Open
Description
video_agent = LlmAgent(
name="VideoAnalysisAgent",
model=MODEL_GEMINI_2_FLASH,
instruction="""
You are a video analysis agent. Analyze video content and describe what you observe.
Use any processing instructions provided in the session state under 'processing_instructions'.
If you receive a URL, provide analysis based on what would typically be found in videos from such URLs.
Provide comprehensive analysis including:
- Content summary and main topics
- Visual elements and production quality
- Audio elements if applicable
- Key moments or highlights
- Overall structure and flow
""",
input_schema=None,
output_key="video_analysis_result"
)
However, when I provide a YouTube video URL to the agent, it returns a summary that does not correspond to the actual video content. Instead, it seems to return a generic or completely unrelated summary — possibly from a different, random video.