-
Notifications
You must be signed in to change notification settings - Fork 0
Installation
| Component | Minimum |
|---|---|
| Python | 3.10 + |
| Terminal | 256-color (true-color recommended for arena gradients) |
| Network | Reachable subnet to your model hosts |
The runtime dependencies are slim — see requirements.txt:
rich>=13.7.0
httpx>=0.26.0
asyncio-throttle>=1.0.0
prompt-toolkit>=3.0.0
No compilers, no native extensions, no GPU drivers required on the client side.
git clone https://github.com/mtecnic/model-chat-cli.git
cd model-chat-cli
python -m venv venv
source venv/bin/activate
pip install -r requirements.txtVerify:
python main.pyYou should see a network scan progress bar, followed by the discovered-models table.
run.sh activates the venv and launches the app in one go:
./run.shUse this if you want a one-shortcut launch (e.g. terminal launcher, tmux pane, desktop entry).
The repo isn't yet packaged for PyPI, but you can install from source into an isolated env:
pipx install --editable git+https://github.com/mtecnic/model-chat-cli.gitThen run with model-chat-cli (if a console_scripts entry is added later) or by invoking python -m main inside the pipx venv.
| Path | What it is |
|---|---|
~/.model_chat_cache.json |
Discovered servers / models cache |
~/.model_chat_history/ |
Saved conversations (when storage/history.py is wired) |
./logs/stress_test_*.log |
Stress-test logs (per-run, INFO console / DEBUG file) |
./arena_*.md |
Exported arena tournaments |
./chat__*.md |
Exported chat sessions |
Delete ~/.model_chat_cache.json to force a full rescan on next launch.
git pull
pip install -r requirements.txt --upgradeCache format is forward-compatible; if a future change breaks it, the app will detect the mismatch and rescan rather than crashing.
Nothing is installed system-wide. To fully remove:
rm -rf ~/model-chat-cli ~/.model_chat_cache.json ~/.model_chat_historyGetting started
Features
Internals
Operating