folder-mcp

# GitHub Copilot Prompts for Windows TMOAT Testing This file provides specific prompts to use with GitHub Copilot when testing folder-mcp on Windows machines. ## Getting Started Prompts ### Initial Setup (RTX 3080) ``` I'm testing folder-mcp (a TypeScript Node.js project) on Windows with RTX 3080 GPU. The project has TMOAT tests that: - Detect machine capabilities (should detect RTX 3080 with 10GB VRAM and CUDA) - Download and test ONNX ML models (~120MB) - Test model compatibility scoring with 60/32/8 weight distribution - Handle Ollama integration gracefully when offline I have the repo cloned and synced to origin/main. I need to switch to phase-8-task-11.5 branch. Help me set up the environment and leverage the RTX 3080 for optimal testing. ``` ### Dependency Installation ``` I'm running `npm install` for folder-mcp on Windows and getting [paste error]. The project uses: - TypeScript + Vitest for testing - @xenova/transformers for ONNX models - systeminformation for hardware detection - Node.js fetch for downloads What's the Windows-specific fix for this dependency issue? ``` ## Test Execution Prompts ### Running TMOAT Tests ``` I need to run TMOAT tests for folder-mcp on Windows. These are comprehensive integration tests that: - Test machine capabilities detection (should detect Windows GPU/CPU) - Download ONNX models from HuggingFace (120MB files) - Test embedding generation (384-dimension vectors) - Validate model scoring (Language 60%, Accuracy 32%, Speed 8%) Command I'm running: `npm test -- tests/**/*.tmoat.test.ts` Help me understand what successful output should look like on Windows vs Mac. ``` ### Specific Test Suite Prompts #### Machine Capabilities Test (RTX 3080) ``` I'm running the machine capabilities TMOAT test on Windows with RTX 3080: `npm test -- tests/domain/models/model-system.tmoat.test.ts` This test detects CPU, GPU, and memory using the systeminformation package. On Mac it detects Apple M3 + Metal support (integrated GPU, unified memory). What should I expect to see with RTX 3080? Should detect: - NVIDIA GeForce RTX 3080 - 10GB VRAM - CUDA support - Discrete GPU type Help me interpret the RTX 3080 detection results vs Mac M3. ``` #### ONNX Integration Test ``` I'm running ONNX integration tests on Windows: `npm test -- tests/infrastructure/embeddings/onnx/onnx.tmoat.test.ts` This downloads Xenova/multilingual-e5-small (120MB) and generates embeddings. The download is taking [X minutes] and I'm seeing [paste progress output]. Is this normal Windows performance vs Mac? What should I troubleshoot? ``` #### Architecture Fixes Test ``` I'm testing the intelligent recommendations architecture fixes: `npm test -- tests/domain/models/intelligent-recommendations.tmoat.test.ts` This validates: - No Ollama models in curated config (clean architecture) - Hardware as binary filter (not scored in compatibility) - Scoring weights: Language 60%, Accuracy 32%, Speed 8% The test output shows [paste results]. Help me interpret if these scoring weights are correct. ``` ## Debugging Prompts ### Test Failure Analysis ``` My TMOAT test is failing on Windows with this error: [paste full error message and stack trace] The test was: [test name and description] My system specs: [Windows version, CPU, GPU, RAM] Node.js version: [version] This is testing [machine capabilities/ONNX downloads/model scoring] functionality. What are the most likely Windows-specific causes and fixes? ``` ### Performance Issues ``` My TMOAT tests are running very slowly on Windows compared to expected Mac performance: - ONNX model download: [X minutes] (expected: Y minutes) - Embedding generation: [X tokens/sec] (expected: Y tokens/sec) - Overall test suite: [X minutes] (expected: Y minutes) My specs: [CPU, GPU, RAM, disk type] Are these performance differences normal? How can I optimize for Windows? ``` ### Path and File Issues ``` I'm getting file path errors in TMOAT tests on Windows: [paste error message] The test is trying to access files at: [paste paths] This involves [model downloads/cache directories/temp files]. What's the Windows path conversion issue and how do I fix it? ``` ## Architecture Validation Prompts ### Scoring System Validation ``` I need to verify the intelligent recommendations scoring system is working correctly on Windows. The architectural requirements are: - Language compatibility: 60% weight (0-60 points) - Accuracy (MTEB scores): 32% weight (0-32 points) - Speed benchmarks: 8% weight (0-8 points) - Hardware: Binary filter only (not scored) My test results show: [paste scoring output] Help me validate these weights are implemented correctly and working as expected. ``` ### Clean Architecture Validation ``` I need to verify the Ollama architecture cleanup is working on Windows: - curated-models.json should have NO ollamaModels section - ModelCompatibilityEvaluator should exclude all Ollama references - OllamaDetector should provide basic runtime info only - Assisted mode should never show Ollama models My test results: [paste output] Help me confirm the clean separation between curated models (GPU/ONNX) and runtime models (Ollama) is working. ``` ## Windows-Specific Troubleshooting Prompts ### PowerShell Execution Policy ``` I'm getting PowerShell execution policy errors when running npm scripts for folder-mcp TMOAT tests: [paste error message] What's the minimal PowerShell policy change needed to run these Node.js/TypeScript tests safely? ``` ### Long Path Names ``` I'm getting "path too long" errors during TMOAT tests on Windows: [paste error message] This happens during [model downloads/cache operations/temp file creation]. What's the Windows long path fix for Node.js projects like folder-mcp? ``` ### Network/Download Issues ``` The ONNX model download is failing on Windows during TMOAT tests: Download URL: https://huggingface.co/Xenova/multilingual-e5-small/resolve/main/onnx/model_quantized.onnx Error: [paste error] Network: [corporate/home/etc.] What Windows network/security settings might block ML model downloads? ``` ## Expected Results Prompts ### Successful Test Output ``` Help me understand what successful TMOAT test output should look like on Windows. I should see 5 test suites passing: 1. model-system.tmoat.test.ts (machine capabilities) 2. onnx.tmoat.test.ts (model downloads) 3. model-download-management.tmoat.test.ts (coordination) 4. ollama.tmoat.test.ts (runtime detection) 5. intelligent-recommendations.tmoat.test.ts (architecture) What are the key success indicators for each suite on Windows vs Mac? ``` ### Performance Benchmarks ``` What are reasonable performance expectations for folder-mcp TMOAT tests on Windows? Test components: - Machine capabilities detection: [expected time] - 120MB ONNX model download: [expected time on typical Windows setup] - Embedding generation (384-dim vectors): [expected tokens/sec] - Full test suite: [expected total runtime] Help me set realistic Windows performance baselines vs Mac performance. ``` ## Final Validation Prompts ### Architecture Completeness ``` After running all TMOAT tests successfully on Windows, help me verify these architectural fixes are working: ✅ Ollama Clean Architecture: - No ollamaModels in curated config: [status] - Pure runtime detection only: [status] - Assisted mode exclusion: [status] ✅ Hardware Binary Filter: - Incompatible models get score=0: [status] - No hardware points in total score: [status] ✅ Scoring Weights (60/32/8): - Language prioritization working: [status] - Proper weight distribution: [status] ✅ Windows Compatibility: - Hardware detection working: [status] - Model downloads working: [status] - Performance acceptable: [status] What final checks should I do to confirm Phase 8 Task 11.5 is complete on Windows? ``` ## Usage Instructions 1. **Copy relevant prompts** based on your current testing phase 2. **Paste your specific error messages** or output where indicated 3. **Include your Windows system specs** when asking about performance 4. **Be specific about which test suite** you're running when debugging 5. **Include the full error stack trace** for debugging prompts These prompts are designed to give GitHub Copilot the context it needs to provide Windows-specific guidance for folder-mcp TMOAT testing.