Social Media Agent System

An AI-powered agent system for TikTok influencer marketing and trend research. Built with Next.js, Prisma, and Gemini AI.

Overview

This project demonstrates a multi-agent system that routes queries to specialized agents:

• Database Search Agent: Query influencers by tier, genre, location, and price
• Trend Research Agent: Research TikTok trends and content ideas
• YouTube Video Finder Agent (Student Challenge): Find YouTube videos on any topic

Video Walkthrough

Live Walkthrough Video

Prerequisites

• Node.js 20.x or higher
• Yarn or npm
• API Keys (see setup below)

Quick Start

1. Clone and Install

# Install dependencies
yarn install
# or
npm install

2. Environment Setup

Create a .env file in the root directory with the following keys:

# Required: Get your Gemini API key
GEMINI_API_KEY=your_gemini_key_here

# Required: Get your SerpAPI key for trend research
SERP_API_KEY=your_serpapi_key_here

# Database URL (shared for students - read-only access)
DATABASE_URL="postgresql://neondb_owner:npg_Rt2Mena8ZVwA@ep-billowing-shape-a4a0p4zw-pooler.us-east-1.aws.neon.tech/neondb?sslmode=require&channel_binding=require"

3. Get API Keys

Gemini API Key (Required)

1. Go to Google AI Studio
2. Sign in with your Google account
3. Click "Create API Key"
4. Copy the key and paste it in your .env file as GEMINI_API_KEY

Free Tier: 60 requests per minute, sufficient for development and testing

SerpAPI Key (Required for Trend Research)

1. Go to SerpAPI
2. Sign up for a free account
3. Navigate to your Dashboard
4. Copy your API key and paste it in your .env file as SERP_API_KEY

Free Tier: 100 searches per month

4. Database Setup

Option A: Use Shared Database (Recommended for Workshop)

The DATABASE_URL provided above connects to a pre-seeded database with 1000 influencers.

Simply run:

npx prisma generate

This generates the Prisma Client with TypeScript types. No migration or seed needed!

Option B: Create Your Own Database

If you want your own Postgres instance:

1. Create a Postgres Database (options):

   • Neon (Free tier available)
   • Supabase (Free tier available)
   • Local Postgres installation

2. Update DATABASE_URL in .env with your connection string

3. Generate Prisma Client:

   npx prisma generate

4. Run Migrations:

   npx prisma migrate dev

5. Seed the Database (creates 1000 influencers):

   npx prisma db seed

5. Run the Development Server

yarn dev
# or
npm run dev

Open http://localhost:3000 in your browser.

Project Structure

app/
├── agents/
│   ├── executeAgent.ts          # Main router agent
│   ├── databaseSearchAgent.ts   # Database queries
│   ├── trendResearchAgent.ts    # Trend research
│   ├── videoFinderAgent.ts      # Student challenge (to implement)
│   └── agentTypes.ts            # Type definitions
├── libs/
│   ├── gemini.ts                # Gemini AI client
│   └── prisma.ts                # Database client
├── actions.ts                   # Server actions
└── page.tsx                     # Main UI

prisma/
├── schema.prisma                # Database schema
└── seed.ts                      # Seed script

Using the Agents

Example Queries

Database Search:

• "Find micro influencers in beauty"
• "Show me gaming influencers under $500"
• "I need nano tier influencers in Los Angeles"

Trend Research:

• "What are the latest TikTok dance trends?"
• "Trending fitness content ideas"
• "Popular hashtags for beauty campaigns"

Student Challenge: YouTube Video Finder Agent

See TODOS.md for a guided challenge to build a new agent.

What you'll build: An agent that finds YouTube videos on any topic using SerpAPI and Gemini AI.

Database Schema

The database includes:

• 1000 Influencers with metadata (tier, genre, location)
• 5 Tiers: nano, micro, mid, macro, mega
• 10 Genres: pop, hiphop, rock, electronic, country, gaming, beauty, fitness, comedy, tech
• Pricing data for different content types

View Database

# Open Prisma Studio to browse data
npx prisma studio

Update Database Schema

If you make changes to prisma/schema.prisma:

# Generate Prisma client
npx prisma generate

# Create and apply migration (if using your own DB)
npx prisma migrate dev --name your_migration_name

Troubleshooting

"Invalid API Key" errors

• Double-check your GEMINI_API_KEY and SERP_API_KEY in .env
• Ensure there are no quotes around the keys
• Restart the dev server after changing .env

"Can't reach database" errors

• Using shared DB: Check your internet connection
• Using own DB: Verify your DATABASE_URL is correct
• Run npx prisma db push to sync schema

"No results found" errors

• Database Search: Make sure database is seeded (npx prisma db seed)
• Trend Research: Check SERP_API_KEY is valid
• Check console logs for detailed error messages

Prisma Client errors

# Regenerate Prisma client
npx prisma generate

# Reset database (warning: deletes all data)
npx prisma migrate reset

Learn More

Building Effective Agents

Read Anthropic's guide on agent design patterns: Building Effective Agents

Key concepts covered:

• Workflows vs Agents
• When to use agents
• Prompt engineering for agents
• Structured outputs
• Agent orchestration patterns

Technology Stack

• Next.js 16 - React framework
• Prisma - Database ORM
• Gemini AI - Language model
• Zod - Schema validation
• SerpAPI - Search results API

Contributing

This is an educational project. Feel free to:

• Add new agents
• Improve existing agents
• Enhance the UI
• Add more seed data

License

MIT License - feel free to use this for learning and teaching!

---

Want to Go Deeper?

If you enjoyed building this agent system and want to take your AI development skills to the next level, check out Parsity's 30-Day AI Dev Cohort.

What you'll learn:

• RAG (Retrieval-Augmented Generation) agents
• LLM operations and deployment
• Linear algebra fundamentals for ML
• Model fine-tuning techniques

What you'll get:

• Live support and office hours
• Build an amazing portfolio project or startup idea
• Learn from experienced AI engineers
• Join a community of web developers leveling up their AI skills

Perfect for web developers who want to become AI engineers.

---

Built for learning about AI agents