Pluggable CLI and web interface for AI-powered tools. Currently supports document compression with plans for more tools.

• 🔧 Pluggable Architecture - Easy to add new tools
• 🌐 Dual Interface - CLI and modern web UI
• 🔄 Multi-Provider Support - Ollama (ready), OpenAI & Anthropic (coming soon)
• 📦 File Upload - Support for text, code, PDF, and images with OCR
• 🎨 Dark Mode - Beautiful UI with light/dark themes
• 📊 Diff Viewer - Visual comparison of changes
• 💾 History & Presets - Save and reuse configurations
• ⚡ Streaming - Real-time output generation

# Clone the repository
git clone <your-repo-url>
cd ai-tools

# Install root dependencies (CLI)
npm install

# Install web server dependencies
cd src/web/server
npm install
cd ../../..

# Install web client dependencies
cd src/web/client
npm install
cd ../../..

# Install CLI globally (optional)
npm link

• Node.js 18+
• Ollama (default provider)

Install Ollama:

# Visit https://ollama.ai for installation
# Then pull recommended model
ollama pull qwen2.5-coder:14b

# Compress documentation
ai-tools compress input.md -o output.md

# With specific provider and model
ai-tools compress input.md --provider ollama --model llama3.1:8b

# From stdin
cat docs.md | ai-tools compress > compressed.md

# With streaming
ai-tools compress input.md --stream

# List available tools and providers
ai-tools list

# Manage configuration
ai-tools config show
ai-tools config init
ai-tools config set provider ollama

# Start both server and client (recommended)
npm run web:dev

# Or start separately
npm run web:server  # Backend on port 3842
npm run web:client  # Frontend on port 5847

# Then open http://localhost:5847

Web Features:

• Drag & drop file upload
• Real-time streaming output
• Side-by-side and diff views
• Save presets for quick access
• History of recent compressions
• Provider/model configuration

Create .ai-toolsrc in your home directory or project root:

{
  "provider": "ollama",
  "models": {
    "ollama": "qwen2.5-coder:14b",
    "openai": "gpt-4-turbo",
    "anthropic": "claude-sonnet-4"
  },
  "temperature": 0.2,
  "maxTokens": 8192
}

Config priority:

1. CLI flags (highest)
2. .ai-toolsrc file
3. Default to Ollama

Compress verbose technical documentation while preserving accuracy.

Features:

• Removes marketing language and fluff
• Preserves all code, commands, and technical details
• Maintains structure and formatting
• Optimizes for information density

CLI:

ai-tools compress docs.md -o compressed.md --target-length 50

API:

import { execute } from "ai-tools";

const result = await execute("compress", inputText, {
  provider: "ollama",
  model: "qwen2.5-coder:14b",
});

ai-tools/
├── src/
│   ├── core/                 # Shared AI provider logic
│   │   ├── config.js
│   │   ├── ai-provider.js
│   │   ├── provider-factory.js
│   │   └── providers/
│   │       ├── ollama-provider.js
│   │       ├── openai-provider.js  (stub)
│   │       └── anthropic-provider.js  (stub)
│   ├── tools/                # Tool definitions
│   │   ├── tool-base.js
│   │   ├── compress.js
│   │   └── index.js
│   ├── cli.js                # CLI interface
│   ├── index.js              # Programmatic API
│   └── web/
│       ├── server/           # Express backend (port 3842)
│       │   ├── routes/
│       │   ├── services/
│       │   ├── middleware/
│       │   └── server.js
│       └── client/           # React frontend (port 5847)
│           ├── src/
│           │   ├── components/
│           │   ├── pages/
│           │   ├── services/
│           │   └── context/
│           └── vite.config.js
├── package.json
└── README.md

// src/tools/your-tool.js
import ToolBase from "./tool-base.js";

class YourTool extends ToolBase {
  getName() {
    return "your-tool";
  }

  getDescription() {
    return "Description of what your tool does";
  }

  getSystemPrompt() {
    return "System prompt for the AI...";
  }

  buildUserPrompt(input, options) {
    return `Process this: ${input}`;
  }
}

export default YourTool;

// src/tools/index.js
import YourTool from "./your-tool.js";

const TOOLS = {
  compress: CompressTool,
  "your-tool": YourTool, // Add here
};

ai-tools your-tool input.md

import { execute, executeStream } from "ai-tools";

// Execute tool
const result = await execute("compress", input, {
  provider: "ollama",
  model: "qwen2.5-coder:14b",
  temperature: 0.2,
});

// Stream output
const stream = await executeStream("compress", input);
for await (const chunk of stream) {
  console.log(chunk.message.content);
}

POST   /api/tools/:toolName/execute
POST   /api/tools/:toolName/upload
GET    /api/tools/:toolName/stream    (SSE)
GET    /api/tools
GET    /api/config
PUT    /api/config
POST   /api/config/test-provider
GET    /api/providers
GET    /api/providers/:name/models
GET    /api/health

# Run CLI with auto-reload
npm run dev

# Run web with auto-reload
npm run web:dev

# Build web client for production
npm run web:build

# Preview production build
npm run web:preview

Upload Support:

• Text: .txt, .md, .json, .yaml, .xml, .csv, .log, etc.
• Code: .js, .py, .java, .cpp, .rs, .go, and 50+ more
• Documents: .pdf (text extraction)
• Images: .png, .jpg, .gif (OCR via Tesseract)

Max File Sizes:

• Text/Code: 10MB (configurable)
• PDF: 20MB (configurable)
• Images: 5MB (configurable)

• OpenAI provider implementation
• Anthropic provider implementation
• More tools (summarize, translate, refactor, etc.)
• History export/import
• Batch processing
• API authentication
• Database persistence
• Tool marketplace

MIT

Contributions welcome! The pluggable architecture makes it easy to add new tools and providers.