Skip to content

Latest commit

 

History

History
101 lines (77 loc) · 4.13 KB

HOW_TO_USE_LMSTUDIO.md

File metadata and controls

101 lines (77 loc) · 4.13 KB

🚀 How to Use LMStudio with ComfyUI-Copilot

LMStudio allows you to run powerful language models locally on your machine. Here's a step-by-step guide to set it up with ComfyUI-Copilot.

📥 Step 1: Install LMStudio

  1. Download LMStudio from https://lmstudio.ai/
  2. Install the application on your computer
  3. Launch LMStudio

🤖 Step 2: Download a Model

  1. Open LMStudio
  2. Go to the "Discover" tab
  3. Search for a model. Recommended options:
    • Llama 3.1 8B Instruct (good balance of performance and quality)
    • CodeLlama 7B (excellent for code-related tasks)
    • Mistral 7B Instruct (fast and efficient)
    • Phi-3.5 Mini (very fast, smaller model)
  4. Click "Download" for your chosen model

🔧 Step 3: Start LMStudio Server

  1. Go to the "Developer" tab in LMStudio
  2. Select your downloaded model from the dropdown
  3. Click "Start Server"
  4. Note the server URL (usually http://localhost:1234)
    • If you changed the port in LMStudio, use that port number instead

⚙️ Step 4: Configure ComfyUI-Copilot

  1. Open ComfyUI-Copilot in your browser
  2. Click the settings/configuration button (usually in the top-right area)
  3. Look for "LLM Configuration" or "API Configuration" section
  4. Expand the configuration panel if it's collapsed
  5. Configure the settings:
    • API Key: Leave this completely empty (LMStudio doesn't need one)
    • Server URL: Enter http://localhost:1234/v1 (or your custom port)
      • If you changed LMStudio's port to 1235, use: http://localhost:1235/v1
      • Always add /v1 at the end
  6. Click "Verify" to test the connection
  7. Save the configuration

✅ Step 5: Verify It's Working

  1. Test the connection - you should see "LMStudio connection successful!"
  2. Try a simple chat - ask ComfyUI-Copilot a question
  3. Check model availability - the model dropdown should show your LMStudio model

🔍 Troubleshooting

❌ "Connection Failed" Error

  • Check LMStudio is running: Make sure the server is started in LMStudio
  • Verify the URL: Ensure you're using the correct port and added /v1
  • Check firewall: Make sure port 1234 (or your custom port) isn't blocked
  • Restart LMStudio: Sometimes restarting the server helps

❌ "No Models Found"

  • Load a model: Make sure you've selected and loaded a model in LMStudio
  • Wait for loading: Some models take time to load completely
  • Check LMStudio logs: Look at the LMStudio console for error messages

❌ "API Key Error"

  • Leave API key empty: LMStudio doesn't require API keys
  • Clear any existing key: Remove any text from the API key field

🎯 Common LMStudio URLs

  • Default: http://localhost:1234/v1
  • Custom port: http://localhost:YOUR_PORT/v1
  • Network access: http://YOUR_IP:1234/v1 (if accessing from another machine)

💡 Pro Tips

  1. Model Performance: Larger models give better results but need more RAM/GPU
  2. GPU Acceleration: Enable GPU acceleration in LMStudio for faster responses
  3. Context Length: Adjust context length in LMStudio based on your needs
  4. Multiple Models: You can switch models in LMStudio without reconfiguring ComfyUI-Copilot
  5. Offline Usage: Once set up, everything works offline!

🆚 LMStudio vs OpenAI

Feature LMStudio OpenAI API
Cost Free (after initial setup) Pay per token
Privacy 100% local Data sent to OpenAI
Speed Depends on your hardware Usually very fast
Models Open source models GPT-4, GPT-3.5, etc.
Setup Requires installation Just API key
Offline ✅ Works offline ❌ Needs internet

🎉 You're Ready!

Once configured, you can use all ComfyUI-Copilot features with your local LMStudio models:

  • Workflow generation and debugging
  • Parameter suggestions
  • Model recommendations
  • Chat assistance
  • And more!

Everything runs locally on your machine for complete privacy and control.