fix(oci): fix embed embedding_types casing and handle embeddingsByType response

- embedding_types: OCI expects lowercase (float, int8) not uppercase.
  The .upper() was breaking all embedding_types requests.
- Response: OCI returns "embeddingsByType" (not "embeddings") when
  embeddingTypes is specified. Handle both response keys.
- Unit test updated to expect lowercase.
- Integration tests added: embedding_types=["float"] and truncate modes.

Author: fede-kamel

src/cohere/oci_client.py

@@ -669,7 +669,8 @@ def transform_request_to_oci(
             oci_body["truncate"] = cohere_body["truncate"].upper()
 
         if "embedding_types" in cohere_body:
-            oci_body["embeddingTypes"] = [et.upper() for et in cohere_body["embedding_types"]]
+            # OCI expects lowercase embedding types (float, int8, binary, etc.)
+            oci_body["embeddingTypes"] = [et.lower() for et in cohere_body["embedding_types"]]
         if "max_tokens" in cohere_body:
             oci_body["maxTokens"] = cohere_body["max_tokens"]
         if "output_dimension" in cohere_body:
@@ -875,7 +876,8 @@ def transform_oci_response_to_cohere(
         Transformed response in Cohere format
     """
     if endpoint == "embed":
-        embeddings_data = oci_response.get("embeddings", {})
+        # OCI returns "embeddings" by default, or "embeddingsByType" when embeddingTypes is specified
+        embeddings_data = oci_response.get("embeddingsByType") or oci_response.get("embeddings", {})
 
         if isinstance(embeddings_data, dict):
             normalized_embeddings = {str(key).lower(): value for key, value in embeddings_data.items()}

tests/test_oci_client.py

@@ -509,6 +509,30 @@ def test_embed_search_query_input_type(self):
         self.assertIsNotNone(response.embeddings.float_)
         self.assertEqual(len(response.embeddings.float_[0]), 1024)
 
+    def test_embed_with_embedding_types(self):
+        """Test embed with explicit embedding_types parameter."""
+        response = self.client.embed(
+            model="embed-english-v3.0",
+            texts=["Hello world"],
+            input_type="search_document",
+            embedding_types=["float"],
+        )
+        self.assertIsNotNone(response.embeddings.float_)
+        self.assertEqual(len(response.embeddings.float_[0]), 1024)
+
+    def test_embed_with_truncate(self):
+        """Test embed with truncate parameter."""
+        long_text = "hello " * 1000
+        for mode in ["NONE", "START", "END"]:
+            response = self.client.embed(
+                model="embed-english-v3.0",
+                texts=[long_text],
+                input_type="search_document",
+                truncate=mode,
+            )
+            self.assertIsNotNone(response.embeddings.float_)
+            self.assertEqual(len(response.embeddings.float_[0]), 1024)
+
     def test_command_r_plus_chat(self):
         """Test command-r-plus-08-2024 via V1 client."""
         v1_client = cohere.OciClient(
@@ -772,7 +796,7 @@ def test_transform_embed_request(self):
         self.assertEqual(result["inputs"], ["hello", "world"])
         self.assertEqual(result["inputType"], "SEARCH_DOCUMENT")
         self.assertEqual(result["truncate"], "END")
-        self.assertEqual(result["embeddingTypes"], ["FLOAT", "INT8"])
+        self.assertEqual(result["embeddingTypes"], ["float", "int8"])
         self.assertEqual(result["compartmentId"], "compartment-123")
         self.assertEqual(result["servingMode"]["modelId"], "cohere.embed-english-v3.0")