Shipressure ⚡

Universal AI Context Compression

Compress documents, repositories, documentation sites, and spreadsheets into highly compressed AI-ready context. One command. Zero cloud dependency. Optimized for AI coding agents.

npx shipressure ./api.pdf
npx shipressure ./docs/
npx shipressure https://github.com/user/repo

Vision

Instead of manually uploading files into Claude, Codex, Gemini, or OpenCode — provide a file path, directory, URL, or repository. Shipressure automatically reads content, extracts important information, removes noise, compresses context, and produces token-efficient plain text optimized for AI agents.

Installation

npm install -g shipressure-cli

Or run directly:

npx shipressure-cli@latest install

Usage

Files

shipressure ./api.pdf
shipressure ./document.docx
shipressure ./data.csv
shipressure ./presentation.pptx

Directories

shipressure ./docs/
shipressure ./src/

Documentation Websites

shipressure https://docs.api.nvidia.com
shipressure https://docs.example.com

GitHub Repositories

shipressure https://github.com/vercel/next.js
shipressure https://github.com/user/repo

Modes

| Mode | Flag | Use Case | |------|------|----------| | Balanced | (default) | General purpose | | Full | --full | Preserve everything | | Ultra | --ultra | Maximum compression |

Examples

shipressure file.pdf              # Balanced (default)
shipressure file.pdf --full       # Preserve everything
shipressure file.pdf --ultra      # Maximum compression

Supported Formats

• Documents: PDF, DOCX, TXT, MD, RTF
• Spreadsheets: XLSX, XLS, CSV, ODS
• Presentations: PPTX, PPT
• Codebases: JS, TS, Python, Rust, Solidity, Go (via GitHub)
• Documentation: GitBook, Mintlify, Docusaurus, Readme, OpenAPI (via URL)

How It Works

User Input → Parser Engine → Normalizer → Compressor → Context Builder → AI Output

1. Input Resolver — Detects input type (file, directory, URL, GitHub)
2. Parser Engine — Extracts text from PDF, DOCX, XLSX, CSV, MD, PPTX
3. Documentation Scraper — Crawls docs sites (GitBook, Mintlify, Docusaurus, Readme)
4. GitHub Analyzer — Analyzes repos (README, structure, config, source files)
5. Normalizer — Converts all content to standard format
6. Compression Engine — Removes noise, deduplicates, condenses text
7. Context Builder — Outputs compressed AI-optimized context

Output Format

[SHIPRESSURE]

PROJECT: NVIDIA NIM

AUTH
OAuth2 → token exchange
API Keys supported

ENDPOINTS
POST /chat/completions
GET /models

MODELS
Llama 3.3
DeepSeek R1

[/SHIPRESSURE]

// ⚡ ~350 tokens | 82% compression

Token Reduction

| Mode | Typical Reduction | |------|------------------| | Full | 0% (preserve) | | Balanced | 40-60% | | Ultra | 60-95% |

AI Agent Integration

Claude Code

/shipressure ./api.pdf

Codex

use shipressure ./api.pdf

OpenCode

use shipressure ./api.pdf

Cursor

use shipressure ./api.pdf

API Usage

import { shipressure } from 'shipressure-cli';

const result = await shipressure('./api.pdf', { mode: 'balanced' });
console.log(result.content);
console.log(`Compressed: ${result.context.tokenEstimate} tokens`);
console.log(`Reduction: ${result.context.compressionRatio}%`);

Project Structure

shipressure/
├── src/
│   ├── cli/              # CLI entry point
│   ├── parsers/          # File parsers (PDF, DOCX, XLSX, CSV, MD, PPTX)
│   ├── scrapers/         # Documentation & GitHub scrapers
│   ├── normalizers/      # Content normalization
│   ├── compressors/      # Token compression engine
│   ├── builders/         # Context builder
│   ├── resolvers/        # Input resolver
│   └── types/            # TypeScript types
├── skill.md              # AI agent skill manifest
├── package.json
└── README.md

Future Versions

• v1 — Document compression
• v2 — Repository intelligence
• v3 — RAG generation
• v4 — Vector embeddings
• v5 — Knowledge graph generation

License

MIT