repository·main·Indexed 19 days ago
https://github.com/motiadev/motia-examplesA collection of 60+ practical examples for the Motia backend framework, demonstrating event-driven, real-time, and AI-powered applications. Featured examples include the Room-Renovate AI Home Renovation Planner (v1.0.0), which uses a multi-agent workflow and Google Gemini for design planning and rendering, and a Competitor Price Scraper implementing a RAG system with OpenAI, Anthropic Claude, and Supabase.
This collection contains expert-level implementations of Motia's event-driven architecture. Key examples include:
Motia provides three main categories of AI agent examples to help you progress from simple to complex implementations:
The repository provides several intermediate-level examples for different monitoring domains:
The foundational directory contains intermediate examples for building production-ready applications. These examples are categorized into four main areas:
todo-app, image-resizer, sentimental-analysis).human-in-the-loop).blog-to-tweet, telegram-gmail-automation, currency-rate-monitor).motia-docker, motia-parallel-execution, motia-uptime-monitor).The Motia examples collection is designed to teach the following core architectural concepts:
The motia-examples repository is organized into several functional categories to help you find specific implementation patterns. You can browse examples based on your current development goal:
The research-report.md file provides a curated bibliography of academic papers relevant to building memory-centric AI applications like Mem0. It categorizes research into several key areas:
The Content and Marketing directory contains three primary specialized examples:
This document is a research report generated by a research assistant application. It analyzes various academic papers related to Retrieval-Augmented Generation (RAG) and long-term memory mechanisms for Large Language Models (LLMs), specifically in the context of the Mem0 framework.
Key research areas covered include:
In Motia, you can avoid redundant token verification by using the Middleware Context Passing pattern. Instead of verifying a JWT in every handler, follow this workflow:
userId, role) directly to the request object.declare module 'motia' to extend the ApiRequest interface so that your custom properties (like tokenInfo) are type-safe throughout your application.req object without re-verifying the token.This pattern ensures efficiency, reduces code duplication, and maintains consistency across your API.
// Example of the pattern in a handler
export const handler: Handlers['Profile'] = async (req, { logger }) => {
// Access tokenInfo directly - no re-verification needed!
const { tokenInfo } = req
logger.info('Fetching profile', {
userId: tokenInfo!.userId,
role: tokenInfo!.role
})
return {
status: 200,
body: { userId: tokenInfo!.userId }
}
}Motia achieves parallel execution through an event-driven pub/sub architecture. Instead of manual thread management, you define steps that subscribe to the same topic.
ApiRouteConfig step receives input and emits a specific topic (e.g., ['parallel-processing']).EventConfig where the subscribes array contains that same topic. Because they all listen to the same broadcast, they execute concurrently.Example API Trigger Configuration:
export const config: ApiRouteConfig = {
type: 'api',
method: 'POST',
path: '/process-data',
emits: ['parallel-processing'],
bodySchema: z.object({
text: z.string().min(1, 'Text content is required'),
id: z.string().optional()
})
}Example Parallel Processor Configuration:
export const config: EventConfig = {
type: 'event',
subscribes: ['parallel-processing'], // Same topic = parallel execution
emits: ['processing-result'],
input: z.object({
text: z.string(),
processingId: z.string()
})
}export const config: ApiRouteConfig = {
type: 'api',
method: 'POST',
path: '/process-data',
emits: ['parallel-processing'],
bodySchema: z.object({
text: z.string().min(1, 'Text content is required'),
id: z.string().optional()
})
}
export const config: EventConfig = {
type: 'event',
subscribes: ['parallel-processing'],
emits: ['processing-result'],
input: z.object({
text: z.string(),
processingId: z.string()
})
}