Releases: michael-borck/talk-buddy
Releases · michael-borck/talk-buddy
2.11.0
2.10.1
Talk Buddy 2.10.1 — Supply chain, auto-updates, dark mode Three small additions on top of 2.10.0: Dependabot: Weekly grouped PRs for npm + GitHub Actions deps, Electron major bumps blocked from automation (those need a dedicated session). Security advisories still come through immediately. Auto-updater: Production builds check GitHub Releases 30s after launch via electron-updater. When a newer version exists, downloads in the background and prompts on app quit. First release where this is wired — users on 2.10.1+ will auto-detect future releases. Older users need one manual upgrade to enter the loop. Dark mode: Settings → Appearance → Light / Dark / Match system (default). Uses the Studio Calm dark tokens (warm near-black paper, brighter sage accent for WCAG AA contrast). Designed for late-night practice. Theme flips live without reload.
2.10.0
Talk Buddy 2.10.0 — Polish release Quality-of-life arc on top of 2.9.7's streaming pipeline. What's new: - Push-to-talk toggle mode (Settings → Voice) — friendlier than hold-to-talk for longer turns or motor-difficulty users - Replay button works on every AI message, including resumed sessions from the library (re-synthesises on first click via TTS, caches result for subsequent clicks) - Visualiser voice rings now ripple in time with replay audio - 60s connection timeout + 1 retry on hosted-provider streams (OpenAI / Groq / Anthropic / Gemini) — flaky calls fail clean instead of hanging silently until the user hits Esc - Streaming token cap raised 400 → 2048 (barge-in handles long responses naturally now) - Pause now also stops an active replay (matches Esc and spacebar) Cleanup: - ~280 lines of orphan code deleted (webspeech.ts whole file plus nine unused exports across chat / speechProvider / embedded)
2.9.7
Talk Buddy 2.9.7 — Polish release Builds on v2.9.6's streaming voice pipeline with the UX layer that makes it feel natural rather than just functional. Live transcript: - New transcript pane on the left of the conversation page, auto-scrolls as the AI's reply streams in token-by-token. - Partial AI responses are preserved on Esc or spacebar barge-in instead of being discarded — you keep what you saw. Rehear / replay: - Every AI message in the transcript gets a small speaker icon to replay it (using the cached audio from the original synthesis, no second TTS call). Works for interrupted messages too — replays whatever was synthesised before the cut. - Clear playing state: button switches to a stop square with REPLAYING label, the message gets a sage left border. Esc, the button, or starting a new turn all stop the replay. Pause / Save & exit / End session: - Three honest actions instead of one misleadingly-named "Pause": Pause = freeze in place (audio paused mid-utterance, resume continues exactly where it left off); Save & exit = persist and navigate away (was the old "Pause"); End session = finalise + analysis. Keyboard: - Esc silences the AI without recording — for when you've got the gist and want quiet to think. - Spacebar still records and barge-ins as before. Status hints beneath the visualiser update per state to remind you which key does what. Landing page: - Hero rewritten to "Practice the conversation before the conversation." - Right column replaces the editorial "Edition / № 2026" with a keystroke cheat sheet so every visit reinforces the patterns. Settings: - Test button now actually validates the API key with an authenticated call to each provider's model-list endpoint instead of pinging the base URL with no headers (which always passed and gave false confidence). - LLM errors now surface verbatim in the failure toast instead of a generic "Voice pipeline failed." Fixes: - Ghost audio leak: navigating away mid-response no longer leaves an orphaned TTSPipeline playing in the background. - npm run electron:dev no longer launches two app windows. - Initial scenario greeting now also gets a replay button.
2.9.6
v2.9.6 Talk Buddy 2.9.6 — Pass --config explicitly so CI loads electron-buil…
2.6.2
Bump version to 2.6.2 - Previous build completed but assets weren't uploaded - Triggering new build with fixed workflow 🤖 Generated with [Claude Code](https://claude.ai/code) Co-Authored-By: Claude <noreply@anthropic.com>
v2.6.1 - Zero TypeScript Errors Achievement
🎯 Key Achievement
ZERO TypeScript Errors - From 107 errors to 0!
✨ What's New Since v2.5.9
🤖 Multi-Provider Chat Support
- Anthropic (Claude) - Full support with correct API headers
- OpenAI (GPT) - Complete integration
- Groq - Fast inference support
- Ollama - Local model support
- Custom Endpoints - Flexible configuration
📊 Code Quality Improvements
- ✅ Complete type safety throughout codebase
- ✅ Commercial-grade TypeScript standards
- ✅ Proper error handling and null safety
- ✅ Clean, maintainable architecture
🐛 Bug Fixes
- Fixed IPC response parsing (Buffer/Uint8Array conversion)
- Resolved embedded server port allocation issues
- Fixed all type comparison errors in switch statements
- Corrected session state management
- Fixed Date constructor null safety issues
🏗️ Architecture Enhancements
- New TypeScript interfaces for settings (
types/settings.ts) - Custom React hooks (
useSettings,useModelFetcher) - Started component extraction (
STTSettings.tsx) - Foundation for complete SettingsPage refactoring
👨💻 Developer Experience
- Zero errors in strict TypeScript mode
- Proper type annotations everywhere
- Improved code organization and structure
- Ready for future enhancements
📦 Installation
Download the appropriate installer for your platform from the assets below.
🔄 Upgrading
This release maintains full backward compatibility. Simply install over your existing version.
📝 Notes
This release represents a major milestone in code quality and reliability. The codebase now meets commercial-grade standards with zero TypeScript errors.
Full Changelog: v2.5.9...v2.6.1
v2.6.0 - Commercial-Grade Code Quality & Multi-Provider Support
🎉 Major Improvements
✨ Zero TypeScript Errors
- Fixed ALL 107 TypeScript errors - now ZERO errors!
- Achieved commercial-grade code quality standards
- Complete type safety throughout the codebase
🤖 Multi-Provider Chat Support
- ✅ Anthropic (Claude) - Full support with proper headers
- ✅ OpenAI (GPT models) - Complete integration
- ✅ Groq - Fast inference support
- ✅ Ollama - Local model support
- ✅ Custom endpoints - Flexible configuration
🎤 Enhanced Speech Services
- Improved embedded STT/TTS with pywhispercpp
- Dynamic port allocation for embedded server
- Proper service isolation - no more cross-contamination
- Fallback mechanisms for reliability
🏗️ Code Architecture Improvements
- New TypeScript interfaces for settings management
- Custom React hooks (
useSettings,useModelFetcher) - Component extraction started (
STTSettings.tsx) - Foundation laid for complete SettingsPage refactoring
- Proper separation of concerns
🐛 Bug Fixes
- Fixed IPC response parsing (Uint8Array to string conversion)
- Fixed embedded server port mismatch issues
- Fixed session end reason types
- Fixed Date constructor null safety
- Resolved all unused variable warnings
👨💻 Developer Experience
- Zero errors and warnings in TypeScript
- Proper type annotations everywhere
- Clean, maintainable, scalable code
- Ready for future enhancements
💥 Breaking Changes
None - Full backward compatibility maintained
📝 Notes
This release represents a major step forward in code quality and reliability. The codebase is now production-ready with commercial-grade standards.
🙏 Acknowledgments
Special thanks to all contributors and testers who helped make this release possible.
Full Changelog: v2.5.9...v2.6.0
Talk Buddy v2.5.10
Talk Buddy Release v2.5.10
Downloads
Windows
Talk Buddy-Setup-*.exe- Windows installer
macOS
Talk Buddy-*.dmg- macOS installer
Linux
Talk Buddy-*.AppImage- Universal Linux appTalk Buddy-*.deb- Debian/Ubuntu package
Installation Instructions
Windows: Download and run the .exe installer
macOS: Download the .dmg file, open it, and drag Talk Buddy to Applications
Linux:
- AppImage: Make executable with
chmod +xand run - DEB: Install with
sudo dpkg -i Talk\ Buddy-*.deb
Requirements
- Microphone access for speech recognition
- Internet connection for AI services
- Speaches server for text-to-speech (configurable)
- Ollama or OpenAI API for conversation AI (configurable)
What's New
- Collapsible dark-themed sidebar
- Resume in-progress sessions
- Improved TTS with gender-specific models
- Better error handling for STT/TTS services
- Configurable speech speed
- Many UI/UX improvements
2.5.3
Bump version to 2.5.3 🤖 Generated with [Claude Code](https://claude.ai/code) Co-Authored-By: Claude <noreply@anthropic.com>