@guidemode/session-processing

The analytics engine that powers GuideMode.

Parses AI coding sessions from any provider and generates actionable insights. One parser for all AI tools.

Why This Matters

Before: Each AI tool had different formats → 5+ custom parsers → maintenance nightmare

Now: Unified canonical format → 1 parser → consistent analytics for all providers

Installation

npm install @guidemode/session-processing

Usage

Parse Any Provider

import { ParserRegistry } from '@guidemode/session-processing/parsers'

const registry = new ParserRegistry()
const parser = registry.getParser('claude-code') // or gemini, copilot, codex, opencode

const parsed = parser.parse(sessionContent)
// Works the same for all providers!

Calculate Metrics

import { CanonicalSessionProcessor } from '@guidemode/session-processing/processors'

const processor = new CanonicalSessionProcessor()
const metrics = await processor.processMetrics(sessionContent, context)

// Get: performance, usage, quality, engagement, error metrics

Key Features

• ✅ One Parser - Handles all AI providers (Claude, Gemini, Copilot, Codex, OpenCode)
• ✅ Canonical Format - Universal message structure
• ✅ Rich Metrics - Performance, usage, quality, engagement, errors, context
• ✅ Provider Metadata - Preserves provider-specific features
• ✅ UI Components - React components for session display

Supported Providers

All via the same unified parser:

• Claude Code - Anthropic
• Gemini Code - Google
• GitHub Copilot - GitHub
• Codex - AI assistant
• OpenCode - Open source

For Developers

Build from Source

git clone https://github.com/guidemode/session-processing.git
cd session-processing
pnpm install
pnpm build

See CLAUDE.md for:

• Parser architecture details
• How canonical format works
• Adding new providers
• Metrics processor documentation

Tech Stack

• TypeScript with strict type safety
• Unified canonical parser
• 6 metric processors
• React UI components
• Dual build (ESM + CommonJS)

What Gets Analyzed

• Performance - Response times, token usage
• Usage - Commands executed, files modified, tools used
• Quality - Success rates, error handling
• Engagement - User interactions, turn counts
• Errors - Error detection and categorization
• Context - Context window usage, file tracking

Related Packages

• @guidemode/desktop - Converts provider formats to canonical
• @guidemode/types - Shared type definitions

License

MIT License - see LICENSE

Support

• 💬 Discussions - Ask questions, share ideas
• 🐛 Issues - Report bugs, request features
• 📧 Email: [email protected]