SRE Components Reference

The Smyth Runtime Environment (SRE) provides over 40 production-ready components, building blocks you can combine to create robust AI agents.
Each component encapsulates a reusable function, like calling a language model, fetching web data, or filtering JSON.

How to use components

You can add and connect components visually in Studio or programmatically in the SDK Guide.
Most agents are built by wiring together multiple components in a workflow.

AI & LLM Components

Direct access to cutting-edge AI models and multimodal processing.

Component List:

• GenAILLM — Unified LLM interface (OpenAI, Anthropic, Google, etc.)
• VisionLLM — Image analysis and understanding
• MultimodalLLM — Text, image, audio together
• LLMAssistant — Multi-turn conversational assistant
• HuggingFace — Hugging Face models
• ImageGenerator — Text-to-image generation

Example:

const llm = SDK.GenAILLM({ model: 'gpt-4', prompt: 'Summarize this article' }, agent);
const vision = SDK.VisionLLM({ image: imageBuffer }, agent);

External Integrations

Connect your agents to live web data and APIs.

Component List:

• APICall — HTTP(S) API requests, OAuth support
• WebSearch — Live search via web engines
• WebScrape — Website content extraction
• ZapierAction — Zapier triggers/automations
• MCPClient — Model Context Protocol client (advanced integration)

Example:

const search = SDK.WebSearch({ query: 'SmythOS news', maxResults: 5 }, agent);
const api = SDK.APICall({
  url: 'https://api.example.com/data',
  method: 'GET'
}, agent);

Data Processing Components

Clean, index, filter, and classify your data before using it in workflows.

Component List:

• DataSourceIndexer — Index a batch of documents
• DataSourceLookup — Fast search/lookup from index
• DataSourceCleaner — Data normalization/cleaning
• JSONFilter — Extract/filter from JSON structures
• Classifier — Classify data by custom rules or AI

Example:

const indexer = SDK.DataSourceIndexer({ documents: myDocs }, agent);
const classifier = SDK.Classifier({ classes: ['finance', 'tech', 'health'] }, agent);
classifier.in({ Input: llm.out.Reply });

Logic & Control Flow

Control agent workflow logic: branching, iteration, asynchronicity.

Component List:

• LogicAND, LogicOR, LogicXOR — Boolean logic
• LogicAtLeast, LogicAtMost — Threshold-based logic
• ForEach — Iterate over a list or array
• Async, Await — Asynchronous workflow handling

Example:

const forEach = SDK.ForEach({ items: [1,2,3] }, agent);
// Process each item in the array through a step in your workflow

Storage & File Operations

Store, retrieve, and execute code or files inside your agent’s environment.

Component List:

• FileStore — Read/write files (local/S3)
• Code — Execute code snippets dynamically
• ServerlessCode — Serverless compute (cloud function-style)

Example:

const fileStore = SDK.FileStore({ filename: 'output.txt', content: 'Hello world' }, agent);
const code = SDK.Code({ language: 'python', source: 'print("Hi")' }, agent);

System Integration

Integrate agents with system or plugin-level functionality.

Component List:

• ComputerUse — Automate computer/system tasks
• GPTPlugin — Use OpenAI-compatible plugins
• APIEndpoint — Create custom HTTP endpoints in the agent

Example:

const endpoint = SDK.APIEndpoint({ path: '/report', handler: async (req) => ({ ok: true }) }, agent);

nt Summary Table

A complete overview of all available components, organized by category.

Category	Component Name	Description
AI & LLM	GenAILLM	Unified interface for multiple LLMs
	VisionLLM	Analyze and understand images
	MultimodalLLM	Process text, image, and audio
	LLMAssistant	Conversational AI assistant
	HuggingFace	Hugging Face model APIs
	ImageGenerator	Generate images from text
External Integration	APICall	HTTP requests (OAuth support)
	WebSearch	Real-time web search
	WebScrape	Extract website content
	ZapierAction	Zapier automations
	MCPClient	Model Context Protocol client
Data Processing	DataSourceIndexer	Index documents for lookup
	DataSourceLookup	Retrieve from indexed sources
	DataSourceCleaner	Clean and normalize input
	JSONFilter	Manipulate JSON data
	Classifier	Categorize content into classes
Logic & Control Flow	LogicAND, LogicOR, LogicXOR	Boolean operations
	LogicAtLeast, LogicAtMost	Conditional logic by threshold
	ForEach	Iterate over lists
	Async, Await	Manage async steps
Storage & Files	FileStore	File operations (read, write)
	Code	Execute custom code
	ServerlessCode	Run serverless functions
System Integration	ComputerUse	Automate system-level tasks
	GPTPlugin	OpenAI plugin integration
	APIEndpoint	Custom API endpoints

How to Add and Use Components

You can add components to your agent either:

• Visually: Drag and connect in Studio
• Programmatically: Use the SDK Guide

  import { SDK } from 'smyth-runtime';
  const agent = new SDK.Agent({ name: 'FileBot', model: 'gpt-4' });
  const fileStore = SDK.FileStore({ filename: 'log.txt', content: 'Test' }, agent);

See the Building Agents guide for more advanced patterns.

What's Next?

• Building Agents: Combine these components into agent workflows.
• SDK Guide: Programmatically add and configure components.
• Architecture: See how these components fit into the overall SRE system.