llm_tool_macros

Crate: llm_tool_macros · Path: crates/llm_tool_macros
Description (Cargo.toml): Attribute macros to register Rust functions as LLM tools with JSON Schema via schemars

Role

Proc-macro crate that provides the #[llm_tool] attribute. Annotating a plain Rust function generates companion functions that expose the tool’s JSON Schema (via schemars) and an executor, ready to be collected by llm_tools::registry_from!.

Feature flags

From crates/llm_tool_macros/Cargo.toml:

No feature flags declared.

Macro: `#[llm_tool]`

Attributes

Attribute	Type	Required	Purpose
`name`	`String`	No	Override the tool name (defaults to function name)
`description`	`String`	No	Tool description sent to the LLM
`context`	flag	No	Marks the tool as context-aware (requires a `ToolContext` parameter)

Generated items

For a function my_func, the macro generates:

Generated item	Signature	Purpose
`my_funcParams`	`struct`	Serde + JsonSchema struct derived from function parameters
`my_func_llm_tool_info()`	`() -> (String, String, Value)`	Returns `(name, description, json_schema)`
`my_func_llm_tool_exec(args)`	`(Value) -> Value`	Deserializes args and calls the original function
`my_func_llm_tool_needs_context()`	`() -> bool`	Whether the tool needs a `ToolContext`
`my_func_llm_tool_exec_ctx(args, ctx)`	`(Value, Box<dyn Any + Send>) -> Value`	Context-aware executor

Example sketch

use llm_tool_macros::llm_tool;

#[llm_tool(name = "get_weather", description = "Get current weather for a city")]
fn get_weather(city: String, units: String) -> String {
    format!("Weather in {city} ({units}): sunny, 22°C")
}

// The macro generates get_weather_llm_tool_info, get_weather_llm_tool_exec, etc.
// Collect into a registry:
let registry = llm_tools::registry_from!(get_weather);

Context-aware tools

#[llm_tool(name = "search", description = "Search with context", context)]
fn search(query: String, ctx: ToolContext) -> String {
    // ctx is injected at runtime via the context executor
    format!("Searching for: {query}")
}

See the LLM Tools guide for end-to-end examples of defining, registering, and invoking LLM tools.

Full API reference: cargo doc -p llm_tool_macros --no-deps