricCreator:
+                from langkit.metrics.input_context_similarity import input_context_similarity
+
+                return partial(input_context_similarity, onnx=onnx, input_column_name="response")
+
         class topics:
             def __init__(self, topics: List[str], hypothesis_template: Optional[str] = None, onnx: bool = True):
                 self.topics = topics
diff --git a/tests/langkit/metrics/test_input_context_similarity.py b/tests/langkit/metrics/test_input_context_similarity.py
index 1c26d00..ce97d13 100644
--- a/tests/langkit/metrics/test_input_context_similarity.py
+++ b/tests/langkit/metrics/test_input_context_similarity.py
@@ -29,6 +29,28 @@ def test_similarity():
     assert metrics["prompt.similarity.context"][0] == pytest.approx(0.7447172999382019)  # pyright: ignore[reportUnknownMemberType]
 
 
+def test_similarity_repoonse():
+    wf = Workflow(metrics=[lib.response.similarity.context()])
+
+    context: InputContext = {
+        "entries": [
+            {"content": "Some source 1", "metadata": {"source": "https://internal.com/foo"}},
+            {"content": "Some source 2", "metadata": {"source": "https://internal.com/bar"}},
+        ]
+    }
+
+    df = pd.DataFrame({"response": ["Some source"], "context": [context]})
+
+    result = wf.run(df)
+
+    metrics = result.metrics
+
+    metric_names: List[str] = metrics.columns.tolist()  # pyright: ignore[reportUnknownMemberType]
+
+    assert metric_names == ["response.similarity.context", "id"]
+    assert metrics["response.similarity.context"][0] == pytest.approx(0.7447172999382019)  # pyright: ignore[reportUnknownMemberType]
+
+
 def test_similarity_missing_context():
     # The metric should not be run in this case since the context is missing
     wf = Workflow(metrics=[lib.prompt.similarity.context()])

From 94b8fb78c90c28da708b2895d4b8b285ce5f2aab Mon Sep 17 00:00:00 2001
From: Anthony Naddeo <anthony@whylabs.ai>
Date: Wed, 10 Apr 2024 10:56:34 -0700
Subject: [PATCH 2/2] bump version

---
 .bumpversion.cfg | 2 +-
 pyproject.toml   | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/.bumpversion.cfg b/.bumpversion.cfg
index b5d79d7..e9def7f 100644
--- a/.bumpversion.cfg
+++ b/.bumpversion.cfg
@@ -1,5 +1,5 @@
 [bumpversion]
-current_version = 0.0.28.dev8
+current_version = 0.0.28.dev9
 tag = False
 parse = (?P<major>\d+)\.(?P<minor>\d+)\.(?P<patch>\d+)(\.(?P<release>[a-z]+)(?P<build>\d+))?
 serialize = 
diff --git a/pyproject.toml b/pyproject.toml
index a2673e2..a9c934d 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "langkit"
-version = "0.0.28.dev8"
+version = "0.0.28.dev9"
 description = "A language toolkit for monitoring LLM interactions"
 authors = ["WhyLabs.ai <langkit@whylabs.ai>"]
 homepage = "https://docs.whylabs.ai/docs/large-language-model-monitoring"