Merge 0f61117e4b into 88cb83484b

litellm/responses/litellm_completion_transformation/transformation.py

@@ -1476,9 +1476,18 @@ class LiteLLMCompletionResponsesConfig:
                                     "image_url": {"url": image_url_val},
                                 }
                             )
+                    elif part_type in ("input_file", "file"):
+                        file_item = LiteLLMCompletionResponsesConfig._transform_input_file_item_to_file_item(
+                            part
+                        )
+                        if file_item.get("file"):
+                            normalized_blocks.append(file_item)
 
-                # Prefer structured blocks if we have images; otherwise return a string.
-                if any(b.get("type") == "image_url" for b in normalized_blocks):
+                # Prefer structured blocks if we have any non-text part
+                # (images or files); otherwise return a string.
+                if any(
+                    b.get("type") in ("image_url", "file") for b in normalized_blocks
+                ):
                     # Ensure we include any accumulated text as text blocks too
                     return normalized_blocks
                 if text_acc:

tests/test_litellm/responses/litellm_completion_transformation/test_function_call_output_normalization.py

@@ -40,3 +40,91 @@ def test_function_call_output_string_passthrough():
     )
     assert len(out) == 1
     assert out[0]["content"] == '{"ok":true}'
+
+
+def test_function_call_output_input_file_is_preserved_as_file_block():
+    """A PDF returned from a tool call (input_file part inside
+    function_call_output.output) must survive the lowering to a chat-completions
+    tool message as a structured file content block. Previously the file part
+    was silently dropped and only the input_text sibling reached the model."""
+    pdf_data_url = "data:application/pdf;base64,JVBERi0xLjQKJfb=="
+
+    out = LiteLLMCompletionResponsesConfig._transform_responses_api_tool_call_output_to_chat_completion_message(
+        tool_call_output={
+            "type": "function_call_output",
+            "call_id": "call_pdf",
+            "output": [
+                {"type": "input_text", "text": "Here is the PDF."},
+                {
+                    "type": "input_file",
+                    "file_data": pdf_data_url,
+                    "filename": "test.pdf",
+                },
+            ],
+        }
+    )
+
+    assert len(out) == 1
+    msg = out[0]
+    assert msg["role"] == "tool"
+    assert msg["tool_call_id"] == "call_pdf"
+
+    content = msg["content"]
+    assert isinstance(content, list), (
+        "expected structured content list when a file part is present, "
+        f"got {type(content).__name__}: {content!r}"
+    )
+
+    text_blocks = [b for b in content if b.get("type") == "text"]
+    file_blocks = [b for b in content if b.get("type") == "file"]
+
+    assert any(b.get("text") == "Here is the PDF." for b in text_blocks)
+    assert len(file_blocks) == 1
+    assert file_blocks[0]["file"]["file_data"] == pdf_data_url
+
+
+def test_function_call_output_input_file_with_file_id_is_preserved():
+    """Same as above but for the file_id form (no inline bytes)."""
+    out = LiteLLMCompletionResponsesConfig._transform_responses_api_tool_call_output_to_chat_completion_message(
+        tool_call_output={
+            "type": "function_call_output",
+            "call_id": "call_pdf_id",
+            "output": [
+                {"type": "input_text", "text": "See attached."},
+                {
+                    "type": "input_file",
+                    "file_id": "file-abc123",
+                    "filename": "report.pdf",
+                },
+            ],
+        }
+    )
+
+    assert len(out) == 1
+    content = out[0]["content"]
+    assert isinstance(content, list)
+
+    file_blocks = [b for b in content if b.get("type") == "file"]
+    assert len(file_blocks) == 1
+    assert file_blocks[0]["file"]["file_id"] == "file-abc123"
+
+
+def test_function_call_output_input_file_without_data_is_skipped():
+    """An input_file part with no file_id, file_url, or file_data carries no
+    payload — appending it as ``{"type": "file", "file": {}}`` would activate
+    the structured-content path and emit a hollow block that downstream
+    adapters (Gemini, Bedrock) reject or silently ignore. Skip it instead."""
+    out = LiteLLMCompletionResponsesConfig._transform_responses_api_tool_call_output_to_chat_completion_message(
+        tool_call_output={
+            "type": "function_call_output",
+            "call_id": "call_empty_file",
+            "output": [
+                {"type": "input_text", "text": "Just text here."},
+                {"type": "input_file", "filename": "missing.pdf"},
+            ],
+        }
+    )
+
+    assert len(out) == 1
+    # No usable file payload -> fall back to the plain-string text path.
+    assert out[0]["content"] == "Just text here."