System Requirements

CiteKit is a hybrid cloud/local system. The "heavy lifting" of understanding content happens in the cloud via the configured mapper (Gemini by default), but the "heavy lifting" of extracting content happens locally on your machine.

Supported Operating Systems

CiteKit is tested and supported on:

• macOS: 12+ (Apple Silicon & Intel)
• Windows: 10, 11 (PowerShell or WSL2 recommended)
• Linux: Ubuntu 20.04+, Debian 11+, Alpine 3.14+

Language Runtimes

Python SDK

• Python 3.10+ (Required for modern type hinting features)

JavaScript SDK

• Node.js 18+ (Required for native fetch and crypto APIs)

External Dependencies

FFmpeg (Optional)

Why is it needed? CiteKit uses FFmpeg to slice video and audio files locally (Physical Resolution). Without it, you cannot resolve video/audio into new clip files.

When is it NOT needed?

• Virtual Resolution: If you only need timestamps/metadata.
• Modality: If you only use PDFs, Images, or Text/Code.
• Serverless: Use Virtual Mode for zero-binary deployments.

Requirement:

• FFmpeg 4.0+
• Must be available in the system PATH.

How to verify:

ffmpeg -version

PDF Tools

• Python: pymupdf is installed automatically.
• Node.js: pdf-lib is an optional peer dependency. Install it when resolving PDFs (npm install pdf-lib).

Image & Video Tools (Node.js)

• Image: sharp is an optional peer dependency (npm install sharp).
• Video/Audio: fluent-ffmpeg is an optional peer dependency (npm install fluent-ffmpeg) and requires FFmpeg installed system-wide.

Storage

CiteKit creates two local directories by default:

1. .resource_maps/: Stores JSON maps. Tiny (few KB per file).
2. .citekit_output/: Stores extracted clips/PDFs.
   • Best Practice: Configure your agent to delete these files after use if disk space is a concern.