Quick Reference Guide¶

Interactive Model Selection¶

All model-related make commands now feature interactive selection menus. No need to remember or type long paths!

Chat with a Model¶

make chat

What happens: 1. Shows numbered list of all available models 2. You select which model to chat with 3. Starts an interactive chat session 4. Type messages and get responses 5. Use /bye or Ctrl+D to exit

Example output:

💬 Chat with a model

📦 Available models:

  [1] llama3.2:1b (1.3 GB)
  [2] mistral:7b (4.1 GB)
  [3] my-chatbot (1.3 GB) ⭐
  [4] my-code-helper (1.3 GB) ⭐

  [0] Cancel

Select a model to chat with [0-4]: 3

🚀 Starting chat with: my-chatbot

💡 Tips:
   - Type your messages and press Enter
   - Use /bye to exit the chat
   - Use Ctrl+D to exit

───────────────────────────────────────────────────────

>>> Hello! How are you?
I'm doing great, thanks for asking! How can I help you today?

>>> Tell me a joke
Why don't scientists trust atoms? Because they make up everything!

>>> /bye
👋 Chat ended

Create a Custom Model¶

make create-model

What happens: 1. Shows numbered list of all Modelfiles in models/examples/ and models/custom/ 2. You select by number 3. You enter a name for your new model 4. Model is created

Example output:

🔨 Create a custom model

📁 Available Modelfiles:

  [1] examples/chatbot/Modelfile
  [2] examples/code-assistant/Modelfile
  [3] examples/creative-writer/Modelfile
  [4] examples/personal-assistant/Modelfile
  [5] examples/translator/Modelfile
  [6] custom/my-custom-bot/Modelfile

  [0] Enter custom path

Select a Modelfile [0-6]: 2
Enter name for the new model: my-code-helper

Save a Model for Deployment¶

make save-model

What happens: 1. Shows numbered list of all models in your Ollama instance 2. You select which model to save 3. Saves Modelfile to models/saved/

Example output:

💾 Save a model for deployment

📦 Available models:

  [1] llama3.2:1b
  [2] mistral:7b
  [3] my-chatbot
  [4] my-code-helper
  [5] production-assistant

  [0] Enter custom name

Select a model [0-5]: 3

💾 Saving model: my-chatbot
Output: ./models/saved/my-chatbot.Modelfile

✅ Model saved successfully!

Deploy a Saved Model¶

make deploy-model

What happens: 1. Shows numbered list of saved Modelfiles in models/saved/ 2. You select which one to deploy 3. Optionally rename it during deployment 4. Model is created in your Ollama instance

Example output:

🚀 Deploy a saved model

💾 Available saved models:

  [1] my-chatbot.Modelfile
  [2] production-bot.Modelfile
  [3] team-assistant.Modelfile

  [0] Enter custom path

Select a saved model [0-3]: 1
Enter model name (press Enter to use filename):

🚀 Deploying model: my-chatbot
✅ Model deployed successfully!

Direct Script Usage¶

If you prefer command-line arguments, you can still use the scripts directly:

# Create model (specify path)
bash scripts/create-custom-model.sh my-bot ./models/examples/chatbot/Modelfile

# Save model (specify name)
bash scripts/save-model.sh my-chatbot

# Deploy model (specify path)
bash scripts/deploy-model.sh ./models/saved/my-chatbot.Modelfile

# Or deploy with custom name
bash scripts/deploy-model.sh ./models/saved/my-chatbot.Modelfile renamed-bot

Cheat Sheet¶

Task	Command	Interactive?
Check requirements	`make preflight`	No
Start all services (Ollama + Chat UI)	`make up`	No
Start Ollama only (no web UI)	`make up-core`	No
Start with NVIDIA GPU	`make up-gpu`	No
Stop services	`make down`	No
View logs	`make logs`	No
Shell access	`make shell`	No
Pull base models	`make pull-base`	No
List models	`make list-models`	No
Create custom model	`make create-model`	Yes - select Modelfile
Chat with model	`make chat`	Yes - select model
Save model	`make save-model`	Yes - select model
Deploy model	`make deploy-model`	Yes - select saved file
Publish model to registry	`make publish-model`	Yes - select model
Export model + weights (tar)	`make export-full`	Yes - select model
Import full model archive	`make import-full`	Yes - select archive
Backup all models	`make backup-models`	No
Quick test	`make quick-test`	Yes - confirmation
Validation tests	`make test`	No
Clean up	`make clean`	Yes - confirmation

Choosing a Base Model (2026)¶

The project's examples default to the lightweight llama3.2:1b — keep using it for learning and quick iteration. Modern lightweight alternatives worth trying with ollama pull:

Model	Approx. size	Approx. RAM needed	Notes
`llama3.2:1b`	~1.3 GB	~4 GB	Project default
`qwen3:0.6b` / `qwen3:1.7b`	~0.5 / ~1.4 GB	~2 / ~4 GB	Small, strong; optional thinking mode
`gemma3:1b`	~815 MB	~4 GB	Very compact
`gemma3:4b`	~3.3 GB	~8 GB	Multimodal, 128K context
`phi4-mini`	~2.5 GB	~8 GB	Compact and efficient
`qwen3:8b`	~5 GB	~16 GB	Heavier, for capable machines
`gpt-oss:20b`	~13 GB	~16 GB	Heavier, for capable machines

See Essential Concepts - Choosing a Base Model for details.

Directory Structure¶

models/
├── examples/          # Pre-configured example Modelfiles
│   ├── chatbot/
│   │   └── Modelfile
│   ├── code-assistant/
│   │   └── Modelfile
│   └── ...
├── custom/            # Your custom Modelfiles
│   └── my-bot/
│       └── Modelfile
└── saved/             # Saved models ready for deployment
    ├── my-chatbot.Modelfile
    └── production-bot.Modelfile

Tips¶

Option [0] in any selection menu lets you enter a custom path/name
Saved models are stored in models/saved/ by default
Backups include timestamps: backups/models/YYYYMMDD_HHMMSS/
You can still use scripts directly with paths if you prefer
All interactive commands can be cancelled with Ctrl+C

Common Workflows¶

Quick Test of Example Model¶

make up                 # Start Ollama
make pull-base          # Get base models
make create-model       # Select example, name it
make chat               # Select your model and start chatting!

Deploy to Production¶

# On dev server:
make save-model         # Select your model

# Transfer file:
scp ./models/saved/my-model.Modelfile user@prod:/opt/ollama/models/saved/

# On prod server:
make deploy-model       # Select the transferred file

Regular Backup¶

make backup-models      # Backs up all custom models
# Files saved to: ./backups/models/YYYYMMDD_HHMMSS/

Quick Test¶

Test the complete workflow:

make quick-test

What happens: 1. Asks for confirmation 2. Creates a temporary test model 3. Sends a test prompt: "Hello! Can you introduce yourself in one sentence?" 4. Displays the response 5. Deletes the test model 6. Shows success summary

Example output:

🧪 Quick Model Test

This will:
  1. Create a test model from an example
  2. Send a test prompt to it
  3. Delete the test model

Continue? (y/N): y

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
📝 Step 1/3: Creating test model 'test-chatbot-1701234567'
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Test Details:
  Base Model: llama3.2:1b
  Modelfile:  ./models/examples/chatbot/Modelfile
  Test Model: test-chatbot-1701234567

🔨 Creating custom model: test-chatbot-1701234567
✅ Model created successfully!

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
💬 Step 2/3: Testing model with a prompt
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Sending test prompt: 'Hello! Can you introduce yourself in one sentence?'

Response:
─────────────────────────────────────────────────────
I'm a helpful AI assistant designed to provide accurate and
clear answers to your questions.
─────────────────────────────────────────────────────

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
🗑️  Step 3/3: Cleaning up test model
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

deleted 'test-chatbot-1701234567'

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
✅ Quick test completed successfully!
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Summary:
  ✓ Model creation works (base: llama3.2:1b)
  ✓ Model responds to prompts
  ✓ Model cleanup works

You can now create your own models with:
  make create-model

Note: Output is automatically cleaned for readability (ANSI escape codes removed).