fix[experimental]: Fix text splitter with gradient (langchain-ai#26629)

Fixes langchain-ai#26221

---------

Co-authored-by: Erick Friis <erick@langchain.dev>

libs/experimental/langchain_experimental/text_splitter.py

@@ -217,6 +217,12 @@ def split_text(
         # np.percentile to fail.
         if len(single_sentences_list) == 1:
             return single_sentences_list
+        # similarly, the following np.gradient would fail
+        if (
+            self.breakpoint_threshold_type == "gradient"
+            and len(single_sentences_list) == 2
+        ):
+            return single_sentences_list
         distances, sentences = self._calculate_sentence_distances(single_sentences_list)
         if self.number_of_chunks is not None:
             breakpoint_distance_threshold = self._threshold_from_clusters(distances)

libs/experimental/tests/unit_tests/test_text_splitter.py

@@ -0,0 +1,54 @@
+import re
+from typing import List
+
+import pytest
+from langchain_core.embeddings import Embeddings
+
+from langchain_experimental.text_splitter import SemanticChunker
+
+FAKE_EMBEDDINGS = [
+    [0.02905, 0.42969, 0.65394, 0.62200],
+    [0.00515, 0.47214, 0.45327, 0.75605],
+    [0.57401, 0.30344, 0.41702, 0.63603],
+    [0.60308, 0.18708, 0.68871, 0.35634],
+    [0.52510, 0.56163, 0.34100, 0.54089],
+    [0.73275, 0.22089, 0.42652, 0.48204],
+    [0.47466, 0.26161, 0.79687, 0.26694],
+]
+SAMPLE_TEXT = (
+    "We need to harvest synergy effects viral engagement, but digitalize, "
+    "nor overcome key issues to meet key milestones. So digital literacy "
+    "where the metal hits the meat. So this vendor is incompetent. Can "
+    "you champion this? Let me diarize this. And we can synchronise "
+    "ourselves at a later timepoint t-shaped individual tread it daily. "
+    "That is a good problem"
+)
+
+
+class MockEmbeddings(Embeddings):
+    def embed_documents(self, texts: List[str]) -> List[List[float]]:
+        return FAKE_EMBEDDINGS[: len(texts)]
+
+    def embed_query(self, text: str) -> List[float]:
+        return [1.0, 2.0]
+
+
+@pytest.mark.parametrize(
+    "input_length, expected_length",
+    [
+        (1, 1),
+        (2, 2),
+        (5, 2),
+    ],
+)
+def test_split_text_gradient(input_length: int, expected_length: int) -> None:
+    embeddings = MockEmbeddings()
+    chunker = SemanticChunker(
+        embeddings,
+        breakpoint_threshold_type="gradient",
+    )
+    list_of_sentences = re.split(r"(?<=[.?!])\s+", SAMPLE_TEXT)[:input_length]
+
+    chunks = chunker.split_text(" ".join(list_of_sentences))
+
+    assert len(chunks) == expected_length