Where Human Meets AI Intelligence
Build once, deploy anywhere. Multi-cloud AI platform with agentic workflows that scale effortlessly.
- Design System: Centralized UI component library with glassmorphism effects, vibrant gradients, and smooth animations
- Enhanced UI/UX: Modern dark theme with indigo-purple-pink color palette across all applications
- AppHeader Component: Consistent navigation header with home button across all internal apps
- Code Quality Fixes: Resolved auth gating bugs, invalid HTML, and duplicate component issues
- Agentic Data Analysis: Multi-source ingestion (CSV/S3/Postgres), automated ETL pipeline (Glue), agentic Q&A with self-healing queries, chart visualizations
- LLMs Fine-Tuning (Tinker API): Run fine-tuning workflows with live logs (generate data → train → sample/predictions)
- Email Notifications (Gmail SMTP): Central opt-in email service (default Off) for key events (Barista order confirmed, LMS course enrolled, Insurance policy created/claim filed)
- Voice Input (Speech-to-Text): Mic input option for AI Chat, Agentic Barista, Insurance Claims, and Agentic LMS
- ML Predictor: Multi-algorithm ML with automatic model selection, dataset management, streaming pipeline
- Evaluation Dashboard: Auth-gated eval summary with metrics, issues, safety, and model usage
- Infrastructure Security: Private subnets with NAT gateways, RDS SSL + storage encryption, S3 AES256, EBS encryption
- Redis Cluster: ElastiCache with at-rest and in-transit encryption
- Observability: X-Ray daemonset, EKS control plane logging
- IAM Enhancements: IRSA support, GuardDuty (optional), secret rotation (optional)
- K8s: Backend IRSA service account, ECR lifecycle policies
- New Apps: Agentic LMS, Agentic Tutor
- Multi-Cloud Support: Added GCP and Azure deployment
- Serverless Execution: AWS Lambda, GCP Cloud Functions, Azure Functions
- 7 AI Models: Across 3 providers (Gemini, Groq, AWS Bedrock)
- Backend: FastAPI + Tortoise ORM + LangGraph 1.0.1 + App Registry System
- Frontend: Next.js 14 (App Router) + Centralized Design System + Inter Font
- Database: PostgreSQL (AWS RDS, GCP Cloud SQL, Azure Flexible Server)
- AI Models: Gemini 2.5 Flash (Lite/Flash/Pro), Groq Compound + Llama 4 Scout, AWS Bedrock Nova (Lite/Pro)
- Code Execution: AWS Lambda, GCP Cloud Functions, Azure Functions
- Infrastructure: Multi-cloud (AWS EKS, GCP GKE, Azure AKS)
| Service | AWS | GCP | Azure |
|---|---|---|---|
| IaC | CloudFormation | Terraform | Terraform |
| Kubernetes | EKS | GKE | AKS |
| Database | RDS PostgreSQL | Cloud SQL | PostgreSQL Flexible Server |
| Container Registry | ECR | Artifact Registry | ACR |
| Serverless | Lambda | Cloud Functions | Azure Functions |
| Storage | S3 | Cloud Storage | Storage Account |
| Secrets | Secrets Manager | Secret Manager | Key Vault |
- 🤖 AI Chat - Switch between 7 AI models across 3 providers
- 📄 Document Analysis - PDF, DOCX, TXT support with text extraction
- 🌐 Web Search - Real-time internet search (Tavily integration)
- ⚡ Code Execution - AI automatically runs Python code when needed
- 💬 Streaming Responses - Real-time AI responses
- ☕ LangGraph Workflow - Multi-agent system with state management
- 🤖 3 Specialized Agents - Menu, Order, and Confirmation agents
- 🧠 AI Reasoning - Intent detection with conversational handling
- 🛒 Cart Management - Add/remove items, view totals
- 📋 Menu Discovery - Browse coffee, pastries, and food items
- ✅ Order Confirmation - Complete orders with database persistence
- 💬 Floating Chat UI - Modal popup interface with agent status display
- 🏥 Role-Based Workflow - Multi-role system (customer, agent, adjuster, manager, admin)
- 📋 Policy Management - Create and manage insurance policies
- 📝 Claim Submission - Submit claims with incident details
- 🔄 Status Workflow - Role-based claim status transitions
- 👥 Adjuster Assignment - Managers assign claims to adjusters
- 💰 Damage Assessment - Track estimated and approved amounts
- 📎 Notes & Documents - Add notes and attachments to claims
- 🔐 Access Control - Role-based permissions and data visibility
- 🎓 AI Course Discovery - Natural language course search
- 📚 Natural Language Enrollment - Conversational enrollment process
- 📊 Progress Tracking - Track learning progress
- 🤖 LangGraph Agents - Multi-agent orchestration for learning
- 🎯 Interactive Learning - AI tutor for Python, AI, Data Science, and more
- 📝 Practice Assessments - Generate quizzes and coding challenges
- 🤖 Multi-Agent System - Tutor, Assessor, Grader, Hint, and Progress agents
- 📊 Progress Tracking - Track scores, strengths, and improvement areas
- 📚 15 Topics - Across 5 categories with beginner to advanced levels
- 🧠 Multi-Algorithm ML - Classification and regression with automatic model selection
- 📈 Dataset Management - Upload files or text, preview samples, and track user datasets
- 🚀 Streaming Pipeline - Live status updates for training, evaluation, and saving
- 🧮 Metrics & Visuals - Key metrics, progress bars, and single-prediction utility
- 🌐 Model Selector - Switch AI models for pipeline guidance
- 🧪 Mini-Apps - Multilingual classification, instruction tuning (SFT), and RL (importance sampling)
- ✅ Dataset Validation - Schema checks for SFT messages[] and RL prompt/required_keys inputs
- 🧠 LoRA Training - Fine-tune small Llama checkpoints with reproducible runs
- 💬 Sampling UI - Ask questions and sample from the latest fine-tuned checkpoint
- 📊 Job Runs - Track status, logs, and outputs per step (generate/train/sample)
- 📊 Multi-Source Ingestion - Upload CSV, connect S3 buckets, or query Postgres databases
- ⚙️ Automated ETL Pipeline - AWS Glue transforms data to Parquet, catalogs in Glue Data Catalog
- 🤖 Agentic Q&A - ReAct agent with tools (get_schema, run_sql, sample_data, create_chart)
- 🔄 Self-Healing Queries - Agent auto-corrects failed SQL using schema context
- 📈 Chart Visualizations - Agent generates bar/line charts for aggregations and trends
- 💡 Suggested Questions - AI-generated question suggestions based on dataset columns
- 📥 Export Results - Download query results as CSV
- 🔐 Secure Authentication - Cookie-based (httpOnly access/refresh), rotation, and RBAC
- ✉️ Email Notifications (Opt-in) - Toggle in landing page user menu; uses Gmail app password (
GMAIL_SMTP_USER,GMAIL_SMTP_APP_PASSWORD) - 💬 Slack Notifications (Opt-in) - Set
SLACK_WEBHOOK_URLand enable the master + per-app toggles in the user menu - 🎙️ Voice Input (STT) - Optional mic input in supported apps (browser speech-to-text)
- 👤 User Profile Header - Reusable AppHeader component with logout functionality
- 📊 Evaluation Dashboard - Auth-gated eval summary with metrics, issues, safety and model usage
- ☁️ Cloud Native - Deployed on AWS EKS with auto-scaling
- 🧩 Modular Architecture - Add new apps in 10 minutes
- 🔄 App Registry System - Auto-discovery and registration of apps
- 🎨 Design System - Centralized UI components with glassmorphism styling, vibrant gradients, and modern animations (Card, Modal, Button, AppCard, AppHeader)
- 🪝 Custom Hooks - useAuth hook for centralized authentication
- 📦 Shared Base Models - Timestamp and soft delete mixins
- AWS CLI configured
- Docker installed
- kubectl installed
- jq installed
- Node.js 20+
- Python 3.11+
See DEPLOYMENT_GUIDE.md for complete deployment instructions. See rds_failover.md for RDS Multi-AZ failover testing. See rds_login.md for RDS login steps from EKS.
# Start local environment
docker-compose up
# Notification delivery runs in a separate worker service.
# Backend has NOTIFICATION_WORKER_ENABLED=false; worker has it true.
# Backend: http://localhost:8000
# Frontend: http://localhost:3000
# API Docs: http://localhost:8000/docsPOST /api/auth/register- Register new userPOST /api/auth/login- Login userPOST /api/auth/forgot-password- Send password reset emailPOST /api/auth/reset-password- Reset password with tokenGET /api/auth/me- Get current user
POST /api/apps/ai-chat/sessions- Create chat sessionGET /api/apps/ai-chat/sessions- List sessionsGET /api/apps/ai-chat/sessions/{id}/messages- Get messagesPOST /api/apps/ai-chat/chat- Send message
POST /api/apps/agentic-barista/chat- Chat with barista agentGET /api/apps/agentic-barista/menu- Get menu itemsGET /api/apps/agentic-barista/orders/{session_id}- Get order history
- Node Group: Auto-scales from 2 to 3 t3.medium instances
- Pods: HPA scales backend/frontend from 1 to 3 replicas at 70% CPU
# Check pod status
kubectl get pods
# View logs
kubectl logs -f deployment/backend
kubectl logs -f deployment/frontend
# Check HPA status
kubectl get hpa- PostgreSQL SSL: For private VPC connections, SSL is disabled via RDS parameter group (
rds.force_ssl=0) - Code Execution: Lambda function uses whitelisted safe modules (math, json, datetime, etc.) with custom
__import__for security
# Delete K8s resources
kubectl delete -f k8s/
# Delete CloudFormation stack
aws cloudformation delete-stack --stack-name co-intelligence --region us-east-1# 1. Scaffold new app
./create_app.sh my-app "My App" "🚀" "#ec4899"
# 2. Add import to backend/main.py
import apps.my_app
# 3. Add to frontend/app/config/apps.ts
{
id: 'my-app',
name: 'My App',
description: ['Feature 1', 'Feature 2', 'Feature 3', 'Feature 4'],
icon: '🚀',
color: '#ec4899',
route: '/apps/my-app',
status: 'active',
requiresAuth: true
}
# Done! App is live.See docs/NEW_APP_TEMPLATE.md for detailed guide.
- ✅ Auto-registered backend routes
- ✅ Database models with timestamps
- ✅ Frontend page with auth
- ✅ Appears on homepage automatically
- ✅ Design System components available (Card, Button, Modal, AppHeader)
MIT