daily-ai-times：基于多智能体集群的 serverless AI 新闻聚合平台

Modern serverless AI-powered news aggregation platform with advanced multi-agent swarm intelligence and consensus-driven content curation.

Live Demo

Production Site: https://www.dailyai.wtf

Architecture

Serverless Backend (GitHub Actions) → Static Frontend (GitHub Pages)

┌─────────────────┐    ┌──────────────────┐    ┌─────────────────┐
│   Data Sources  │    │  GitHub Actions  │    │  GitHub Pages   │
│                 │    │  (Serverless     │    │  (Static        │
│ • RSS Feeds     │───▶│   Backend)       │───▶│   Frontend)     │
│ • News APIs     │    │                  │    │                 │
│ • Research APIs │    │ • Data Collection│    │ • Static Site   │
└─────────────────┘    │ • AI Processing  │    │ • JSON APIs     │
                       │ • API Generation │    │ • Asset Serving │
                       └──────────────────┘    └─────────────────┘

Backend Pipeline (Every 4 Hours)

1. Collection - Intelligent gathering from 30+ configured news sources
2. Bulk Intelligence Swarm - Multi-agent parallel processing with specialized AI models
3. Initial Consensus - Advanced consensus algorithms for content filtering
4. Deep Intelligence Analysis - Sophisticated fact-checking, bias detection, and impact analysis
5. Final Consensus - Weighted combination of initial and deep intelligence results
6. Content Classification - Automatic categorization into headlines, articles, and research papers
7. API Generation - Dynamic creation of optimized JSON endpoints
8. Deployment - Seamless push to GitHub Pages with validation

Frontend Features

• Responsive Design - Newspaper-style layout optimized for all devices
• Real-time Loading - Dynamic content updates from JSON APIs
• Offline Support - Service worker caching for offline access
• Performance Optimized - Minimal JavaScript, fast loading

Sidebar data charts

The sidebar bar charts pull from small, independently-refreshed JSON feeds under src/frontend/api/, each with a static in-code fallback so the page never breaks if a feed is missing:

• AI Data Center Buildout (capex.json, via capex_collector.py, workflow Update Data Center Capex) — annualized data center / AI-infrastructure spend, $B. A hybrid feed with no API key required:
  • Public companies (Amazon, Microsoft, Alphabet, Meta, Oracle) — trailing-12-month capital expenditure fetched live from the keyless SEC EDGAR XBRL API.
  • Private companies (OpenAI, Anthropic) — file nothing with the SEC, so their figures come from src/backend/collectors/capex_curated.json, a committed, source-cited file. These are flagged as estimates (a * on the chart).
  • Note: no filing breaks out data-center-only capex, so public figures are total company capex (overwhelmingly AI/data-center spend for these firms). To show a data-center-only estimate for any company, add a value + source/basis override to that company in capex_curated.json.
• Benchmark Leaderboard (leaderboard.json) and the masthead Ticker (ticker.json) follow the same collector → JSON → fallback pattern.

Benefits

• Zero hosting costs - GitHub Pages + Actions free tier
• Global CDN - Automatic scaling and edge distribution
• Secure & reliable - GitHub's enterprise infrastructure
• Fast deployment - Changes live in minutes
• Built-in monitoring - Workflow status and analytics

Quick Start

For Users

Visit the live site - no installation required.

For Developers

git clone https://github.com/SiddanthEmani/daily-ai-times.git
cd daily-ai-times

# Install Python dependencies
pip install -r src/backend/requirements.txt

# Test the complete pipeline locally
./orchestrator.run

# Or run with Python directly
python src/backend/orchestrator.py

For Contributors

1. Fork the repository
2. Configure GitHub Secrets:
   • GROQ_API_KEY - Required for AI processing
   • GOOGLE_ANALYTICS_ID - Optional analytics
   • ALPHA_VANTAGE_API_KEY - Optional, powers the live masthead ticker (falls back to static placeholder quotes if unset)
   • ARTIFICIAL_ANALYSIS_API_KEY - Optional, powers the live benchmark leaderboard chart via the Artificial Analysis API (falls back to static placeholder figures if unset)
   • No secret is needed for the Data Center Capex chart — it uses the keyless SEC EDGAR API plus a committed curated file (see below).
3. Enable GitHub Pages with "GitHub Actions" source
4. Test locally before pushing changes
5. Push changes - Automatic deployment via workflow

Features

🤖 Advanced AI Processing

• Multi-Agent Swarm Intelligence - Distributed processing with specialized AI models
• Consensus Algorithms - Advanced voting and confidence-based filtering
• Deep Intelligence Analysis - Fact-checking, bias detection, and credibility scoring
• Adaptive Batch Processing - Dynamic optimization for model rate limits

📰 Content Management

• Smart Classification - Automatic sorting into headlines, articles, and research papers
• Multi-Source Integration - Research papers, industry news, open-source updates
• Advanced Deduplication - Intelligent content cleanup and similarity detection
• Quality Gates - Multi-tier confidence scoring and validation

⚡ Performance & Reliability

• Fresh Content - Updates every 4 hours with modern orchestrator
• Robust Error Handling - Built-in timeouts, retries, and graceful degradation
• Performance Monitoring - Detailed pipeline metrics and usage tracking
• Serverless Architecture - Zero infrastructure management with GitHub Actions

Development Commands

Pipeline Configuration

The modern AI pipeline is configured via YAML files in src/shared/config/:

Sources: sources/*.yaml - Individual source configurations for each category
Application: app.yaml - General pipeline parameters and collection limits
AI Swarm: swarm.yaml - Multi-agent configuration, consensus rules, and model settings

AI Swarm Configuration

• Bulk Intelligence Agents - Multiple specialized models for parallel processing
• Deep Intelligence Agents - Advanced analysis models with enhanced capabilities
• Consensus Engine - Voting algorithms and confidence thresholds
• Final Consensus - Weighted combination rules and quality gates
• Rate Limiting - Intelligent model-specific TPM management and batching

Workflow Automation

The GitHub Actions workflow (collect-news.yml) runs every 4 hours with modern serverless architecture:

Automated Processing

1. AI News Processing Pipeline - Single orchestrator handles complete workflow
2. Multi-Agent Swarm Intelligence - Distributed AI processing with consensus algorithms
3. Automatic API Generation - Dynamic creation of frontend-ready JSON endpoints
4. GitHub Pages Deployment - Seamless content delivery with validation
5. Pipeline Monitoring - Detailed metrics, usage tracking, and performance analytics

Manual Triggers

Available via GitHub Actions interface with advanced options:

• Source Selection - Choose specific sources or process all
• Force Refresh - Override caching and force complete refresh
• Debug Mode - Enable detailed deployment structure logging
• Skip Deployment - Run processing without deploying (useful for testing)
• Pipeline Metrics - Real-time processing statistics and agent performance

License

MIT License