# Flash ⚡

Fast AI-powered code completion and chat using Cerebras API - No API key setup required!

## Installation

```bash
npm install flash-tool-oss
```

## Quick Start

### Command Line Usage

```bash
# Interactive mode
npx flash-tool-oss interactive

# Complete code
npx flash-tool-oss complete "def fibonacci"

# Chat with AI
npx flash-tool-oss chat "How do I create a React component?"

# Complete code in specific language
npx flash-tool-oss complete "function sortArray" --language javascript
```

### Programmatic Usage

```javascript
const Flash = require('flash-tool-oss');
const flash = new Flash();

// Complete code
const completion = await flash.complete('def fibonacci', 'python');
console.log(completion);

// Chat with AI
const response = await flash.chat('How do I create a React component?');
console.log(response);

// Process "using oss" commands (for Cursor integration)
const result = await flash.processUsingOss('using oss explain async/await');
console.log(result);
```

## Features

- ⚡ **Fast AI completions** using Cerebras gpt-oss-120b model
- 💬 **Interactive chat** with AI programming assistant
- 🔧 **Multi-language support** (Python, JavaScript, TypeScript, etc.)
- 🚀 **Zero configuration** - works immediately after installation
- 📦 **Easy installation** via npm
- 🎯 **Cursor integration** with "using oss" commands via MCP
- 🤖 **Automatic code application** - Cursor applies suggestions directly

## Configuration

The tool uses these default settings:
- Model: `gpt-oss-120b`
- Max tokens: 1024 (completion), 2048 (chat)
- Temperature: 0.3 (completion), 0.7 (chat)
- Top-p: 0.9
- Reasoning effort: medium

## Cursor Integration

### MCP Server Setup

To integrate with Cursor chat using MCP (Model Context Protocol):

1. **Install the package:**
   ```bash
   npm install flash-tool-oss
   ```

2. **Create MCP configuration** (see `CURSOR_SETUP.md` for detailed instructions):
   ```json
   {
     "mcpServers": {
       "flash-tool-oss": {
         "command": "node",
         "args": ["node_modules/flash-tool-oss/src/mcp-server.js"],
         "env": {}
       }
     }
   }
   ```

3. **Use in Cursor chat:**
   ```
   using oss create a React component that displays a todo list
   ```

### Programmatic Usage

You can also use the `processUsingOss` method programmatically:

```javascript
const Flash = require('flash-tool-oss');
const flash = new Flash();

// This will detect "using oss" commands and respond appropriately
const response = await flash.processUsingOss('using oss how do I debug this React error?');
```

## Troubleshooting

- Ensure Python is installed and accessible via `python` command
- Install Python dependencies: `pip install cerebras-cloud-sdk`
- Check your internet connection
- The tool uses a pre-configured API key - no setup required

## Examples

### Code Completion Examples

```bash
# Python
npx flash-tool-oss complete "def quick_sort"

# JavaScript
npx flash-tool-oss complete "function debounce" --language javascript

# React
npx flash-tool-oss complete "const MyComponent = () => {" --language javascript
```

### Chat Examples

```bash
# Programming questions
npx flash-tool-oss chat "What's the difference between let, const, and var in JavaScript?"

# Code review
npx flash-tool-oss chat "Review this code: function add(a, b) { return a + b; }"

# Debugging help
npx flash-tool-oss chat "I'm getting a TypeError in my React app, how do I debug it?"
```