mirror of
https://github.com/hanzoskill/smart-followups.git
synced 2026-10-30 21:07:59 +00:00
2.6 KiB
2.6 KiB
Changelog
All notable changes to Smart Follow-up Suggestions will be documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
[1.0.0] - 2026-01-20
🎉 Initial Release
Added
-
CLI Tool (
cli/followups-cli.js)- Standalone command-line interface for generating follow-ups
- Support for multiple output modes: JSON, Telegram, text, compact
- Context parsing from various input formats
- Integration with Claude Haiku API
- Proper error handling and validation
-
Bot Integration (
handler.js)/followupscommand support- Auto-trigger mode (optional)
- Channel detection (inline buttons vs text mode)
- Support for Telegram, Discord, Slack, Signal, iMessage, SMS
-
Documentation
- README.md: Feature overview and quick start
- SKILL.md: Comprehensive Bot integration guide
- examples.md: Channel-specific output examples
- INTERNAL.md: Architecture and design decisions
- QUICKSTART.md: 5-minute setup guide
-
Features
- 3 contextual suggestions per generation (1 per category)
- 3 categories: Quick (⚡), Deep Dive (🧠), Related (🔗)
- Mobile-optimized UI (3 buttons = no scrolling on Telegram)
- Context-aware analysis of last 1-3 exchanges
- Sub-second latency with Claude Haiku
- Cost-effective (~$0.0001 per generation)
Technical Details
- Uses
@anthropic-ai/sdkv0.32.0 - Node.js 18+ required
- Temperature: 0.7 for optimal diversity
- Max tokens: 1024
- Context window: Last 3 exchanges
Design Decisions
- 3 suggestions (not 6): Mobile UX testing showed 3 buttons are cleaner and less cluttered on Telegram mobile, reducing decision fatigue while maintaining category diversity
- One per category: Quality over quantity - one well-crafted suggestion per category beats multiple mediocre ones
Known Issues
- None
Migration Guide
- N/A (initial release)
[Unreleased]
Planned for v1.1.0
- Caching layer for repeated contexts
- Rate limiting implementation
- User feedback tracking
- Personalization based on user profile
- Multi-language support (i18n)
- Improved error messages
- Unit tests
- Integration tests
Under Consideration
- Fine-tuned domain-specific models
- Conversation memory (avoid repetitive suggestions)
- Batch processing for high-traffic scenarios
- Webhook support for external integrations
- Analytics dashboard
Version History
- 1.0.0 (2026-01-20): Initial release
Note: For detailed technical changes, see INTERNAL.md