ollama serve

./MatOllama.py

# Pull a model (now with concurrent download bars!)
pull qwen3

# Run by name or index
run qwen3
# or
run 1

# One-shot prompt
run qwen3 "Explain quantum computing in simple terms"

# Start conversation with one model
run qwen3:30b
> "Let's discuss machine learning basics"

# Switch to specialized model while keeping context
/switch qwen3-coder:30b
# Conversation history is preserved and transferred

# Remove models with automatic list refresh
rm 1
# ✓ Model deleted
# Updated model list automatically displayed
# No more index confusion!

# Also works with model names
rm old-experimental-model

# In chat mode, cleanly exit and unload
/unload
# ✓ Model unloaded, exited chat mode
# Memory freed, ready for next model

# Quick model switching with context
/switch 3
# ✓ Previous conversation transferred to new model

# Set theme (survives restarts)
theme magenta

# Configure temperature (persistent)
temp 0.9

# Settings stored in config.json

# Rename models efficiently
rename 1 production-model
rename 2 dev-model

# Export conversations in different formats
export json training-data.json
export text readable-chat.txt

# Save current work
save important-research

# Load previous session
load important-research.json
# Restores: model, history, settings, theme

# Export for sharing
export text research-summary.txt

./MatOllama.py --help

Options:
  --host TEXT          Ollama server URL (default: http://localhost:11434)
  --timeout FLOAT      Request timeout in seconds (default: 300.0)
  --version           Show version and exit

# Custom Ollama host
export OLLAMA_HOST=http://192.168.1.100:11434

# Run with custom settings
./MatOllama.py --host $OLLAMA_HOST --timeout 600

# Pull research-focused model
pull deepseek-coder:33b

# Start research session
run deepseek-coder:33b

# Chat about complex topics, then switch for different perspective
/switch llama3.1:70b

# Save valuable conversation
save research-session-$(date +%Y%m%d)

# Export for sharing
export text research-summary.txt

# Interactive model selection
select
# Use arrow keys to choose

# Quick code generation
run codellama "Write a Python function for binary search"

# Switch to general model for documentation
/switch llama3.1 
"Now write documentation for that function"

# Export as training data
export json code-examples.json

# List all models with clear formatting
list
# or use shorthand
ls

# Clean up old models (with auto-refresh)
rm old-model-v1
rm 3  # Remove by index, shows updated list

# Create production copies
copy experimental-model production-v1
rename development-model staging-v2

# Monitor system resources
ps

# Clean unload when switching contexts
/unload  # In chat mode
unload unused-model  # In command mode

# Check Ollama server status
curl http://localhost:11434/api/version

# Test with custom host
./MatOllama.py --host http://localhost:11434

# Verify Ollama is running
ollama serve

# Check running models
ps

# Unload unused models
unload old-model

# Clean exit from chat with unload
/unload  # Frees memory and exits chat

# Monitor during operations
run llama3 --verbose

# Models not deleting? They might be loaded
ps  # Check what's running
unload problematic-model  # Unload first
rm problematic-model     # Then delete

# Use the auto-refresh feature
rm 1  # Automatically shows updated list after deletion

# Terminal too narrow?
# Resize terminal - tables auto-adjust to width

# Color issues?
theme cyan  # Try different theme

# Text corruption?
clear  # Clear terminal buffer

# Table formatting issues?
ls  # New responsive tables adapt to terminal size

# Model not found?
pull missing-model

# Corrupted model?
rm problematic-model
pull problematic-model  # Re-download

# Out of memory?
/unload  # In chat mode - frees memory and exits
unload   # In command mode
ps       # Check what's running