viseeker 0.1.0

A collection of computer-vision tools for LLM Agents (image/video processing & understanding)

viseeker

A collection of computer vision utilities for image and video processing, designed to be easy for AI agents to use.

Quick Start

Install Dependencies

uv sync --dev

Some tools require FFmpeg:

brew install ffmpeg

Run a Tool

uv run python -m viseeker.video_metadata ./demo.mp4

Available Tools

See TOOLS.md for detailed usage and examples.

• video_metadata: Extract video metadata (duration, codec, resolution, framerate, etc.)
• video_keyframes: Extract keyframes using multiple algorithms
• video_resize: Resize videos with various aspect ratio strategies
• video_remove_audio: Remove audio track from videos
• video_convert_mp4: Convert videos to MP4 format
• video_adaptive_compress: Adaptive video compression to target file size
• video_split: Split videos by I-frames or fixed duration

Features

• Unified Input: Supports local paths, HTTP/HTTPS URLs, and S3 URLs (s3://bucket/key)
• JSON Output: All tools return JSON-serializable results
• Composable: Single-responsibility tools that can be combined in workflows