How brutal-45/nexus-assistant works
Nexus is a cross-platform mobile AI assistant centered on downloading models, chatting with local or remote inference engines, and managing assistant personas. The graph also preserves documented web-search, page-reading, and voice capabilities. Several store, search-provider, and screen implementations are absent from the supplied tree or excerpts, so those links are left out rather than inferred.
Main entry points: ChatScreen.tsx Chat screen · useOnboardingHandlers.ts Onboarding · App.tsx App shell · WebSearchEngine.ts Web search tool
Components 19
App experience
- App shell
App.tsx - Onboarding
src/screens/OnboardingScreens/useOnboardingHandlers.ts - Chat screen
src/screens/ChatScreen/ChatScreen.tsx - Pal interface
src/components/PalsSheets/PalSheet.tsx - Pals Hub
src/components/PalsHub/index.ts
Models and chat
- Model discovery
src/screens - Model downloads
src/services/downloads/DownloadManager.ts - Completion engines
src/api/completionEngines.ts - Chat sessions
src/repositories/ChatSessionRepository.ts - App database
src/database/index.ts
Assistant tools
- Web search tool
src/services/talents/WebSearchEngine.ts - Page reader
src/services/talents/ReadUrlEngine.ts - Voice setup
src/components/TTSSetupSheet/TTSSetupSheet.tsx - Speech engines
src/services/tts/engineRegistry.ts
Platform capabilities
- iOS shortcuts
ios/PocketPal/AppIntents/AskPalIntent.swift - Shortcut inference
ios/PocketPal/AppIntents/LlamaInferenceEngine.swift
Other components
- User
- Search provider
- Speech runtime
Connections 24
- User → App shell: opens
- App shell → Onboarding: presents
- App shell → Chat screen: navigates
- User → Onboarding: chooses setup
- Onboarding → Model discovery: registers model
- Onboarding → Pal interface: creates or updates
- User → Model discovery: browses models
- Model discovery → Model downloads: starts download
- User → Chat screen: sends message
- Chat screen → Completion engines: requests completion
- Chat screen → Chat sessions: loads and saves
- Chat sessions → App database: reads and writes
- Completion engines → Shortcut inference: uses local runtime
- Chat screen → Web search tool: invokes tool
- Chat screen → Page reader: invokes tool
- Web search tool → Search provider: searches
- Page reader → Search provider: fetches page
- Web search tool → Page reader: allows result URLs
- User → Voice setup: configures voice
- Voice setup → Speech engines: selects engine
- Speech engines → Speech runtime: dispatches speech
- User → Pals Hub: browses assistants
- User → iOS shortcuts: asks assistant
- iOS shortcuts → Shortcut inference: requests inference