ML Jupyter MCP

#!/bin/bash # Test exact Claude usage patterns # This validates the patterns documented in the README echo "Testing Claude Usage Patterns" echo "==============================" # Colors for output GREEN='\033[0;32m' RED='\033[0;31m' NC='\033[0m' # Test counter TESTS_PASSED=0 TESTS_FAILED=0 # Function to run a Claude test run_claude_test() { local test_name="$1" local test_input="$2" local expected_pattern="$3" echo -e "\n📝 Testing: $test_name" # Run the test OUTPUT=$(echo "$test_input" | claude 2>&1) EXIT_CODE=$? if [ $EXIT_CODE -ne 0 ]; then echo -e "${RED}✗ Command failed${NC}" TESTS_FAILED=$((TESTS_FAILED + 1)) return 1 fi # Check for expected pattern if echo "$OUTPUT" | grep -qi "$expected_pattern"; then echo -e "${GREEN}✓ Pattern found: $expected_pattern${NC}" TESTS_PASSED=$((TESTS_PASSED + 1)) return 0 else echo -e "${RED}✗ Pattern not found: $expected_pattern${NC}" echo "Output preview: ${OUTPUT:0:200}..." TESTS_FAILED=$((TESTS_FAILED + 1)) return 1 fi } # Pattern 1: Always start with initialization echo -e "\n### Pattern 1: Initialization First ###" run_claude_test "Initialization Pattern" \ "Use jupyter-executor to: 1. Call jupyter_initialize(working_dir='.') 2. Get the session_id from the result 3. Print 'Initialized with session:' and the session_id" \ "session" # Pattern 2: Execute code with rich feedback echo -e "\n### Pattern 2: Code Execution ###" run_claude_test "Execution Pattern" \ "Use jupyter-executor to: 1. Initialize first 2. Execute this code: import pandas as pd; print('Pandas imported') 3. Show the output" \ "pandas" # Pattern 3: ImportError → UV install → retry echo -e "\n### Pattern 3: Error Recovery via UV ###" TEST_INPUT='Use jupyter-executor to: 1. Initialize the environment 2. Try to import matplotlib: import matplotlib 3. If it fails with ImportError: - Use jupyter_ensure_dependencies(session_id, ["matplotlib"]) to install via UV - NEVER use pip install - Retry the import 4. Confirm success' run_claude_test "UV Package Installation" "$TEST_INPUT" "matplotlib" # Check that pip was NOT used echo -e "\n🔍 Verifying UV-only approach..." LAST_OUTPUT=$(echo "$TEST_INPUT" | claude 2>&1) if echo "$LAST_OUTPUT" | grep -q "pip install" && ! echo "$LAST_OUTPUT" | grep -q "jupyter_ensure_dependencies"; then echo -e "${RED}✗ CRITICAL: pip was used instead of UV!${NC}" TESTS_FAILED=$((TESTS_FAILED + 1)) else echo -e "${GREEN}✓ UV-first principle maintained${NC}" TESTS_PASSED=$((TESTS_PASSED + 1)) fi # Pattern 4: Get guidance when needed echo -e "\n### Pattern 4: Guidance System ###" run_claude_test "Guidance Pattern" \ "Use jupyter-executor to: 1. Initialize 2. Call jupyter_get_guidance('fix_error', context={'error_type': 'ModuleNotFoundError'}) 3. Show the guidance" \ "jupyter_ensure_dependencies\|guidance\|fix" # Pattern 5: Complete workflow echo -e "\n### Pattern 5: Complete Workflow ###" WORKFLOW_TEST='Use jupyter-executor following this exact pattern: # 1. Always start with initialization result = jupyter_initialize(working_dir=".") session_id = result["session_id"] # 2. Execute code with rich feedback jupyter_execute_cell(session_id, "x = 42") # 3. Show persistence jupyter_execute_cell(session_id, "print(f\"x = {x}\")") Print the final result' run_claude_test "Complete Workflow" "$WORKFLOW_TEST" "42" # Summary echo -e "\n==============================" echo "Test Summary" echo "==============================" echo -e "Passed: ${GREEN}$TESTS_PASSED${NC}" echo -e "Failed: ${RED}$TESTS_FAILED${NC}" if [ $TESTS_FAILED -eq 0 ]; then echo -e "\n${GREEN}✅ All Claude pattern tests passed!${NC}" exit 0 else echo -e "\n${RED}❌ Some Claude pattern tests failed${NC}" exit 1 fi