Tools & Registry

FlowgentraAI includes a tool system with built-in tools and a registry for managing them.

Tool Registry

from flowgentra_ai import ToolRegistry

# Create with built-in tools
registry = ToolRegistry.with_builtins()
print(registry.list_names())  # ["calculator", "search", ...]
print(len(registry))          # number of tools

# Call a tool
result = registry.call_tool("calculator", {
    "operation": "add",
    "a": 2,
    "b": 3,
})
print(result)  # 5

# Validate input without calling
registry.validate_input("calculator", {
    "operation": "add",
    "a": 2,
    "b": 3,
})

# Empty registry
registry = ToolRegistry()

Built-in Tools

from flowgentra_ai import CalculatorTool, SearchTool, WebRequestTool, FilesTool

Tool	Description
CalculatorTool	Basic arithmetic operations
SearchTool	Web search
WebRequestTool	HTTP requests
FilesTool	File system operations

Tool Node

Use ToolNode to integrate tool execution into a graph:

from flowgentra_ai import ToolNode, create_tool_node, store_tool_calls, check_tools_condition

JSON Schema

Define schemas for tool input/output validation:

from flowgentra_ai import JsonSchema

# Create schemas
schema = JsonSchema.object()
schema.with_description("Calculator input")
schema.with_required(["operation", "a", "b"])

# Other types
JsonSchema.string()
JsonSchema.number()
JsonSchema.integer()
JsonSchema.boolean()
JsonSchema.array()

# Validate a value
schema.validate({"operation": "add", "a": 1, "b": 2})

LLM Function Calling

Use ToolDefinition with the LLM client for function calling:

from flowgentra_ai import LLMClient, LLMConfig, Message, ToolDefinition

client = LLMClient.from_config(LLMConfig("openai", "gpt-4", api_key="sk-..."))

tools = [
    ToolDefinition(
        "calculator",
        "Perform arithmetic",
        {
            "type": "object",
            "properties": {
                "operation": {"type": "string", "enum": ["add", "subtract", "multiply", "divide"]},
                "a": {"type": "number"},
                "b": {"type": "number"},
            },
            "required": ["operation", "a", "b"],
        },
    )
]

response = client.chat_with_tools(
    [Message.user("What is 42 * 17?")],
    tools,
)

if response.has_tool_calls():
    for tc in response.tool_calls():
        print(f"Call: {tc.name}({tc.arguments})")
        result = registry.call_tool(tc.name, tc.arguments)
        print(f"Result: {result}")