Releases: isair/jarvis
Releases · isair/jarvis
v1.5.0
v1.4.0
v1.3.0
🤖 Agentic Flow & Intelligent Planning
This major update transforms how Jarvis thinks and responds by introducing an advanced agentic architecture that enables dynamic, multi-step reasoning with faster, more adaptive responses.
✨ What's New
🧠 Intelligent Agentic Planning
- Dynamic multi-step reasoning: Jarvis now thinks and plans adaptively, making decisions step-by-step rather than planning everything upfront
- Faster initial responses: No more waiting for complete plans - Jarvis starts working on your request immediately
- Smarter tool usage: Automatically chains tools together based on what it learns at each step
- Self-correcting behavior: Can adjust its approach mid-conversation based on results
⚡ Enhanced Performance
- Significantly faster responses: Planning happens on-the-go, so you get answers sooner
- Reduced redundancy: Smart duplicate detection prevents repeated tool calls
- Efficient context management: Fresh time and location data injected only when needed
- Optimized memory usage: Better clean-up of conversation state between turns
🎯 User Experience Improvements
- More natural conversations: Better handling of multi-turn interactions with proper context preservation
- Clearer progress indication: Visual feedback during multi-step operations
- Improved error recovery: Better handling of stuck states and infinite loops
🔧 Technical Improvements
Architecture
- Complete reply engine refactor: Modern agentic message loop with configurable turn limits (default: 8 turns)
- Streamlined LLM integration: New
chat_with_messagesfunction for direct model communication - Enhanced memory system: Improved
DialogueMemorywith better timing and redaction
Developer Experience
- Comprehensive test coverage: New tests for dialogue memory, redaction, and agentic flows
- Better debugging: Enhanced logging with emoji indicators and structured categories
- Cleaner codebase: Removed legacy planning code (
coach.py,planner.py) for maintainability
v1.2.0
🎤 Voice Cloning & Enhanced Memory Update
This update focuses on making Jarvis more natural and expressive through advanced TTS capabilities while significantly improving the underlying search, memory, and voice interaction systems.
✨ What's New
🎭 Voice Cloning
- Choose between two TTS engines: System TTS or the new experimental Chatterbox AI.
- Clone any voice by uploading a reference audio file! (Chatterbox only) 🎯
- Professional speech quality with configurable voice settings.
👂 More Natural Conversation
- Smarter Echo Cancellation: Jarvis is now much better at ignoring its own echo, preventing it from interrupting itself or getting confused. This works even when the transcription isn't perfect, leading to fewer accidental "stop" commands.
- Fluid Follow-ups: You can now speak in longer, more natural sentences when replying to Jarvis. It will wait for you to finish your complete thought before responding, just like a real conversation.
- More Reliable Listening: The entire voice listening pipeline is more robust, making interactions smoother and more reliable.
🧠 Smarter Memory
- Better conversation recall - Jarvis remembers context much better now.
- Faster responses with improved search capabilities.
- Enhanced keyword matching for finding past conversations.