Crate agentvet 

Validate LLM-generated tool args before execution.

When an LLM emits a tool call, the arguments may have wrong types, missing required fields, or extra fields. agentvet validates them against the same JSON Schema you sent the model and returns a ToolArgError whose for_llm renders a short human-readable hint you can feed back to the model on the next turn so it can self-correct.

Quick start

use agentvet::{ToolArgError, Validator};
use serde_json::json;

// The same JSON Schema you advertised in your tool definition.
let schema = json!({
    "type": "object",
    "properties": {
        "city": {"type": "string"},
        "units": {"type": "string", "enum": ["c", "f"]}
    },
    "required": ["city"],
    "additionalProperties": false
});
let v = Validator::from_schema(&schema).unwrap();

// Good args: validation passes.
v.validate(&json!({"city": "SFO", "units": "c"})).unwrap();

// Bad args: structured error with an LLM-friendly retry hint.
let err = v.validate(&json!({"units": "c"})).unwrap_err();
assert!(err.for_llm().contains("city"));

Structs

ToolArgError

Structured error returned when LLM-supplied tool args fail validation.

ToolArgIssue

One concrete schema violation.

Validator

Validates JSON values against a JSON Schema.