A lightweight, self-hosted replacement for Open-WebUI designed for local hardware. Zero bloat. Minimal dependencies. Optimized for Python 3.13 on Debian 13 (Trixie).

• Persistent User Profile — Automatically injects user context (hardware specs, rules of engagement, identity) into every chat.
• System Prompt Presets — Toggle between "Coding Companion", "Sysadmin", or custom personas instantly.
• Streaming Chat — Real-time token streaming with stop-token support to prevent hallucination.
• Conversation History — Stored locally in SQLite (jarvischat.db). No external DB required.
• Model Switcher — Select any installed Ollama model (DeepSeek, Llama3, Mistral) per request.
• Responsive UI — Full Markdown support, syntax highlighting, and copy-to-clipboard for code blocks.
• LAN Accessible — Host on jarvis, access from llamadev (Windows/WSL), mobile, or any network node.

• OS: Linux (Tested on Debian 13 Trixie / Ubuntu 24.04)
• Python: 3.11+ (Compatible with Python 3.13)
• Backend: Ollama running locally (localhost:11434)

# On jarvis (as your standard user):
cd /opt/jarvischat
./install.sh

Since Debian 13 enforces PEP 668, you must use a virtual environment.

# 1. Prepare directory
mkdir -p ~/jarvischat
cp app.py ~/jarvischat/
cd ~/jarvischat

# 2. Create and activate venv
python3 -m venv venv
source venv/bin/activate

# 3. Install dependencies (FastAPI, Uvicorn, HTTPX)
pip install fastapi uvicorn httpx

# 4. Run manually to test
./venv/bin/python3 app.py

Access the UI via browser at: http://<IP-ADDRESS>:8080

To run JarvisChat as a background service:

1. Copy jarvischat.service to /etc/systemd/system/.
2. Update the User= and ExecStart= paths in the file.
3. Enable and start:

sudo systemctl daemon-reload
sudo systemctl enable --now jarvischat

• app.py — Single-file application (FastAPI backend + Embedded HTML/JS frontend).
• jarvischat.db — SQLite database (Created automatically on first run).
• jarvischat.service — Systemd unit file.
• requirements.txt — Dependency list.