A robust function calling implementation for the Gemma language model, enabling seamless integration with external APIs, databases, and system operations.
- Dynamic Function Registry for runtime function management
- Robust Input Validation with type checking
- Comprehensive Error Handling and logging
- Automatic Schema Management
- Extensible Design for custom function integration
- Comprehensive Test Coverage
- Python 3.8 or higher
- pip (Python package installer)
- Make (optional, for using Makefile commands)
- Clone the repository:
git clone https://github.com/your-org/gemma-fncall-flow.git
cd gemma-fncall-flow- Install dependencies:
make install # or pip install -e ".[test,dev]"# Run tests
make test # or pytest tests/ -v
# Check code quality
make lint # check code quality
make format # auto-format code
# Clean up temporary files
make cleanfrom gemma_fncall_flow import FunctionRegistry, GemmaFunctionCaller
from validation import ParameterType
# Initialize registry and caller
registry = FunctionRegistry()
caller = GemmaFunctionCaller(registry)
# Define function schema
schema = {
"name": "greet",
"description": "Greet a person by name",
"parameters": {
"name": {
"type": ParameterType.STRING.value,
"description": "Name of the person to greet",
"required": True
}
},
"required": ["name"]
}
# Define and register function
def greet(name: str) -> str:
return f"Hello, {name}!"
registry.register("greet", greet, schema)
# Call function
result = caller.call_function("greet", {"name": "Alice"})
print(result) # Output: Hello, Alice!gemma-fncall-flow/
├── tests/ # Test suite
│ ├── test_function_calling.py
│ ├── test_validation.py
│ ├── test_weather_example.py
│ ├── test_file_system_example.py
│ └── test_database_example.py
├── gemma_fncall_flow/ # Core implementation
│ ├── __init__.py
│ ├── function_calling.py
│ └── validation.py
├── setup.py # Package configuration
├── requirements.txt # Dependencies
├── requirements-test.txt # Test dependencies
├── Makefile # Development commands
└── README.md # Documentation
registry = FunctionRegistry()
# Register function
registry.register_function(my_function)
# Get function schema
schema = registry.get_function_schema("function_name")
# List registered functions
functions = registry.list_functions()caller = GemmaFunctionCaller(registry)
# Call function
result = caller.call_function(
"function_name",
{"param1": "value1"}
)
# Get function documentation
doc = caller.get_function_documentation("function_name")- Fork the repository
- Create a feature branch
- Commit your changes
- Push to the branch
- Create a Pull Request
This project is licensed under the MIT License - see the LICENSE file for details.