Local AI Assistant - Changli UI

---

🚧 Status: This project will no longer be developed, because every time Ollama releases an update, new bugs appear.
In the future, I will rebuild it as a standalone application, without relying on any other software. 🚧

---

📖 Description

Local AI Assistant is an offline/local AI chat application featuring:

• Backend built with Flask → API communication and chat storage,
• Frontend (UI) built with PySide6 (Qt) → interactive chat interface.
• Ollama Integration → run local AI models (default: gemma3:4b, now supports model selection).
• Multi-language system with 9 languages and layered fallback.

This project is designed to run AI fully locally, with customizable identity, persona, profile, memory, and chat history.

---

📂 Project Structure

local-ai-assistant/
├─ app.py                    
│
├─ backend/
│  ├─ i18n/
│  │   ├─ ar.json
│  │   ├─ en_gb.json
│  │   ├─ en_us.json
│  │   ├─ id.json
│  │   ├─ es.json
│  │   ├─ ja.json
│  │   ├─ ko.json
│  │   ├─ pt.json
│  │   └─ zh.json
│  ├─ locales/
│  │   ├─ ar.json
│  │   ├─ en_gb.json
│  │   ├─ en_us.json
│  │   ├─ id.json
│  │   ├─ es.json
│  │   ├─ ja.json
│  │   ├─ ko.json
│  │   ├─ pt.json
│  │   └─ zh.json
│  ├─ config.py      
│  ├─ core.py           
│  ├─ i18n.py           
│  ├─ persona.py            
│  ├─ storage.py             
│  ├─ ollama_client.py       
│  └─ routes.py              
│
├─ ui/
│  ├─ main.py               
│  ├─ chat_window.py         
│  ├─ client.py              
│  ├─ worker.py                            
│  └─ widgets/              
│     ├─ settings.py         
│     ├─ history.py          
│     ├─ bubbles.py          
│     └─ identity.py        
│
├─ data/                    
│  ├─ chat_history.json
│  ├─ chat_sessions.json
│  ├─ ui_chat_config.json
│  └─ config.json
├─ config.json
├─ requirement.txt  
└─ README.md                 

🏗 Architecture

---

⚡ Features

• 🖥 Modern UI using PySide6 (Qt)
• 📝 Chat History → rename, delete, or continue past sessions
• 🎭 Custom Persona → change AI name, user name, and prompt
• 👤 Profile System → add personal info (e.g. What do you do, Anything else the AI should know)
• 🧠 Chat Memory → AI remembers up to 32 previous messages
• 🌍 Multi-language Support → 9 languages, native names in UI, with layered fallback (en_us → id → target) English (US, UK), Bahasa Indonesia, 日本語, 한국어, 中文（简体）, Português, Español, العربية.
• 🎨 Custom Background → solid color or custom image
• ⚙️ Flask Backend with /chat, /chats, /config endpoints
• 🤖 Ollama Integration → run local AI models; default gemma3:4b, now supports model selection
• 📂 All data & configs stored locally under data/

---

🛠️ Installation & Setup

1. Clone Repository

git clone https://github.com/rillToMe/local-ai-assistant.git
cd local-ai-assistant

2. Create Virtual Environment (recommended)

python -m venv venv
source venv/bin/activate     # Linux/Mac
venv\Scripts\activate        # Windows

3. Install Dependencies

pip install -r requirements.txt

Main dependencies:

• flask
• flask-cors
• requests
• PySide6

⚠️ Note: Make sure you have installed Ollama and the required models (gemma3:4b by default). Other models can be selected if installed.

4. Run the App

python app.py

• Flask backend will start on http://127.0.0.1:5000
• PySide6 UI will automatically open

---

🎮 Usage

• Settings → change background, update Identity & Prompt
• History → view, rename, or continue previous sessions
• Profile → add info about yourself (used by AI in responses)
• Language → choose UI language from 9 available
• Models → select available Ollama models installed on your system

---

🧩 Model Recommendations

Based on official benchmarks and community tests:

Model	Min RAM (CPU-only)	Approx. GPU VRAM (BF16 / 4-bit)	Notes
gemma3:1b	≥ 2 GB RAM	~1.5 GB / ~0.9 GB	Lightweight - runs on old notebooks, but slow (~7–10 tokens/sec) getdeploying.com, windowscentral.com
qwen3:1.8b	≥ 2 GB RAM	~2 GB / ~1 GB (est.)	Slightly better reasoning - light enough for laptops
gemma3:4b	≥ 4 GB RAM	~6.4 GB / ~3.4 GB	Recommended default - good speed & quality getdeploying.com, ai.google.dev
qwen3:4b	≥ 4 GB RAM	~6 GB / ~3 GB (est.)	Balanced - strong chat & reasoning
gemma3:12b	≥ 9 GB RAM	~20 GB / ~8.7 GB	Requires strong GPU or high RAgetdeploying.com, ai.google.dev
qwen3:8b	≥ 9 GB RAM	~18 GB / ~8 GB (est.)	Good quality & context
deepseek-r1:8b	≥ 9 GB RAM	~18 GB / ~8 GB (est.)	Specialized reasoning
gemma3:27b	≥ 18 GB RAM	~46 GB / ~21 GB	Heavy - best on high-end GPUs or servers getdeploying.com, ai.google.dev
gpt-oss:20b	≥ 32 GB RAM	~40 GB / ~20 GB (est.)	Large - better long context
gpt-oss:120b	≥ 128 GB RAM / Multi-GPU	~120 GB+ / 60 GB+ (est.)	Experimental - extremely heavy compute requirement

---

📌 Roadmap / Todo

• Fix UI crash bugs
• Add multi-tab chat support
• Add chat export/import
• Optimize performance for long requests
• Expand model integration beyond Ollama

---

⚠️ Notes

• This is still in early development, expect frequent bugs and issues
• UI/UX is minimal for now, focused on core functionality
• Default persona is simple, but can be extended with custom prompts