Crate nano_agent 

Structured LLM agents in Rust: JSON-schema input/output, system prompt building, optional chat history, context providers, and optional tools. Backed by genai (e.g. Gemini with JSON response format).

Features

Feature	Purpose
searxng	SearXNG search tool
calculator	Expression evaluation tool
webpage_scraper	Fetch and simplify web pages

Example

use nano_agent::{
    AgentConfig, BasicChatInputSchema, BasicNanoAgent, NanoAgent,
    context::{ChatHistory, SystemPromptGenerator},
};
use schemars::JsonSchema;
use serde::{Deserialize, Serialize};

#[derive(Serialize, Deserialize, JsonSchema)]
struct Out {
    chat_message: String,
}

let mut agent = BasicNanoAgent::<BasicChatInputSchema, Out>::new(
    AgentConfig::new("gemini-2.5-flash-lite")
        .with_system_prompt_generator(SystemPromptGenerator::new())
        .with_chat_history(ChatHistory::new()),
);
let _reply: Out = agent
    .run(BasicChatInputSchema {
        chat_message: "Hello".into(),
    })
    .await?;

Re-exports

pub use genai;

Modules

context

tools

Structs

AgentConfig

Agent configuration.

AgentInner

Inner agent state.

BasicChatInputSchema

Basic chat input schema.

BasicChatOutputSchema

Basic chat output schema.

BasicNanoAgent

Wraps AgentInner and implements NanoAgent with no extra state or hooks.

Enums

AgentError

Agent error.

Traits

NanoAgent

Schema-driven chat agent: system prompt, optional history, structured I/O.

Attribute Macros

async_trait