fix(gemini): wrap video content in types.Part with inline_data for API compatibility

- Fix Pydantic validation error when sending video content to Gemini
- Video data now properly wrapped in types.Part with inline_data containing types.Blob
- Resolves issue #35 where video input failed with 'Extra inputs are not permitted' error
- Maintains backward compatibility with existing video parsing logic
- Updated tests to expect types.Part objects instead of raw dictionaries

Co-Authored-By: Elijah Arbee <ec.arbee1@gmail.com>

Author: devin-ai-integration[bot]

src/providers/google.py

@@ -173,7 +173,12 @@ def _parse_content(self, content: Union[str, List[Dict[str, Any]]]) -> List[Any]
                         video_bytes = base64.b64decode(video_data["data"])
                         mime_type = video_data.get("mime_type", "video/mp4")
 
-                    parts.append({"mime_type": mime_type, "data": video_bytes})
+                    parts.append(types.Part(
+                        inline_data=types.Blob(
+                            mime_type=mime_type,
+                            data=video_bytes
+                        )
+                    ))
                 elif "file_path" in video_data:
                     file_path = video_data["file_path"]
                     if os.path.exists(file_path):
@@ -182,7 +187,12 @@ def _parse_content(self, content: Union[str, List[Dict[str, Any]]]) -> List[Any]
                         mime_type, _ = mimetypes.guess_type(file_path)
                         if not mime_type or not mime_type.startswith("video/"):
                             mime_type = "video/mp4"
-                        parts.append({"mime_type": mime_type, "data": video_bytes})
+                        parts.append(types.Part(
+                            inlineData=types.Blob(
+                                mimeType=mime_type,
+                                data=video_bytes
+                            )
+                        ))
             elif t == "document":
                 doc_data = item.get("document", {})
                 if "data" in doc_data:

tests/test_content_types.py

@@ -124,27 +124,29 @@ def test_video_base64_data_with_prefix(self):
         content = [
             {"type": "video", "video": {"data": f"data:video/mp4;base64,{video_b64}"}}
         ]
-        
+
         result = self.google_completions._parse_content(content)
-        
+
         assert len(result) == 1
-        assert isinstance(result[0], dict)
-        assert result[0]["mime_type"] == "video/mp4"
-        assert result[0]["data"] == self.sample_video_bytes
+        from google.genai import types
+        assert isinstance(result[0], types.Part)
+        assert result[0].inline_data.mime_type == "video/mp4"
+        assert result[0].inline_data.data == self.sample_video_bytes
 
     def test_video_base64_data_with_explicit_mime(self):
         """Test video parsing with explicit mime_type"""
         video_b64 = base64.b64encode(self.sample_video_bytes).decode('utf-8')
         content = [
             {"type": "video", "video": {"data": video_b64, "mime_type": "video/webm"}}
         ]
-        
+
         result = self.google_completions._parse_content(content)
-        
+
         assert len(result) == 1
-        assert isinstance(result[0], dict)
-        assert result[0]["mime_type"] == "video/webm"
-        assert result[0]["data"] == self.sample_video_bytes
+        from google.genai import types
+        assert isinstance(result[0], types.Part)
+        assert result[0].inline_data.mime_type == "video/webm"
+        assert result[0].inline_data.data == self.sample_video_bytes
 
     def test_video_file_path_mime_detection(self):
         """Test video parsing from file path with MIME type detection"""
@@ -159,9 +161,10 @@ def test_video_file_path_mime_detection(self):
             result = self.google_completions._parse_content(content)
 
             assert len(result) == 1
-            assert isinstance(result[0], dict)
-            assert result[0]["mime_type"] == "video/mp4"
-            assert result[0]["data"] == self.sample_video_bytes
+            from google.genai import types
+            assert isinstance(result[0], types.Part)
+            assert result[0].inline_data.mime_type == "video/mp4"
+            assert result[0].inline_data.data == self.sample_video_bytes
 
             os.unlink(tmp_file.name)
 
@@ -178,9 +181,10 @@ def test_video_file_path_unknown_extension_fallback(self):
             result = self.google_completions._parse_content(content)
 
             assert len(result) == 1
-            assert isinstance(result[0], dict)
-            assert result[0]["mime_type"] == "video/mp4"
-            assert result[0]["data"] == self.sample_video_bytes
+            from google.genai import types
+            assert isinstance(result[0], types.Part)
+            assert result[0].inline_data.mime_type == "video/mp4"
+            assert result[0].inline_data.data == self.sample_video_bytes
 
             os.unlink(tmp_file.name)
 
@@ -272,9 +276,10 @@ def test_mixed_content_types(self):
         assert result[1]["mime_type"] == "application/pdf"
         assert result[1]["data"] == self.sample_pdf_bytes
 
-        assert isinstance(result[2], dict)
-        assert result[2]["mime_type"] == "video/mp4"
-        assert result[2]["data"] == self.sample_video_bytes
+        from google.genai import types
+        assert isinstance(result[2], types.Part)
+        assert result[2].inline_data.mime_type == "video/mp4"
+        assert result[2].inline_data.data == self.sample_video_bytes
 
         assert isinstance(result[3], dict)
         assert result[3]["mime_type"] == "text/plain"