Part 3 Exercises: Practice Building Tools¶
Complete these exercises to master custom tool creation.
Exercise 1: Calculator Tool (Easy)¶
Goal: Create a tool that performs mathematical calculations
Task: Create a calculator tool that can add, subtract, multiply, and divide two numbers.
Requirements:
- Function name: calculate
- Parameters: operation (string: "add", "subtract", "multiply", "divide"), num1 (float), num2 (float)
- Should handle division by zero
- Should validate the operation parameter
Ask Bob for help:
Bob, create a calculator tool for watsonx Orchestrate that can add, subtract, multiply, and divide two numbers
Test cases:
- Calculate 10 + 5 = 15
- Calculate 20 - 8 = 12
- Calculate 6 * 7 = 42
- Calculate 100 / 4 = 25
- Calculate 10 / 0 = Error
Success criteria:
- All operations work correctly
- Division by zero is handled gracefully
- Invalid operations return helpful errors
Exercise 2: Email Validator Tool (Easy)¶
Goal: Create a tool that validates email addresses
Task: Build a tool that checks if an email address is valid.
Requirements:
- Function name: validate_email
- Parameter: email (string)
- Should check for @ symbol, domain, and basic format
- Return validation result with explanation
Ask Bob for help:
Test cases:
- user@example.com → Valid
- invalid.email → Invalid (no @)
- user@domain → Invalid (no TLD)
- @example.com → Invalid (no username)
Success criteria:
- Correctly identifies valid emails
- Provides clear error messages for invalid emails
- Handles edge cases
Exercise 3: Temperature Converter Tool (Medium)¶
Goal: Create a tool that converts between temperature units
Task: Build a tool that converts temperatures between Celsius, Fahrenheit, and Kelvin.
Requirements:
- Function name: convert_temperature
- Parameters: value (float), from_unit (string), to_unit (string)
- Support: Celsius, Fahrenheit, Kelvin
- Validate that Kelvin values aren't below absolute zero
Ask Bob for help:
Test cases:
- 0°C to Fahrenheit = 32°F
- 100°C to Kelvin = 373.15K
- 32°F to Celsius = 0°C
- -300K = Error (below absolute zero)
Success criteria:
- All conversions are accurate
- Validates input ranges
- Handles all unit combinations
Exercise 4: Data Formatter Tool (Medium)¶
Goal: Create a tool that formats data in different ways
Task: Build a tool that takes structured data and formats it as JSON, CSV, or a markdown table.
Requirements:
- Function name: format_data
- Parameters: data (list of dicts), format (string: "json", "csv", "table")
- Should handle empty data
- Should create proper formatting for each type
Ask Bob for help:
Test data:
Success criteria:
- JSON output is valid JSON
- CSV output has proper headers and rows
- Table output is properly formatted markdown
- Handles empty data gracefully
Exercise 5: Text Analyzer Tool (Advanced)¶
Goal: Create a tool that analyzes text and provides statistics
Task: Build a tool that analyzes text and returns word count, character count, sentence count, and reading time.
Requirements:
- Function name: analyze_text
- Parameter: text (string)
- Calculate: word count, character count, sentence count, average word length, estimated reading time
- Reading time: assume 200 words per minute
Ask Bob for help:
Bob, create a text analyzer tool that provides statistics about a piece of text including word count, character count, and reading time
Test text:
Expected output:
- Words: 15
- Characters: 68
- Sentences: 2
- Avg word length: ~4.5
- Reading time: ~5 seconds
Success criteria:
- All counts are accurate
- Reading time calculation is reasonable
- Handles empty text
- Handles text with no punctuation
Exercise 6: Product Search Tool (Advanced)¶
Goal: Create a tool that searches a product catalog
Task: Build a tool that searches for products by name, category, or price range.
Requirements:
- Function name: search_products
- Parameters: query (string, optional), category (string, optional), min_price (float, optional), max_price (float, optional)
- Use a mock product database (list of dicts)
- Support partial name matching
- Return sorted results
Ask Bob for help:
Bob, create a product search tool with a mock database that can search by name, category, and price range
Mock database:
products = [
{"id": 1, "name": "Laptop", "category": "Electronics", "price": 999.99},
{"id": 2, "name": "Mouse", "category": "Electronics", "price": 29.99},
{"id": 3, "name": "Desk", "category": "Furniture", "price": 299.99},
{"id": 4, "name": "Chair", "category": "Furniture", "price": 199.99},
]
Test cases:
- Search "Laptop" → Find laptop
- Category "Electronics" → Find laptop and mouse
- Price range $100-$500 → Find desk and chair
- Combined filters → Multiple criteria
Success criteria:
- All search filters work
- Partial matching works
- Results are sorted by relevance
- Returns empty list when no matches
Exercise 7: Appointment Scheduler Tool (Advanced)¶
Goal: Create a tool that checks appointment availability
Task: Build a tool that checks if a time slot is available and books appointments.
Requirements:
- Function name: schedule_appointment
- Parameters: date (string), time (string), duration_minutes (int), customer_name (string)
- Maintain a simple in-memory schedule
- Check for conflicts
- Business hours: 9 AM - 5 PM, Monday-Friday
Ask Bob for help:
Test cases:
- Book 2024-03-20 at 10:00 AM for 30 minutes → Success
- Book same time again → Conflict error
- Book at 8:00 AM → Outside business hours
- Book on Saturday → Weekend error
Success criteria:
- Detects scheduling conflicts
- Validates business hours
- Validates date/time format
- Stores appointments correctly
Bonus Challenge: Multi-Tool Agent¶
Goal: Create an agent that uses multiple tools together
Task: Create a "Personal Assistant" agent that uses at least 3 of your custom tools.
Requirements:
- Agent should intelligently choose which tool to use
- Agent should chain tools when needed
- Agent should handle tool errors gracefully
Example workflow:
User: "Calculate 15% tip on $85.50 and format the result as a table"
Agent:
1. Uses calculator tool: 85.50 * 0.15 = 12.83
2. Uses formatter tool to create table with original amount, tip, and total
Ask Bob for help:
Success criteria:
- Agent uses multiple tools appropriately
- Agent chains tools when logical
- Agent provides coherent responses
- Agent handles tool failures
Debugging Exercise¶
Goal: Fix broken tools
Task: This tool has several bugs. Find and fix them!
from ibm_watsonx_orchestrate.agent_builder.tools import tool
@tool
def broken_calculator(num1, num2, operation):
"""Performs calculations"""
if operation == "add":
return num1 + num2
elif operation == "subtract":
return num1 - num2
elif operation == "multiply":
return num1 * num2
elif operation == "divide":
return num1 / num2
Problems to find:
1. Missing type hints
2. No input validation
3. Division by zero not handled
4. No return type documentation
5. Vague docstring
6. No error handling
7. Doesn't return structured response
Ask Bob for help:
Reflection Questions¶
After completing the exercises:
- What makes a good tool?
- Clear, specific purpose
- Comprehensive input validation
- Structured, consistent output
- Good error handling
-
Helpful documentation
-
How do you test tools effectively?
- Test happy path
- Test edge cases
- Test error conditions
-
Test with real agent
-
When should you create a new tool vs. modifying existing ones?
- New functionality → new tool
- Bug fix/improvement → modify existing
-
Keep tools focused and single-purpose
-
How can Bob help you build better tools?
- Generate initial code
- Review for issues
- Suggest improvements
- Write test cases
Next Steps¶
Once you've mastered custom tools, you're ready to add knowledge bases and create agent collaborators!
Continue to Part 3b: AI Gateway and external model providers →