Adds Video Surfer to autogen_ext

[gagb]

This pull request introduces the new VideoSurferAgent and its associated functionalities to the autogen-ext package. The changes include adding dependencies for video processing, defining the VideoSurferAgent class, and implementing various video-related functions. Additionally, there are some modifications to existing agents and their initializations.

Key Changes:

New Features:

• VideoSurferAgent Introduction:
  • Added VideoSurferAgent class in python/packages/autogen-ext/src/autogen_ext/agents/video_surfer/_video_surfer.py. This agent can answer questions about local video files using various tools for extracting audio, transcribing audio with timestamps, getting video length, saving screenshots, and using OpenAI to transcribe video screenshots.
  • Implemented video processing functions in python/packages/autogen-ext/src/autogen_ext/agents/video_surfer/_action_space.py, including extract_audio, transcribe_audio_with_timestamps, get_video_length, save_screenshot, openai_transcribe_video_screenshot, and get_screenshot_at.

Dependency Updates:

• New Dependencies for Video Processing:
  • Added new dependencies autogen-agentchat, openai, opencv-python, ffmpeg-python, and openai-whisper to pyproject.toml under video-surfer.

Codebase Modifications:

• Agent Initialization Adjustments:
  • Removed MultimodalWebSurfer import and initialization from python/packages/autogen-ext/src/autogen_ext/agents/__init__.py.
  • Re-added MultimodalWebSurfer initialization in python/packages/autogen-ext/src/autogen_ext/agents/web_surfer/__init__.py.

New Module:

• Video Surfer Module Initialization:
  • Created __init__.py for video_surfer module to include VideoSurferAgent and associated action space functions.