Convert text items from 'text' to 'input_text' for Responses API

Co-authored-by: TomeHirata <[email protected]>

dspy/clients/lm.py

@@ -491,16 +491,26 @@ def _convert_content_item_to_responses_format(item: dict[str, Any]) -> dict[str,
     To:
         {"type": "input_image", "image_url": "..."}
 
-    For text and other types, passes through as-is (already in correct format).
+    For text, converts from:
+        {"type": "text", "text": "..."}
+    To:
+        {"type": "input_text", "text": "..."}
+
+    For other types, passes through as-is.
     """
     if item.get("type") == "image_url":
         image_url = item.get("image_url", {}).get("url", "")
         return {
             "type": "input_image",
             "image_url": image_url,
         }
+    elif item.get("type") == "text":
+        return {
+            "type": "input_text",
+            "text": item.get("text", ""),
+        }
 
-    # For non-image items, return as-is
+    # For other items, return as-is
     return item
 
 

tests/clients/test_lm.py

@@ -666,8 +666,8 @@ def test_responses_api_converts_images_correctly():
     content = result["input"][0]["content"]
     assert len(content) == 2
 
-    # First item should be text (passed through as-is since it's already in correct format)
-    assert content[0]["type"] == "text"
+    # First item should be text converted to input_text format
+    assert content[0]["type"] == "input_text"
     assert content[0]["text"] == "What's in this image?"
 
     # Second item should be converted to input_image format