Agent Smith

An api to create local first human friendly agents in the browser or Nodejs

📚 Read the documentation

• Libraries
  • Agent
    • Get started
    • Tools
    • Templates
    • Supervision
  • Task
    • Get started
    • Specification
    • Use tasks
    • Models
    • Templates
    • Variables
    • Tools
  • Transient memory
    • Get started
    • Usage
    • Api
  • Semantic memory
    • Get started
    • Initialize
    • Write operations
    • Read operations
• Terminal client
  • Install
  • Overview
  • Config
  • Tasks
  • Actions
  • Workflows
  • Commands
  • Tools call
  • Mcp
• Plugins
  • Overview
  • Inference
  • Vision
  • Filesystem
  • Code
    • Git
• Server
  • Get started
  • Api

Check the 💻 examples

What is an agent?

An agent is a language model that can take decisions. It can:

• Think: use language model servers to perform inference queries
• Work: manage long running workflows with multiple tasks, using tools
• Remember: use semantic memory to store data
• Interact: perform interactions with the user

Packages

Version	Name	Description	Nodejs	Browser
	@agent-smith/cli	Terminal client	✅	❌
	@agent-smith/agent	Agent	✅	✅
	@agent-smith/task	Task	✅	✅
	@agent-smith/smem	Semantic memory	✅	❌
	@agent-smith/tfm	Templates for models	✅	✅

Terminal client plugins

Philosophy

• Composable: the packages have limited responsibilities and can work together
• Declarative: focus on the business logic by expressing features simply
• Explicit: keep it simple and under user control: no hidden magic

Requirements

Supported inference servers:

• Llama.cpp
• Koboldcpp
• Ollama
• Any server that supports an Openai compatible api

Example

Terminal client

Simple inference query (using the inference plugin):

lm q list the planets of the solar system

Query a thinking model, use qwq (from the models plugin)::

lm think "solve this math problem: ..." m=qwq

Compare images (using the vision plugin):

lm vision img1.jpg img2.jpg "Compare the images"

Generate a commit message in a git repository (using the git plugin):

lm commit

Terminal client plugins

Plugins for the terminal client are available: terminal client plugins

Nodejs example: using an agent with local tools

import { Agent } from "@agent-smith/agent";
import { Lm } from "@locallm/api";

let model;
const serverUrl = "http://localhost:8080/v1"; // Local Llama.cpp
const apiKey = "";
const system = "You are a helpful touristic assistant";
const _prompt = `I am landing in Barcelona in one hour: I plan to reach my hotel and then go for outdoor sport. 
How are the conditions in the city?`;

function run_get_current_weather(args) {
    console.log("Running the get_current_weather tool with args", args);
    return '{ "temp": 20.5, "weather": "rain" }'
}

function run_get_current_traffic(args) {
    console.log("Running the get_current_traffic tool with args", args);
    return '{ "trafic": "normal" }'
}

const get_current_weather = {
    "name": "get_current_weather",
    "description": "Get the current weather",
    "arguments": {
        "location": {
            "description": "The city and state, e.g. San Francisco, CA",
            "required": true
        }
    },
    execute: run_get_current_weather
};

const get_current_traffic = {
    "name": "get_current_traffic",
    "description": "Get the current road traffic conditions",
    "arguments": {
        "location": {
            "description": "The city and state, e.g. San Francisco, CA",
            "required": true
        }
    },
    execute: run_get_current_traffic
};

async function main() {
    const lm = new Lm({
        providerType: "openai",
        serverUrl: serverUrl,
        apiKey: apiKey,
        onToken: (t) => process.stdout.write(t),
    });
    const agent = new Agent(lm);
    await agent.run(_prompt,
        //inference params
        {
            model: model ?? "",
            temperature: 0.5,
            top_k: 40,
            top_p: 0.95,
            min_p: 0,
            max_tokens: 4096,
        },
        // query options
        {
            debug: false,
            verbose: true,
            system: system,
            tools: [get_current_weather, get_current_traffic]
        });
}

(async () => {
    await main();
})();

Server api example

To execute a task using the server http api:

import { useServer } from "@agent-smith/apicli";

const api = useServer({
    apiKey: "server_api_key",
    onToken: (t) => {
        // handle the streamed tokens here
        process.stdout.write(t)
    }
});
await api.executeTask(
    "translate", 
    "Which is the largest planet of the solar system?", 
    { lang: "german" }
)

Libraries

The cli is powered by:

• Locallm for the inference api servers management
• Modprompt for the prompt templates management

The server is powered by:

• Echo