v0.1.5: OAuth TTL Extension & ES Module Fixes

🚀 Key Improvements

🔐 Extended OAuth Token Lifetime

• Problem Solved: Claude Code (CLI) users were experiencing frequent re-authentication prompts (daily)
• Solution: Extended default OAuth access token TTL from 1 hour to 30 days
• Configurable: Set OAUTH_ACCESS_TOKEN_TTL in wrangler.toml for custom durations (e.g., 90 days)
• Note: Web clients (claude.ai, ChatGPT) are unaffected as they handle token refresh automatically

📦 ES Module Compatibility

• Fixed TypeScript script compatibility issues (#4)
• All scripts now properly support ES modules
• Resolved import path and module resolution errors

🛠️ Configuration

To customize OAuth token lifetime, add to your wrangler.toml:

[vars]
OAUTH_ACCESS_TOKEN_TTL = "7776000"  # 90 days in seconds

📚 Documentation Updates

• Enhanced troubleshooting guide with OAuth TTL configuration
• Added note for Claude Code users in README
• Improved configuration examples

🔧 Changes

• Extended default accessTokenTTL to 30 days
• Made TTL configurable via environment variable
• Fixed ES module imports in all TypeScript scripts
• Updated package version to 0.1.5

💡 Upgrade Instructions

1. Pull latest changes: git pull
2. (Optional) Add OAUTH_ACCESS_TOKEN_TTL to your wrangler.toml
3. Deploy: obvec deploy
4. Re-authenticate once to get the new longer-lived token

---

Full Changelog: v0.1.4...v0.1.5

v0.1.4 - ES Module Compatibility Fix

🐛 Bug Fixes

ES Module Compatibility

• Fixed ES module compatibility issues in all TypeScript scripts
• Replaced CommonJS patterns (__dirname, require.main) with ES module equivalents
• All scripts now use import.meta.url and fileURLToPath for proper module resolution

Script Improvements

• Added dotenv loading to cleanup-orphaned.ts and reset-index.ts scripts
• All CLI scripts now properly load environment variables from .env file
• Consistent ES module patterns across all TypeScript files

📚 Documentation

Module System Documentation

• Added MODULE_SYSTEM.md explaining the hybrid module configuration
• Documented why the project uses both CommonJS (for CLI) and ES modules (for Workers/scripts)
• Added clarifying comments to tsconfig.json

🧪 Testing

• All scripts tested and confirmed working:
  • ✅ obvec index - Vault indexing
  • ✅ obvec search - Note searching
  • ✅ obvec stats - Statistics display
  • ✅ obvec cleanup - Orphaned note cleanup
  • ✅ obvec reset - Index reset

🔧 Technical Details

• Package Version: 0.1.4
• Node Compatibility: Requires Node.js with ES module support
• TypeScript: Uses ESNext module compilation for Cloudflare Workers
• CLI: Maintains CommonJS compatibility for global npm installation

📦 Installation

```bash
npm install -g [email protected]
```

---

Full Changelog: v0.1.3...v0.1.4

v0.1.3 - ChatGPT Connector Integration

🎉 New Features

ChatGPT Connector Support

• Native ChatGPT Integration: Added dedicated endpoints for ChatGPT Connectors (/chatgpt/mcp)
• OpenAI-Compatible Tools: Implements exact search and fetch tool specifications required by ChatGPT
• Configurable Score Threshold: Control search sensitivity with CHATGPT_MIN_SCORE environment variable
• Optional QDF Support: Time-based result boosting with CHATGPT_USE_QDF for freshness-aware searches

📚 Documentation

• Added comprehensive ChatGPT Integration Guide with:
  • Step-by-step setup instructions
  • Architecture overview
  • Configuration options
  • Troubleshooting guide
  • QDF (Query Deserves Freshness) explanation

🔧 Improvements

• Enhanced error handling and validation across all endpoints
• Better separation of concerns with dedicated ChatGPT Durable Object
• Improved type safety with proper TypeScript definitions
• Added debug logging for troubleshooting search issues

🔒 Security

• Maintained OAuth 2.1 authentication for all endpoints
• Rate limiting applied to ChatGPT endpoints
• Input validation and sanitization for all parameters

📦 Installation

Follow the upgrade guide for existing installations or see the README for new installations.

Quick Start for ChatGPT

1. Deploy with the new configuration:
   ```bash
   npm run deploy
   ```

2. Add to ChatGPT:

   • Go to ChatGPT Settings → Connectors
   • Add your server URL: `https://your-worker.workers.dev/chatgpt/mcp\`
   • Enter your MCP_PASSWORD when prompted

3. Start searching:

   • "Search my Obsidian notes for meeting notes from last week"
   • "Find my documentation on API design patterns"

🙏 Contributors

Thanks to all contributors who helped make this release possible!

---

Full Changelog: v0.1.2...v0.1.3

v0.1.2: Critical Indexing Fixes

🚨 Critical Fixes

This release fixes critical issues that prevented proper indexing of Obsidian vaults.

What's Fixed

• Rate Limiting: Increased from 10 to 100 requests per 15 minutes
• Path Validation: Now accepts files with spaces and special characters

Impact

• Files with spaces, ampersands, parentheses, and other special characters now index correctly
• Large vaults (100+ notes) can now be indexed in a single session without hitting rate limits
• No more "Invalid path" errors for common filename patterns

Changes

• 🔧 Fixed overly restrictive path sanitization that rejected files with spaces or special characters
• 📈 Increased API rate limits to allow indexing 1000+ notes in one session
• ✅ Only truly dangerous characters are now blocked (null bytes, control chars, path traversal)

Upgrading

If you experienced indexing issues:

1. Update to v0.1.2
2. Run obvec index to re-index your vault
3. All your notes should now index successfully

Full Changelog: v0.1.1...v0.1.2

v0.1.1 - Security Update

🔒 Security Update

This release addresses critical security vulnerabilities and improves overall system robustness.

🛡️ Security Fixes

• XSS Protection: Added HTML escaping for all user inputs in OAuth pages
• CSRF Protection: Implemented one-time tokens for authorization forms
• Path Traversal Prevention: Comprehensive path validation to prevent directory traversal attacks
• Timing Attack Protection: Constant-time comparison for Bearer token authentication

🚀 Performance & Functionality Improvements

• Large Vault Support: R2 pagination now handles vaults with >1000 files correctly
• Memory Optimization: Process embeddings in batches to prevent memory exhaustion
• Error Handling: Robust validation of AI embedding responses
• Input Validation: All search parameters now properly validated (limits 1-50)
• Rate Limiting: All API endpoints now protected (10 requests per 15 minutes)

🔧 Other Improvements

• Dynamic vector dimensions from environment variables
• Improved stats endpoint documentation
• Better TypeScript type safety throughout

📦 Installation

npm install -g [email protected]

🙏 Credits

Thanks to code review feedback that identified these important security issues.

Full Changelog: v0.1.0...v0.1.1

v0.1.0 - Initial Release

🚀 Initial Release of Obsidian Vectorize MCP

Transform your Obsidian vault into a searchable, AI-accessible knowledge base using serverless Cloudflare Workers and the Model Context Protocol (MCP).

✨ Features

• Semantic Search: AI-powered search through your Obsidian notes
• MCP Integration: Works with Claude Desktop and other MCP clients
• Serverless: Runs on Cloudflare Workers with global edge deployment
• Smart Indexing: Automatic re-indexing with timestamp-based change detection
• OAuth Authentication: Secure access with OAuth 2.1 + PKCE
• CLI Tools: Complete command-line interface for management

🛠️ What's Included

• MCP server implementation with 4 semantic search tools
• Cloudflare Workers deployment with Vectorize, R2, and KV storage
• Complete CLI for indexing, searching, and managing your vault
• OAuth 2.1 authentication with secure token management
• Smart re-indexing based on file modification timestamps
• Comprehensive documentation and setup guides

📦 Installation

npm install -g obvec

🔗 Links

• Documentation
• Setup Guide
• MCP Integration

Perfect for researchers, developers, and knowledge workers who want AI-powered access to their personal notes and documentation.