motia-research-assistant

Motia Research Assistant

A multi-agent research system that combines Motia workflow orchestration with Mastra AI agents to deliver comprehensive business research reports in under 60 seconds.

What This Does

Send a research query, get back a professional business report with verified facts and actionable insights. The system uses real web search, AI analysis, fact-checking, and report writing to create executive-ready documents.

Example: Query "Stripe payment infrastructure" → Get back a 15-page report with market analysis, competitive positioning, and strategic recommendations.

Architecture

8-Step Workflow Pipeline

1. API Query Handler - Accepts research requests via REST API
2. Data Collection - Web search using Tavily API for current information
3. Research Analysis - Mastra agent analyzes collected data for insights
4. Summary Generation - Creates executive summaries with business focus
5. Fact Verification - Verifies claims using additional searches and source credibility
6. Report Writing - Generates professional reports with proper structure
7. Results Coordination - Tracks performance and quality metrics
8. Storage & Delivery - Stores results and provides access APIs

Technology Stack

• Motia: Workflow orchestration and event-driven architecture
• Mastra: AI agent framework with tool integration
• OpenAI GPT-4o: Language model for analysis and writing
• Tavily API: Web search for real-time data collection
• TypeScript: Type-safe development
• Zod: Schema validation

Quick Start

Prerequisites

# API Keys needed
OPENAI_API_KEY=your_openai_key
TAVILY_API_KEY=your_tavily_key

Installation

# Initialize project
npx create-motia-app motia-research-assistant
cd motia-research-assistant

# Install dependencies
npm install @mastra/core mastra @ai-sdk/openai tavily uuid @types/uuid

# Copy workflow steps to your steps directory
# (All 8 step files from this repository)

Usage

# Start the workflow
npm run dev

# Make a research request
curl -X POST http://localhost:3000/research/query \
  -H "Content-Type: application/json" \
  -d '{
    "query": "Anthropic business model and competitive analysis",
    "depth": "detailed"
  }'

# Response includes research ID for tracking
{
  "success": true,
  "researchId": "research_abc123",
  "status": "processing",
  "estimatedCompletion": "60 seconds"
}

# Check results
curl http://localhost:3000/api/research/research_abc123/report

Research Depth Options

• basic: Quick overview (800 words, 1 search query, ~30 seconds)
• detailed: Comprehensive analysis (1500 words, 3 search queries, ~45 seconds)
• comprehensive: Deep dive (2500 words, 5 search queries, ~60 seconds)

API Endpoints

Submit Research Query

POST /research/query

Body:

{
  "query": "Research topic or company",
  "depth": "basic|detailed|comprehensive"
}

Get Report

GET /api/research/{researchId}/report

Get Executive Summary

GET /api/research/{researchId}/summary

Get Key Insights

GET /api/research/{researchId}/insights

Response Format

{
  "success": true,
  "researchId": "research_abc123",
  "status": "completed",
  "performance": {
    "processingTime": "47s",
    "targetMet": true
  },
  "quality": {
    "overallScore": 0.87,
    "reliabilityScore": 0.92,
    "verifiedClaims": 18
  },
  "results": {
    "title": "Research Report: Anthropic Business Analysis",
    "executiveSummary": "Anthropic is a leading AI safety company...",
    "keyInsights": [
      {
        "insight": "Focuses on constitutional AI development",
        "confidence": 0.95,
        "source": "verified_sources"
      }
    ],
    "wordCount": 1847,
    "readingTime": "8 minutes"
  },
  "access": {
    "reportUrl": "/api/research/abc123/report",
    "downloadUrl": "/api/research/abc123/download"
  }
}

Quality Metrics

The system tracks several quality indicators:

• Processing Time: Target <60 seconds for all research depths
• Source Reliability: Credibility scoring based on domain authority
• Fact Verification: Claims verified through additional searches
• Report Completeness: Structured sections with proper formatting
• Overall Quality Score: Weighted combination of all metrics

Target thresholds:

• Time: ≤60 seconds
• Quality: ≥0.7 overall score
• Reliability: ≥0.6 verification rate

How It Works

Event Flow

User Request → query.received → data.collected → research.analyzed →
summary.generated → factcheck.completed → report.generated →
results.coordinated → workflow.completed

Data Processing

1. Web Search: Tavily API fetches current articles and reports
2. AI Analysis: Mastra research agent processes raw data for insights
3. Business Focus: Summary agent creates executive-oriented analysis
4. Verification: Fact-check agent validates claims with additional searches
5. Professional Writing: Report writer creates structured business documents
6. Quality Control: Results coordinator tracks metrics and performance

Storage Organization

• completed_research: Full workflow records
• public_reports: Shareable formatted reports
• executive_summaries: Quick access summaries
• key_insights: Structured insights for API consumption
• api_responses: Client response cache (30-day expiry)

Configuration

Search Parameters

Research depth determines search behavior:

const searchConfig = {
  basic: { maxResults: 10, queries: 1 },
  detailed: { maxResults: 15, queries: 3 },
  comprehensive: { maxResults: 20, queries: 5 },
};

Agent Instructions

Each Mastra agent has specialized prompts:

• Research Agent: Extract insights and identify trends
• Summary Agent: Create business-focused summaries
• Fact-Check Agent: Verify claims and assess source credibility
• Report Writer: Generate professional executive reports

This system demonstrates the power of combining Motia's workflow orchestration with Mastra's AI agent capabilities for real-world business applications.