IShowStream 🎮

AI-Powered Multi-Agent Twitch Chat Analytics Platform

Built for the Google Cloud Run Hackathon 2025 - AI Agents Category

IShowStream helps streamers understand their chat in real-time using a 5-agent AI system powered by Gemini AI and deployed on Google Cloud Run.

🌐 Live Demo

🎥 Demo Video

See IShowStream in action:

🎬 Overview

Shows how IShowStream connects to live Twitch chat
Demonstrates real-time message filtering and emotion detection
Displays how insights are generated by the multi-agent AI system
Walks through the dashboard with trends, engagement, and chat analytics

🤖 Multi-Agent System Architecture

IShowStream uses 5 specialized AI agents that work together to analyze Twitch chat in real-time:

Agent Pipeline:

📨 Message → 🔍 SpamFilter → 🎯 Priority → 💬 Engagement → 📊 Dashboard
             (every message)  (every message) (every message)

📊 Batch    → 📈 TrendAgent → Trends/Memes → 📊 Dashboard
             (every 5 minutes, 50 messages)

⏱️ Periodic → 💡 InsightAgent → AI Insights → 📊 Dashboard
             (every 1 minute, last 60s of messages)

Agent 1: SpamFilterAgent

Role: Spam Detection & Content Filtering
Model: Gemini 2.0 Flash
Capabilities: Spam detection, bot detection, pattern matching
Output: Spam classification with confidence scores

Agent 2: PriorityAgent

Role: Message Prioritization & Ranking
Model: Gemini 2.0 Flash
Capabilities: Importance ranking (1-10), category classification
Depends on: SpamFilterAgent (only processes non-spam)
Output: Priority scores for important messages

Agent 3: EngagementAgent

Role: Engagement Potential Prediction
Model: Gemini 2.0 Flash
Capabilities: Predicts conversation potential (1-10), streamer response recommendations
Output: Engagement scores, should_respond flags

Agent 4: TrendAgent

Role: Trend & Pattern Detection
Model: Gemini 2.0 Flash
Capabilities: Trending topics, meme tracking, spam wave detection, chat mood analysis
Runs: Every 5 minutes on last 50 messages (batch analysis)
Output: Top words, emotes, trending topics, overall mood

Agent 5: InsightAgent

Role: High-Level Chat Analysis & Insights
Model: Gemini 2.0 Flash
Capabilities: Emotion clustering, topic summarization, actionable recommendations
Runs: Every 1 minute on last 60 seconds of messages
Output: AI-generated insights with emotion analysis and streamer recommendations

Multi-Agent Orchestration:

The AgentOrchestrator coordinates all agents:

Per-Message Pipeline: Spam → Priority → Engagement (real-time)
Batch Analysis: TrendAgent analyzes patterns across messages (every 5 min)
Insight Generation: InsightAgent generates high-level insights (every 1 min)
WebSocket Updates: Real-time agent activity broadcast to dashboard

Multi-Agent Features:

✅ Agent specialization (5 distinct roles)
✅ Sequential processing pipeline
✅ Batch processing patterns
✅ Conditional agent invocation
✅ State management across agents
✅ Real-time communication between agents and frontend

🏗️ System Architecture

Pipeline overview

Twitch IRC sends live chat messages to the Chat Ingestion Service
Messages are saved to Firestore
The Multi Agent Orchestrator in Python with Gemini AI processes messages and writes insights and trends to Firestore
The Dashboard API in Go serves REST and WebSocket endpoints
The React Dashboard subscribes to real time updates and renders insights

🚀 Quick Start

Prerequisites:

Node.js 18+ and npm
Go 1.24+
Python 3.12+
Google Cloud account with Firestore enabled
Twitch Developer account

1. Clone the Repository:

git clone https://github.com/JosephDavisC/streamsense.git
cd streamsense

2. Setup Environment Variables:

cp config/.env.example config/.env

Edit config/.env with your credentials:

# Google Cloud
GOOGLE_CLOUD_PROJECT=your-project-id
GOOGLE_API_KEY=your-gemini-api-key

# Twitch
TWITCH_CLIENT_ID=your-client-id
TWITCH_CLIENT_SECRET=your-client-secret
TWITCH_CHANNEL=xqc    # Channel to monitor

3. Authenticate with Google Cloud:

gcloud auth application-default login

4. Install Dependencies:

# Backend agents (Python)
cd backend/agents
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
cd ../..

# Frontend
cd frontend/dashboard
npm install
cd ../..

5. Start All Services:

./start-all.sh

Then open: http://localhost:3000

6. Stop All Services:

./stop-all.sh

📁 Project Structure

streamsense/
├── backend/
│   ├── chat-ingestion/     # Go - Twitch IRC ingestion service
│   │   └── main.go
│   ├── dashboard-api/      # Go - REST API + WebSocket
│   │   └── main.go
│   └── agents/             # Python - Multi-Agent System
│       ├── orchestrator.py        # Agent coordinator
│       ├── spam_filter_agent.py   # Agent 1: Spam detection
│       ├── priority_agent.py      # Agent 2: Priority ranking
│       ├── engagement_agent.py    # Agent 3: Engagement prediction
│       ├── trend_agent.py         # Agent 4: Trend detection
│       ├── insight_agent.py       # Agent 5: Insight generation
│       └── requirements.txt
├── frontend/
│   └── dashboard/          # React - Real-time dashboard UI
│       ├── src/
│       │   ├── App.js
│       │   └── components/
│       │       ├── Dashboard.js
│       │       ├── Landing.js
│       │       ├── Blog.js
│       │       ├── History.js
│       │       └── Profile.js
│       └── package.json
├── config/
│   └── .env                # Environment variables
├── start-all.sh            # Start all services
└── stop-all.sh             # Stop all services

🛠️ Tech Stack

Backend:

Go (Golang) - High-performance service layer (Chat Ingestion, Dashboard API)
Python - AI agents with Gemini
Firestore - Real-time NoSQL database
Twitch IRC - Live chat streaming

AI/ML:

Google Gemini 2.0 Flash - LLM for all agent analysis
Multi-Agent Architecture - 5 specialized agents working in parallel

Frontend:

React 19 - Modern UI framework
WebSocket - Real-time updates
Firebase Auth - User authentication

DevOps:

Google Cloud Run - Serverless container platform
Docker - Containerization
Git - Version control

☁️ Cloud Run Deployment

Deploy all services to Google Cloud Run:

# Set your project ID
export PROJECT_ID=your-project-id

# Deploy Chat Ingestion
cd backend/chat-ingestion
gcloud run deploy chat-ingestion \
  --source . \
  --region us-central1 \
  --allow-unauthenticated \
  --project=$PROJECT_ID

# Deploy Dashboard API
cd ../dashboard-api
gcloud run deploy dashboard-api \
  --source . \
  --region us-central1 \
  --allow-unauthenticated \
  --project=$PROJECT_ID

# Deploy AI Agents
cd ../agents
gcloud run deploy agents \
  --source . \
  --region us-central1 \
  --set-env-vars GOOGLE_API_KEY=your-key \
  --project=$PROJECT_ID

# Deploy Frontend
cd ../../frontend/dashboard
gcloud run deploy ishowstream \
  --source . \
  --region us-central1 \
  --allow-unauthenticated \
  --set-env-vars="REACT_APP_API_URL=https://dashboard-api-xxx.run.app,REACT_APP_WS_URL=wss://dashboard-api-xxx.run.app/ws" \
  --project=$PROJECT_ID

🎨 Features

For Streamers:

✅ Real-time chat monitoring - See all messages as they happen
✅ Automatic spam filtering - AI removes spam/scams/bots
✅ Priority highlighting - Important messages stand out
✅ AI-generated insights - Understand chat emotion and topics every minute
✅ Trend detection - See what's trending in your chat
✅ Pause/Resume monitoring - Save API credits when not streaming
✅ Dynamic channel switching - Change Twitch channel without restart

UI Features:

✅ Real-time dashboard - Live stats and metrics
✅ History view - Review past insights
✅ Dark/Light mode - Customizable theme
✅ WebSocket updates - Instant agent activity notifications
✅ Responsive design - Works on desktop and mobile

🏆 Built For

Google Cloud Run Hackathon 2025

Category: AI Agents
Technologies: Cloud Run, Gemini AI, Firestore, Go, Python, React
Team: JJA (Joseph Davis Chamdani, Juwita Jessica Pangestu, Abraham Guan)

📝 License

MIT License - See LICENSE file for details

Made with ❤️ using Google Cloud Run and Gemini AI

Name		Name	Last commit message	Last commit date
Latest commit History 33 Commits
backend		backend
config		config
frontend/dashboard		frontend/dashboard
logs		logs
.dockerignore		.dockerignore
.gitignore		.gitignore
Diagram.png		Diagram.png
LICENSE		LICENSE
README.md		README.md
clean_firestore.js		clean_firestore.js
docker-compose.yml		docker-compose.yml
start-all.sh		start-all.sh
stop-all.sh		stop-all.sh

License

JosephDavisC/IShowStream

Folders and files

Latest commit

History

Repository files navigation