Meet Tian AI: Your Completely Offline AI Assistant for Android
The Problem with Cloud AI
Every time you ask ChatGPT, Claude, or Gemini a question, your conversation gets sent to a remote server. Your data, your privacy concerns, your confidential information — all processed by someone else's computer.
But what if you could run a genuinely intelligent AI assistant entirely on your phone, with no internet connection, no data leaving your device, and no subscription fees?
Meet Tian AI.
What Is Tian AI?
Tian AI is an open-source, completely offline AI assistant that runs on Android (via Termux). It's not a thin client that sends your queries to an API — it's a full-stack local AI system with:
- Local LLM Inference — Runs Qwen2.5-1.5B (GGUF quantized) via llama.cpp on your phone's CPU
- 34GB Local Knowledge Base — Millions of concepts across 100+ domains, indexed and instantly searchable
- Three-Layer Thinking Engine — Fast response, Chain-of-Thought reasoning, and Deep analysis modes
- Self-Evolution System — Accumulates XP, unlocks new capabilities, and upgrades its own version
- Self-Modifying Code — Can analyze and improve its own source code
- Complete Privacy — Zero data leaves your device. Ever.
Architecture
┌──────────────────────────────────────────────────┐
│ Web UI (HTML/JS) │
├──────────────────────────────────────────────────┤
│ Flask API Server (run.py) │
├──────────┬──────────┬──────────┬─────────────────┤
│ Thinker │ Memory │ Search │ Evolution │
│ Fast/CoT │ Short/ │ Web + │ XP + Leveling │
│ /Deep │ Long │ Local │ Tier System │
├──────────┴──────────┴──────────┴─────────────────┤
│ llama.cpp + Qwen2.5-1.5B GGUF (Local LLM) │
├──────────────────────────────────────────────────┤
│ 34GB SQLite Knowledge Base (Knowledge_base.db) │
└──────────────────────────────────────────────────┘
Key Features
🧠 Three-Layer Thinking
- Fast Mode: Quick responses for simple queries
- Chain-of-Thought: Step-by-step reasoning for complex problems
- Deep Mode: Multi-perspective analysis with reflection
📚 34GB Knowledge Base
- Pre-built with 69,000+ concepts across 100 domains
- 30 question patterns per concept for flexible retrieval
- Instant indexed search (0.04-0.1s response time)
🔒 100% Offline & Private
- No internet connection required after setup
- All data stays on your device
- No accounts, no tracking, no surveillance
🤖 Self-Evolving
- XP system rewards deeper conversations
- Version upgrades unlock new capabilities
- Can modify its own source code for continuous improvement
📱 Android Native
- Runs in Termux on any Android device
- Optimized for aarch64 architecture
- ~7 tok/s generation speed on mid-range phones
Getting Started
# Install Termux from F-Droid
# Then:
pkg update && pkg upgrade
pkg install python git clang cmake
git clone https://github.com/yourusername/tian-ai
cd tian-ai
pip install -r requirements.txt
# Download model (Qwen2.5-1.5B GGUF)
# Place in ~/storage/downloads/qwen-1.5b-q4.gguf
# Start llama-server
llama-server -m ~/storage/downloads/qwen-1.5b-q4.gguf --port 8080 -t 4 -c 2048
# Run Tian AI
python run.py
# Or open the standalone HTML in your browser
Tech Stack
| Component | Technology |
|---|---|
| LLM Engine | llama.cpp + Qwen2.5-1.5B GGUF |
| Backend | Flask (Python) |
| Frontend | Pure HTML/CSS/JS (standalone) |
| Knowledge Base | SQLite (34GB, indexed) |
| Self-Modify | AST analysis + LLM suggestions |
| Search | DuckDuckGo + Google (optional) |
| Payments | USDT (TRC-20) / BTC |
Comparison
| Feature | Tian AI | ChatGPT | Ollama |
|---|---|---|---|
| Fully Offline | ✅ | ❌ | ✅ |
| Built-in Knowledge Base | ✅ 34GB | ❌ | ❌ |
| Three-Layer Thinking | ✅ | ❌ | ❌ |
| Self-Evolution | ✅ | ❌ | ❌ |
| Self-Modifying Code | ✅ | ❌ | ❌ |
| Runs on Android | ✅ | ❌ | ❌ |
| Privacy (no data leaves) | ✅ | ❌ | Depends |
| Open Source | ✅ | ❌ | ✅ |
Support the Project
Tian AI is completely free and open source. If you'd like to support development:
USDT (TRC-20): TNeUMpbwWFcv6v7tYHmkFkE7gC5eWzqbrs
BTC: bc1ph7qnaqkx4pkg4fmucvudlu3ydzgwnfmxy7dkv3nyl48wwa03kmnsvpc2xv
Tian AI — Your Private AI, Completely Offline.
Top comments (0)