Stark-style Flutter desktop assistant with voice control, text chat, local LLM responses, memory context, and command automation.
- Stark-inspired HUD interface with interactive panels.
- Voice input (
speech_to_text) and spoken output (flutter_tts). - Text chat command bar for direct typed interaction.
- Local Ollama chat integration.
- Web search summarization flow.
- Local memory + context retrieval (RAG-style prompts).
- Optional terminal command generation/execution.
- Personality modes:
classic,tactical,casual. - Linux TTS fallback (
spd-saythenespeak-ng) when plugin speech is unavailable.
- Flutter + Dart
- Riverpod state management
- Dio for HTTP calls
- SharedPreferences for settings persistence
- Ollama local API (
/api/generate,/api/embeddings)
- Flutter SDK (Dart
^3.11.1) - Ollama running locally at
http://localhost:11434/api - One or more installed local models
- Linux desktop target recommended
Optional Linux speech fallback packages:
sudo apt install -y speech-dispatcher espeak-ngflutter pub get
flutter run -d linuxollama serve
ollama pull jarvis-7b:latest
ollama pull llama3.1:8b
ollama pull llama3.2:3b
ollama pull mistral-nemo:12b- Launch the app.
- Tap the center core to toggle voice listening.
- Use the bottom command bar for text chat.
- Watch response/status logs in the right panel.
- Change model and personality mode from Settings.
classic: calm, witty, JARVIS-like.tactical: direct, mission-focused.casual: friendly, concise.
Location:
- Settings -> Personality Mode
lib/main.dartlib/features/jarvis/jarvis_screen.dartlib/features/jarvis/jarvis_provider.dartlib/core/settings_provider.dartlib/core/ollama_service.dartlib/core/memory_service.dart
- No model response:
- Ensure
ollama serveis running. - Confirm selected model exists locally.
- Ensure
- No voice output on Linux:
- Confirm system audio is working.
- Install
speech-dispatcherandespeak-ng.
- Text chat not sending:
- Use Enter or send button in the bottom command bar.
No license file is currently defined in this repository.