Using Ollama Local and Remote Models with Tool Support

PS Smart Agent v3.0.2 introduces enhanced Ollama support with automatic tool detection, remote server connectivity, and a clear visual indication of which models support agentic coding workflows.

Why Tool Support Matters

For agentic coding workflows, the AI model needs to be able to use tools like:

• Read files - Access your codebase
• Write files - Create and modify code
• Execute commands - Run terminal commands
• Search code - Find relevant code patterns

Not all Ollama models support these tool capabilities. PS Smart Agent now automatically detects and displays this information.

Local Ollama Setup

1. Install Ollama

# On macOS/Linux
curl -fsSL https://ollama.ai/install.sh | sh

# On Windows, download from https://ollama.ai/download

2. Pull a Model with Tool Support

# Models known to support tools
ollama pull llama3.2
ollama pull qwen2.5
ollama pull mistral

# Check model capabilities
ollama show llama3.2 --modelfile

3. Configure PS Smart Agent

1. Open VS Code Settings (Ctrl+,)
2. Search for “PS Smart Agent”
3. Select Ollama as the API Provider
4. The default URL http://localhost:11434 works for local installations
5. Click Test Connection to verify connectivity
6. Select a model with tool support (shown with green “tools” badge)

Remote Ollama Server

Setting Up a Remote Server

If you have Ollama running on another machine (e.g., a powerful GPU server), you can connect to it remotely.

On the Server Machine:

# Start Ollama with host binding
OLLAMA_HOST=0.0.0.0:11434 ollama serve

# Or set environment variable permanently
export OLLAMA_HOST=0.0.0.0:11434

On Your Development Machine:

1. Open PS Smart Agent settings
2. Select Ollama as the API Provider
3. Enter the remote URL: http://YOUR_SERVER_IP:11434
   • Example: http://192.168.1.100:11434
4. Click Test Connection to verify
5. Models will load automatically from the remote server

Network Requirements

• Both machines must be on the same network (or have proper routing)
• Port 11434 must be accessible
• Firewall rules may need adjustment

# Check if remote Ollama is accessible
curl http://192.168.1.100:11434/api/tags

Tool Support Indicators

When you select Ollama as your provider, PS Smart Agent shows clear visual indicators:

Models WITH Tool Support

• ✅ Green “tools” badge
• ✅ Selectable - Click to use the model
• ✅ Full agentic capabilities - Can read, write, execute

Models WITHOUT Tool Support

• ⚠️ Red “no tools” badge
• ⚠️ Disabled (grayed out) - Cannot be selected
• ⚠️ Chat only - Can answer questions but cannot modify code

Models Known to Support Tools

Model	Tool Support	Notes
llama3.2	✅ Yes	Recommended for most tasks
qwen2.5	✅ Yes	Good for code generation
mistral	✅ Yes	Fast and capable
codellama	✅ Yes	Specialized for code
deepseek-coder	✅ Yes	Excellent for coding tasks
llama2	❌ No	Chat only
gemma	❌ No	Chat only
phi3	❌ No	Chat only

Test Connection Feature

The Test Connection button helps diagnose connectivity issues:

Success States

• ✅ Connected - Server is reachable
• ✅ Models loaded - Available models are displayed

Error States

• ❌ Connection refused - Ollama not running or wrong URL
• ❌ No models found - Server running but no models installed
• ❌ Timeout - Network issues or firewall blocking

Troubleshooting

Models Not Showing

1. Check Ollama is running:

   curl http://localhost:11434/api/tags
   

2. For remote servers, verify connectivity:

   curl http://YOUR_SERVER_IP:11434/api/tags
   

3. Check firewall settings:

   # Linux
   sudo ufw allow 11434
      
   # Windows - Add firewall rule for port 11434
   

Connection Errors

Error	Solution
Connection refused	Start Ollama: ollama serve
No models found	Pull a model: ollama pull llama3.2
Timeout	Check firewall and network
Wrong URL	Verify URL format: http://IP:PORT

Model Tool Support Not Detected

If a model you know supports tools shows “no tools”:

1. Update Ollama to the latest version
2. Re-pull the model: ollama pull MODEL_NAME
3. Restart Ollama: ollama serve

Best Practices

For Local Development

• Use models with tool support for coding tasks
• Keep frequently used models pulled locally
• Monitor GPU memory usage with large models

For Remote Servers

• Use a wired connection for stability
• Consider network latency when selecting models
• Run multiple models on a powerful server, access from lightweight machines

Model Selection Tips

• Coding tasks: Use models with tool support (green badge)
• Quick questions: Any model works
• Large codebases: Prefer models with larger context windows

Example Workflow

1. Start Ollama server (local or remote)
2. Open PS Smart Agent settings
3. Select Ollama provider
4. Enter URL (localhost or remote IP)
5. Click Test Connection
6. Select a model with tool support (green badge)
7. Start coding with full agentic capabilities!

Conclusion

PS Smart Agent’s Ollama integration makes it easy to use both local and remote models for agentic coding. The tool support indicators help you choose the right model for the task, ensuring you have full capabilities when you need them.

For more information, visit pyshine.com or check the VS Code Marketplace.