# logpare

> Semantic log compression library for LLM context windows. Uses the Drain algorithm to extract templates from repetitive log data, achieving 60-90% token reduction while preserving diagnostic information.

## Quick Start

```typescript
import { compress } from 'logpare';

const logs = [
  'ERROR Connection to 192.168.1.100 failed',
  'ERROR Connection to 192.168.1.101 failed',
  'INFO Request processed in 45ms',
];

const result = compress(logs);
console.log(result.formatted);
// === Log Compression Summary ===
// Input: 3 lines → 2 templates (66.7% reduction)
//
// Top templates by frequency:
// 1. [2x] ERROR Connection to <*> failed
// 2. [1x] INFO Request processed in <*>
```

## Installation

```bash
npm install logpare
```

## Core API

- `compress(lines: string[], options?): CompressionResult` - Compress log lines into templates
- `compressText(text: string, options?): CompressionResult` - Compress multiline text
- `createDrain(options?): Drain` - Create Drain instance for incremental processing

## Key Options

```typescript
interface DrainOptions {
  depth?: number;           // Parse tree depth (default: 4)
  simThreshold?: number;    // Similarity threshold 0-1 (default: 0.4)
  maxChildren?: number;     // Max children per node (default: 100)
  maxClusters?: number;     // Max templates (default: 1000)
  format?: 'summary' | 'detailed' | 'json';
}
```

## Output Formats

- `summary` - Compact template list with frequencies
- `detailed` - Full templates with sample variables and metadata
- `json` - Machine-readable JSON output

## Template Structure

Each template contains:
- `pattern: string` - Template with `<*>` wildcards
- `occurrences: number` - Match count
- `sampleVariables: string[][]` - Captured variable values
- `severity: 'error' | 'warning' | 'info'` - Auto-detected severity
- `urlSamples`, `statusCodeSamples`, `durationSamples` - Diagnostic metadata

## CLI Usage

```bash
# Compress log file
logpare compress access.log

# With options
logpare compress --format detailed --depth 5 app.log

# Pipe from stdin
cat logs/*.log | logpare compress
```

## Links

- [GitHub Repository](https://github.com/logpare/logpare)
- [npm Package](https://www.npmjs.com/package/logpare)
- [MCP Server](https://www.npmjs.com/package/@logpare/mcp)

## MCP Integration

For AI coding assistants (Claude, Cursor, etc.), use the MCP server:

```bash
npm install -g @logpare/mcp
```

Configure in Claude Desktop or Cursor to enable direct log compression from your AI assistant.