Handle FastEmbed generator results

backend/app/embeddings.py

@@ -86,8 +86,15 @@ def embed_text(text: str) -> List[float]:
         return [0.0] * get_embedding_size()
     
     model = get_embedding_model()
-    embedding = model.embed([text])
-    return embedding[0].tolist()
+    embeddings = iter(model.embed([text]))
+    try:
+        embedding = next(embeddings)
+    except StopIteration as exc:
+        raise RuntimeError("Embedding model returned no vector") from exc
+
+    if hasattr(embedding, "tolist"):
+        return embedding.tolist()
+    return list(embedding)
 
 
 def embed_texts(texts: List[str]) -> List[List[float]]:

tests/test_embeddings.py

@@ -0,0 +1,29 @@
+"""Tests for FastEmbed result normalization."""
+
+from unittest.mock import patch
+
+from backend.app.embeddings import embed_text
+
+
+class ArrayLike:
+    def tolist(self):
+        return [0.1, 0.2, 0.3]
+
+
+def test_embed_text_consumes_generator_result():
+    class Model:
+        def embed(self, texts):
+            assert texts == ["query"]
+            yield ArrayLike()
+
+    with patch("backend.app.embeddings.get_embedding_model", return_value=Model()):
+        assert embed_text("query") == [0.1, 0.2, 0.3]
+
+
+def test_embed_text_accepts_plain_vector_iterables():
+    class Model:
+        def embed(self, texts):
+            return iter([[0.4, 0.5]])
+
+    with patch("backend.app.embeddings.get_embedding_model", return_value=Model()):
+        assert embed_text("query") == [0.4, 0.5]