Prism – Your Offline AI Hub

Download AI models once, use everywhere. Privacy-first assistant with inter-app AI hosting.

Problem • Solution • Features • Screenshots • Tech Stack • Get Started

---

The Problem

Today's mobile AI is fragmented and wasteful:

Issue	Impact
📦 Redundant Downloads	5 AI apps = 5 copies of the same model (100MB-4GB each)
🔋 Battery Drain	Multiple inference engines destroy battery life
🚫 No Interop	Apps can't share AI – each reinvents the wheel

---

The Solution

Prism is a centralized offline AI hub:

Feature	Benefit
📁 Single Model Repo	Download once, use everywhere
🔗 Inter-App Hosting	Other apps POST to localhost:8080
🔒 100% Privacy	All inference on-device, no cloud

Other Apps → POST /v1/chat/completions → Prism → Local GGUF Models

---

Screenshots

Home & Chat

Home – Daily digest with AI status, tasks, and finance summary. Chat – Real-time streaming with function calling (add tasks, log expenses). Local Model – Fully offline conversations powered by on-device GGUF models. → Source: lib/features/home/, lib/features/chat/

Second Brain

Notes – Search, tag filters, FTS5 full-text search. Editor – Inline editing with timestamps. Personas – Custom AI personalities for different use cases. → Source: lib/features/brain/

📱 Apps Hub (click to expand)

Apps – Central hub for Tasks, Finance, Files, Tools. Kanban – Drag-and-drop task board. Finance – Expense tracking with categories. → Source: lib/features/apps/

🛠️ Tools & Gateway (click to expand)

Tools – Function calling registry (add_task, log_expense, search_notes). Gateway – Local HTTP server for inter-app AI. API Playground – Swagger-like UI to test the local API with auth & streaming. MCP – Model Context Protocol server config. → Source: lib/features/apps/tools_sub_screen.dart, lib/features/apps/gateway_sub_screen.dart

⚙️ Settings (click to expand)

Settings – Providers, themes, privacy. Data – Export/import, storage management. → Source: lib/features/settings/

---

Features

Category	Features	Status
AI Providers	Local GGUF, Ollama, OpenAI, Gemini, OpenRouter	✅ Ready
Chat	Streaming, history, temporary chats, voice input	✅ Ready
Tools	add_task, log_expense, search_notes, get_weather	✅ Ready
Brain	Notes with tags, FTS5 search, personas, soul doc	✅ Ready
Apps	Tasks (Kanban), Finance, Files, Gateway	✅ Ready
ML Kit	OCR, entity extraction, smart reply, language ID	✅ Ready
Inter-App	Shelf server at localhost:8080	🔄 Preview

---

Tech Stack

Layer	Technology
Framework	Flutter 3.10.8 + Dart 3
State	Riverpod 2.6
Database	Drift (SQLite) + FTS5
AI	LangChain.dart + llama_sdk (FFI)
Inter-App	Shelf (localhost HTTP)
ML	Google ML Kit

---

Getting Started

# Clone
git clone https://github.com/Abhijeet1520/Prism.git
cd Prism

# Install
flutter pub get

# Generate DB code
dart run build_runner build --delete-conflicting-outputs

# Run
flutter run

Download Models

1. Settings → Providers → Enter Hugging Face token
2. Download Gemma 3 1B or Phi-4 Mini
3. Start chatting – 100% on-device!

---

Architecture

lib/
├── core/
│   ├── ai/           # LangChain.dart, tool registry
│   ├── database/     # Drift tables, FTS5 search
│   └── ml/           # ML Kit service
├── features/
│   ├── chat/         # AI chat interface
│   ├── brain/        # Notes, personas, soul
│   ├── home/         # Dashboard
│   ├── apps/         # Tasks, finance, files
│   └── settings/     # Providers, themes

---

Privacy

• 100% Offline – Local models, no cloud uploads
• No Telemetry – Zero analytics or tracking
• Open Source – Fully auditable

---

Roadmap

• Multi-provider AI
• Chat with tools
• Second Brain (PARA)
• Theme customization
• RAG for knowledge base
• Voice input/output
• iOS support

---

License

MIT – see LICENSE

---

Made with ❤️ by Abhijeet1520