Fusion 360 MCP

# Complete Setup Guide - FusionMCP with Ollama This guide walks you through setting up FusionMCP with Ollama (100% free and offline). ## Why Ollama? - ✅ **100% Free** - No API costs - ✅ **Privacy** - Runs locally, data never leaves your computer - ✅ **Offline** - Works without internet - ✅ **Fast** - Direct local execution - ✅ **Easy** - Simple installation ## Step-by-Step Setup ### Step 1: Install Ollama #### macOS ```bash # Method 1: Download from website # Visit https://ollama.com and download Ollama.app # Drag to Applications folder # Method 2: Using Homebrew brew install ollama # Method 3: Using curl curl -fsSL https://ollama.com/install.sh | sh ``` #### Linux ```bash curl -fsSL https://ollama.com/install.sh | sh ``` #### Windows ```bash # Download from https://ollama.com/download/windows # Run the installer # Or use WSL2 and follow Linux instructions ``` ### Step 2: Verify Ollama Installation ```bash # Check if Ollama is installed ollama --version # Should show something like: # ollama version 0.1.20 ``` ### Step 3: Pull a Model Ollama needs to download a model first. We recommend `llama3` for best results: ```bash # Pull Llama 3 (best for CAD tasks) ollama pull llama3 # Alternative models: ollama pull mistral # Faster, smaller ollama pull codellama # Good for technical tasks ollama pull phi # Smallest, fastest ``` **Model Sizes** (approximate): - `phi` - 1.6GB (fastest, simplest tasks) - `mistral` - 4.1GB (good balance) - `llama3` - 4.7GB (recommended, best quality) - `codellama` - 7GB (technical focus) **Note**: First download will take 5-15 minutes depending on your internet speed. ### Step 4: Test Ollama ```bash # Test the model ollama run llama3 "What is 2+2?" # Should respond with "4" or similar # Exit with Ctrl+D or type /bye ``` ### Step 5: Start Ollama Server (Background) Ollama needs to run as a background service: #### macOS/Linux ```bash # Start Ollama server ollama serve # Or run in background: nohup ollama serve > /dev/null 2>&1 & ``` #### Windows ```bash # Ollama runs automatically as a service after installation # Check if running: curl http://localhost:11434 ``` **Verify Server**: ```bash # Should return Ollama version info curl http://localhost:11434/api/tags ``` ### Step 6: Install FusionMCP #### Clone Repository ```bash cd ~/Desktop # or your preferred directory git clone https://github.com/yourusername/fusion360-mcp.git cd fusion360-mcp ``` Or download ZIP and extract: ```bash # If you downloaded the ZIP cd ~/Downloads unzip fusion360-mcp-main.zip cd fusion360-mcp-main ``` #### Create Virtual Environment ```bash # Create virtual environment python3 -m venv venv # Activate it source venv/bin/activate # macOS/Linux # OR venv\Scripts\activate # Windows ``` You should see `(venv)` in your terminal prompt. #### Install Dependencies ```bash # Install all required packages pip install -r requirements.txt # This will install: # - FastAPI (web server) # - Uvicorn (ASGI server) # - Pydantic (data validation) # - aiohttp (async HTTP) # - loguru (logging) # - and more... ``` ### Step 7: Configure for Ollama #### Create config.json ```bash # Copy example config cp examples/example_config.json config.json ``` #### Edit config.json Open `config.json` in your text editor and modify: ```json { "ollama_url": "http://localhost:11434", "default_model": "ollama:llama3", "fallback_chain": ["ollama:llama3"], "mcp_host": "127.0.0.1", "mcp_port": 9000, "allow_remote": false, "log_level": "INFO", "log_dir": "logs", "cache_enabled": true, "cache_type": "json", "cache_path": "context_cache.json", "timeout_seconds": 60, "max_retries": 3, "retry_delay": 1.0, "models": { "ollama": { "available": ["llama3", "mistral", "codellama", "phi"], "default": "llama3" } } } ``` **Key settings for Ollama**: - `ollama_url`: Should be `http://localhost:11434` - `default_model`: Use `"ollama:llama3"` (or your chosen model) - `fallback_chain`: Just `["ollama:llama3"]` for offline-only setup - Leave other API keys empty (not needed for Ollama) ### Step 8: Start MCP Server #### Option 1: Using Helper Script (Easiest) ```bash # macOS/Linux ./start_server.sh # Windows start_server.bat ``` #### Option 2: Direct Command ```bash # Make sure virtual environment is activated source venv/bin/activate # macOS/Linux # or venv\Scripts\activate # Windows # Start server python -m mcp_server.server ``` **Expected Output**: ``` 2025-01-15 10:30:00 | INFO | Logger initialized with level INFO 2025-01-15 10:30:00 | INFO | Logs will be saved to logs/mcp_server.log 2025-01-15 10:30:00 | INFO | Created JSON cache at context_cache.json 2025-01-15 10:30:00 | INFO | Cache initialized: json 2025-01-15 10:30:00 | INFO | System prompt loaded 2025-01-15 10:30:00 | INFO | Initialized MCP Router with providers: ['ollama'] 2025-01-15 10:30:00 | INFO | MCP Server started on 127.0.0.1:9000 INFO: Started server process [12345] INFO: Waiting for application startup. INFO: Application startup complete. INFO: Uvicorn running on http://127.0.0.1:9000 (Press CTRL+C to quit) ``` **✅ Server is ready!** Keep this terminal open. ### Step 9: Test MCP Server Open a **new terminal** and run: ```bash # Test 1: Health check curl http://127.0.0.1:9000/health # Expected response: # {"status":"healthy","providers":["ollama"],"cache_enabled":true} # Test 2: List models curl http://127.0.0.1:9000/models # Expected response: # {"models":{"ollama":["llama3","mistral",...]}} # Test 3: Ask model to create geometry curl -X POST http://127.0.0.1:9000/mcp/command \ -H "Content-Type: application/json" \ -d '{ "command": "ask_model", "params": { "provider": "ollama", "model": "llama3", "prompt": "Create a 20mm cube" }, "context": { "active_component": "RootComponent", "units": "mm", "design_state": "empty" } }' ``` **Expected Response**: ```json { "status": "success", "message": "Action generated successfully", "actions_to_execute": [ { "action": "create_box", "params": { "width": 20, "height": 20, "depth": 20, "unit": "mm" }, "explanation": "Creating a 20mm cube", "safety_checks": ["dimensions_positive", "units_valid"] } ], "llm_response": { "success": true, "provider": "ollama", "model": "llama3", ... } } ``` ### Step 10: Install Fusion 360 Add-in #### Find Fusion Add-ins Folder **macOS**: ```bash ~/Library/Application Support/Autodesk/Autodesk Fusion 360/API/AddIns/ ``` **Windows**: ```bash %APPDATA%\Autodesk\Autodesk Fusion 360\API\AddIns\ ``` #### Copy Add-in **macOS**: ```bash # From fusion360-mcp directory cp -r fusion_addin ~/Library/Application\ Support/Autodesk/Autodesk\ Fusion\ 360/API/AddIns/FusionMCP # Verify ls ~/Library/Application\ Support/Autodesk/Autodesk\ Fusion\ 360/API/AddIns/FusionMCP/ # Should show: main.py, ui_dialog.py, fusion_actions.py, FusionMCP.manifest, utils/ ``` **Windows**: ```bash # From fusion360-mcp directory xcopy /E /I fusion_addin "%APPDATA%\Autodesk\Autodesk Fusion 360\API\AddIns\FusionMCP" # Verify dir "%APPDATA%\Autodesk\Autodesk Fusion 360\API\AddIns\FusionMCP" ``` **Important**: The folder name MUST be `FusionMCP` (case-sensitive). ### Step 11: Load Add-in in Fusion 360 1. **Open/Restart Fusion 360** 2. **Go to Scripts and Add-Ins**: - Click **Tools** → **Add-Ins** → **Scripts and Add-Ins** - Or press **Shift+S** 3. **Find FusionMCP**: - Click **Add-Ins** tab - Look for **FusionMCP** in the list - If you don't see it, click the **+** (green plus) button and browse to the folder 4. **Run the Add-in**: - Select **FusionMCP** - Click **Run** button - You should see: "FusionMCP Add-in Started! Use the 'MCP Assistant' command..." 5. **Optional - Auto-run on Startup**: - Check the **Run on Startup** checkbox - Add-in will load automatically next time ### Step 12: Use FusionMCP in Fusion 360 #### Open MCP Assistant 1. Look for **MCP Assistant** button in toolbar 2. Or go to **Tools** menu → Find **MCP Assistant** 3. Click to open the assistant dialog #### Test Commands Try these natural language commands: **Simple Geometry**: ``` Create a 20mm cube ``` **Cylinder**: ``` Make a cylindrical shaft 10mm diameter and 50mm long ``` **Complex Design**: ``` Design a mounting bracket 100x50mm with 4 M5 holes at the corners ``` **With Position**: ``` Create a 5mm hole at position x=25, y=25 ``` **Material**: ``` Apply aluminum material to all bodies ``` #### What Happens 1. You type the command 2. Add-in sends to MCP server 3. Ollama (llama3) processes the request 4. Returns structured JSON action 5. Add-in executes in Fusion 360 6. Geometry appears! ### Step 13: Monitor and Debug #### Check MCP Server Logs ```bash # View logs in real-time tail -f logs/mcp_server.log # Or open in text editor cat logs/mcp_server.log ``` #### Check Conversation History ```bash # View cached conversations curl http://127.0.0.1:9000/history?limit=5 ``` #### Debug Issues **Issue**: "Connection refused" ```bash # Make sure MCP server is running curl http://127.0.0.1:9000/health # If not working, restart server python -m mcp_server.server ``` **Issue**: "Ollama not found" ```bash # Check Ollama is running curl http://localhost:11434/api/tags # If not running, start it ollama serve ``` **Issue**: "Model not found" ```bash # Pull the model again ollama pull llama3 # List available models ollama list ``` ## Complete Startup Checklist Every time you want to use FusionMCP: ### Terminal 1 - Ollama (if not auto-running) ```bash ollama serve ``` ### Terminal 2 - MCP Server ```bash cd ~/Desktop/fusion360-mcp # or your path source venv/bin/activate python -m mcp_server.server ``` ### Fusion 360 1. Open Fusion 360 2. Load FusionMCP add-in (if not auto-run) 3. Click MCP Assistant 4. Start designing! ## Performance Tips ### Faster Response Times 1. **Use Smaller Models** (if llama3 is slow): ```bash ollama pull phi # Smallest, fastest ``` Update `config.json`: ```json "default_model": "ollama:phi" ``` 2. **Lower Temperature** (more consistent): ```json "temperature": 0.3 // More precise, less creative ``` 3. **Reduce Max Tokens**: ```json "max_tokens": 1000 // Faster responses ``` ### GPU Acceleration Ollama automatically uses GPU if available: ```bash # Check GPU usage (macOS) sudo powermetrics --samplers gpu_power # Check GPU usage (Linux with nvidia) nvidia-smi # Check GPU usage (Windows) # Task Manager → Performance → GPU ``` ### Memory Management Large models use RAM: - `phi`: ~2GB RAM - `llama3`: ~6GB RAM - `codellama`: ~8GB RAM If slow, use smaller model or close other apps. ## Example Workflows ### Workflow 1: Simple Box ``` User: "Create a 50mm cube" → Ollama generates: {"action": "create_box", "params": {"width": 50, ...}} → Fusion creates 50x50x50mm box ⏱️ Time: ~3-5 seconds ``` ### Workflow 2: Mounting Bracket ``` User: "Design a mounting plate 100x50mm with 4 holes" → Ollama generates action sequence: 1. Create base plate 100x50x5mm 2. Add hole 1 at (10,10) 3. Add hole 2 at (90,10) 4. Add hole 3 at (10,40) 5. Add hole 4 at (90,40) → Fusion executes all 5 actions ⏱️ Time: ~8-12 seconds ``` ### Workflow 3: Parametric Design ``` User: "Create a shaft where diameter is 1/5 of the length, length 100mm" → Ollama calculates: diameter = 100/5 = 20mm → Generates: {"action": "create_cylinder", "params": {"radius": 10, "height": 100}} → Fusion creates 20mm diameter x 100mm shaft ⏱️ Time: ~5-8 seconds ``` ## Troubleshooting ### Common Issues #### 1. MCP Server Won't Start **Error**: `Address already in use` **Solution**: ```bash # Find process using port 9000 lsof -i :9000 # macOS/Linux netstat -ano | findstr :9000 # Windows # Kill the process kill -9 <PID> # macOS/Linux taskkill /PID <PID> /F # Windows # Or change port in config.json "mcp_port": 9001 ``` #### 2. Ollama Connection Failed **Error**: `Connection refused to localhost:11434` **Solution**: ```bash # Start Ollama server ollama serve # Or check if it's running ps aux | grep ollama # macOS/Linux tasklist | findstr ollama # Windows ``` #### 3. Slow Responses **Solution**: ```bash # Use faster model ollama pull phi # Update config.json: "default_model": "ollama:phi" # Or reduce max_tokens # In config.json: "max_tokens": 500 ``` #### 4. Fusion Add-in Not Visible **Solution**: - Verify folder name is exactly `FusionMCP` - Check manifest file exists: `FusionMCP.manifest` - Restart Fusion 360 - Try manual add: Scripts & Add-Ins → + → Browse to folder #### 5. Import Errors in Python **Error**: `ModuleNotFoundError: No module named 'fastapi'` **Solution**: ```bash # Make sure virtual environment is activated source venv/bin/activate # Reinstall dependencies pip install -r requirements.txt ``` ## Next Steps 1. **Try More Models**: ```bash ollama pull mistral ollama pull codellama ``` 2. **Customize System Prompt**: - Edit `prompts/system_prompt.md` - Adjust behavior and output format 3. **Add API Keys** (optional, for cloud models): - Edit `config.json` - Add OpenAI/Gemini/Claude keys - Enable fallback chain 4. **Explore Examples**: - Check `examples/test_conversation.json` - Try example commands 5. **Run Tests**: ```bash ./run_tests.sh # or pytest tests/ -v ``` ## Quick Reference ### Essential Commands ```bash # Start Ollama ollama serve # Pull model ollama pull llama3 # Activate venv source venv/bin/activate # Start MCP server python -m mcp_server.server # Test server curl http://127.0.0.1:9000/health # View logs tail -f logs/mcp_server.log # Run tests pytest tests/ -v ``` ### File Locations ``` ~/Desktop/fusion360-mcp/ # Project root ├── config.json # Your config ├── venv/ # Virtual environment ├── logs/mcp_server.log # Server logs └── context_cache.json # Conversation history ~/Library/Application Support/Autodesk/Autodesk Fusion 360/API/AddIns/FusionMCP/ # Fusion add-in (macOS) %APPDATA%\Autodesk\Autodesk Fusion 360\API\AddIns\FusionMCP\ # Fusion add-in (Windows) ``` ## Success Checklist - [ ] Ollama installed and running - [ ] Model downloaded (llama3 or phi) - [ ] Python 3.11+ installed - [ ] Virtual environment created and activated - [ ] Dependencies installed (`pip install -r requirements.txt`) - [ ] config.json configured for Ollama - [ ] MCP server running (http://127.0.0.1:9000) - [ ] Health check passes (`curl http://127.0.0.1:9000/health`) - [ ] Fusion add-in copied to correct folder - [ ] Fusion add-in loaded and running - [ ] MCP Assistant button visible in Fusion - [ ] Test command works ("Create a 20mm cube") ## 🎉 You're Ready! Once all checkboxes are ✅, you can: - Design with natural language - Create parametric CAD models - Use AI for geometric reasoning - Work 100% offline and free - Maintain complete privacy **Have fun building with FusionMCP + Ollama!** 🚀 --- **Need Help?** Check README.md or open an issue on GitHub