Implement youtube tool

multimodality_tools.py

@@ -142,13 +142,35 @@ def get_csv_analysis_tool():
         description="Analyse a csv file that is identified by its file id and get common statistics such as mean or max per column."
     )
 
-def watch_video(video_url: str) -> str:
+def watch_video(question: str, youtube_url: str) -> str:
+    """Answer a question about a YouTube video identified by its url."""
+    # TODO our Gemini Key cannot be used to watch videos.
     return "You are not able to watch a Video yet. Reply with 'I don't know' to the question."
+    try:
+        from google import genai
+        from google.genai import types
+
+        client = genai.Client(api_key="GOOGLE_API_KEY")
+        response = client.models.generate_content(
+            model='models/gemini-2.0-flash',
+            contents=types.Content(
+                parts=[
+                    types.Part(
+                        file_data=types.FileData(file_uri=youtube_url)
+                    ),
+                    types.Part(text=question)
+                ]
+            )
+        )
+    except Exception as e:
+        print(e)
+        return "You are not able to watch a Video yet. Reply with 'I don't know' to the question."
+    return str(response.text)
 
 def get_video_tool():
     return FunctionTool.from_defaults(
         fn=watch_video,
-        description="Watch a video and get a content description as a string."
+        description="Answer a question about a YouTube video identified by its url."
     )
 
 def _build_file_url(task_id: str) -> str:

requirements.txt

@@ -16,4 +16,5 @@ pandas
 openpyxl
 huggingface_hub
 transformers
-tavily-python
+tavily-python
+google-genai