The Schema Macro
Derive JSON Schema from Rust types for tool specs.
#[derive(Schema)] generates a schemars::JsonSchema implementation for your Rust type. Appam uses that schema when building tool specs so the model can see the expected shape of tool arguments.
It does not add standalone JSON Schema validation at runtime by itself; runtime behavior still comes from normal deserialization in the generated tool code.
Basic Usage
use appam::prelude::*;
#[derive(Deserialize, Schema)]
struct SearchInput {
#[description = "The search query string"]
query: String,
#[description = "Maximum number of results to return"]
max_results: u64,
#[description = "Include archived items in the results"]
include_archived: bool,
}#[description = ...]
Use #[description = "..."] on fields to populate the schema's description field:
#[derive(Deserialize, Schema)]
struct FileInput {
#[description = "Absolute or relative path to the file"]
path: String,
#[description = "Whether to create parent directories if they do not exist"]
create_dirs: bool,
}Supported Shapes
The derive delegates to schemars, so the usual JsonSchema-compatible types work well:
- strings
- booleans
- integers and floats
Option<T>Vec<T>- nested structs
- enums
Optional Fields
Option<T> fields are omitted from the generated "required" set:
#[derive(Deserialize, Schema)]
struct QueryInput {
#[description = "The search query"]
query: String,
#[description = "Maximum number of results"]
max_results: Option<u64>,
}Pairing with #[tool]
This derive is most useful when a tool takes a single structured input:
use appam::prelude::*;
#[derive(Deserialize, Schema)]
struct GrepInput {
#[description = "Regex pattern to search for"]
pattern: String,
#[description = "File path to search"]
path: String,
}
#[tool(description = "Search for a regex inside a file")]
fn grep(input: GrepInput) -> Result<String> {
Ok(format!("searching {} in {}", input.pattern, input.path))
}In that shape, the generated tool implementation uses the struct's schema when producing ToolSpec::parameters.