youtube transcript

agents/video_analyzer_agent.py

@@ -405,10 +405,14 @@ def initialize_video_analyzer_agent() -> FunctionAgent:
         agent = FunctionAgent(
             name="video_analyzer_agent",
             description=(
-                "VideoAnalyzerAgent inspects video files using Gemini's multimodal "
-                "video understanding capabilities, producing factual scene analysis, "
-                "temporal segmentation, and concise summaries as guided by the system "
-                "prompt."
+                "VideoAnalyzerAgent is a domain-specialist in multimodal video understanding, "
+                "leveraging Gemini’s vision capabilities to deliver precise, frame-level analyses. "
+                "It performs chronological segmentation of visual events, identifies key objects "
+                "and actions, and generates concise executive summaries—all based solely on visual data. "
+                "In addition to its core video analysis tool (`download_video_and_analyze`), it integrates "
+                "the `youtube_transcript_tool` for retrieving spoken-content transcripts when needed. "
+                "Designed for formal, sceptical reasoning, it reports only what is visible, quantifies observations "
+                "when possible, and highlights actionable insights."
             ),
             llm=llm,
             system_prompt=system_prompt,