🔷 PRISM.ai

Product Route Intelligence and Supply Mapping

An AI-powered supply chain digital twin platform built on Google ADK with autonomous multi-agent orchestration, real-time weather & geopolitical intelligence, chaos simulation, and strategic decision-making.

✨ Overview

PRISM.ai is a next-generation supply chain intelligence platform that combines autonomous AI agents, digital twin visualization, and multi-agent orchestration to help businesses proactively identify, assess, and mitigate supply chain risks before they cause damage.

Built on Next.js 16 with Google Gemini 2.5 Flash and the Google Agentic Development Kit (ADK), the platform provides an immersive experience for supply chain professionals to model their networks, simulate disruptions, and receive actionable intelligence — all autonomously and in real time.

Deployed on Google Cloud Run via a multi-stage Dockerized pipeline for production-grade scalability.

🚀 Key Features

🗺️ Interactive Digital Twin & Copilot

Interactive supply chain network visualization using React Flow with drag-and-drop node management (suppliers, factories, warehouses, distributors, retailers)
Real-time connection mapping with transport modes, costs, and risk multipliers
Geographic mapping with Leaflet integration showing live node positions
Live Status Monitoring: Nodes dynamically change state and appearance based on real-time intelligence, risk levels, and active disruptions
CopilotKit Integration: A natural language AI copilot embedded directly in the digital twin — query your supply chain health, ask for risk summaries, and get instant answers conversationally

🤖 Autonomous Multi-Agent AI System (Google ADK)

Thirteen specialized AI agents powered by Google Gemini 2.5 Flash via the Google ADK (@google/adk) framework, operating both on-demand and autonomously in the background:

Agent	Type	Purpose
Intelligence	On-demand	Gathers real-time news, weather, and market intelligence via Tavily web search
Forecast	On-demand	Predictive analytics and trend analysis using historical data with persistent storage
Scenario	On-demand	What-if modeling and disruption simulation generation
Impact	On-demand	Quantitative risk assessment and financial impact scoring
Strategy	On-demand	Mitigation planning and strategic recommendations with execution tracking
Strategy Execution	On-demand	Tracks and monitors the implementation of approved strategies
Orchestrator	On-demand	Coordinates all agents for comprehensive multi-step analysis
Route Optimization	On-demand	Calculates optimal alternate transit paths when nodes fail, using graph traversal
Automated Alerts	Autonomous	Continuously scans global news (Tavily) every 3 minutes, correlating threats to specific supply chain nodes with AI-powered severity classification
Weather Intelligence	Autonomous	Scans all node coordinates and transit route midpoints via OpenWeather API every 3 hours, classifying adverse conditions with AI severity assessment
News Polling	Autonomous	Fetches and deduplicates breaking supply chain news every 2 minutes with URL-level deduplication
News Simulation	On-demand	Generates AI-powered news impact scenarios for simulation exercises
Live Intelligence	On-demand	Real-time intelligence gathering for specific nodes and regions

Agent Infrastructure

Tracing & Observability: Every agent execution is wrapped in withTrace(), logging session ID, duration, success/failure, and token usage to the agent_traces table in Supabase
Audit Logging: All agent actions are recorded in the audit_logs table via a fire-and-forget logAudit() pipeline for full accountability
Multi-Level Deduplication: URL-based, node-based, and content-based deduplication prevents notification spam across all autonomous agents
Deterministic Cooldowns: Both in-memory timers and database-timestamp checks prevent redundant API calls
Intelligent Rate-Limit Fallbacks: If Gemini API quota is exhausted, agents automatically fall back to playbook-based mitigation strategies instead of failing, ensuring 100% operational uptime

🔬 Simulation & Chaos Engine

Monte Carlo risk simulations with multi-variable sensitivity analysis
Disruption scenario modeling with template-based and custom configurations
Professional Template Library: Pre-built disruption templates (port closure, supplier bankruptcy, natural disaster, etc.)
Inject artificial disruptions (latency, price spikes, node failures) and watch the digital twin react in real-time with pulsing red states
Cascading Failure Maps: Visualize how a single node failure propagates through the entire supply chain graph
Strategy Finalization: Approved mitigation strategies are tracked through implementation with execution roadmaps
Forecast Scenarios: AI-generated forward-looking risk projections persisted to Supabase for longitudinal analysis

📰 News Room & Timeline

Chronological Timeline: All intelligence (geopolitical threats, weather alerts, live news) organized in a time-ordered, scrollable feed
Chain Segregation: Alerts are strictly isolated to their affected supply chain — no cross-contamination between chains
Smart Severity Filtering: Defaults to showing only CRITICAL and HIGH impact events; MEDIUM and LOW accessible via dropdown selector
Alert Detail Sheets: Click any timeline event to open a full analysis sheet with source citations, affected nodes, credibility scores, and AI-generated impact summaries
Category Tags: Events are tagged as GEOPOLITICAL, WEATHER, Live News, or Supply Chain Alert for instant visual classification

📊 Dashboard & Analytics

Real-time supply chain health metrics with trend indicators
Risk score visualization with Recharts and D3.js
Notification Feed: Live-streaming threat alerts with read/unread state management and optimistic UI updates
Activity Log: Full audit trail of all agent actions, user operations, and system events fetched from the audit_logs table
Strategy execution tracking with progress monitoring

🧠 Memory & Trend Analysis

Mem0 Integration: Persistent memory layer that remembers past disruptions and intelligence to identify longitudinal patterns
ADK Session Memory: Agent sessions maintain context across interactions for smarter, contextual responses
Delta Risk Calculation: Compares current risk levels against historical baselines to highlight escalating threats
Pattern Recognition: Detects recurring failure points in the supply chain lifecycle

🏗️ Architecture

prism.ai/
├── app/
│   ├── (main)/                         # Authenticated routes
│   │   ├── dashboard/                  # Analytics dashboard + notification feed
│   │   ├── digital-twin/              # Supply chain visualization + Copilot
│   │   │   └── view/[id]/             # Individual twin view (dynamic route)
│   │   ├── news-room/                 # Segregated intelligence timeline
│   │   ├── risk-prediction/           # Risk prediction UI
│   │   ├── simulation/                # Chaos simulation engine
│   │   │   ├── mitigationstrategy/    # Strategy execution dashboard
│   │   │   └── result/                # Simulation results view
│   │   └── profile/                   # User profile management
│   ├── api/
│   │   ├── agent/                     # AI Agent endpoints (13 agents)
│   │   │   ├── automated-alerts/      # Autonomous threat scanner
│   │   │   ├── weather-intelligence/  # Autonomous weather monitor
│   │   │   ├── news-polling/          # Autonomous news fetcher
│   │   │   ├── news-simulation/       # News impact simulator
│   │   │   ├── live-intelligence/     # Real-time intelligence
│   │   │   ├── forecast/              # Forecasting agent
│   │   │   ├── impact/                # Impact assessment agent
│   │   │   ├── info/                  # Intelligence gathering agent
│   │   │   ├── orchestrator/          # Multi-agent orchestrator
│   │   │   ├── route-optimization/    # Route optimization agent
│   │   │   ├── scenario/              # Scenario generation agent
│   │   │   ├── strategy/              # Strategy planning agent
│   │   │   │   └── finalize/          # Strategy finalization
│   │   │   └── strategy-execution/    # Strategy execution tracker
│   │   ├── audit-logs/                # Audit log retrieval API
│   │   ├── risk-prediction/           # Risk prediction proxy
│   │   ├── coordination/             # Agent coordination layer
│   │   ├── copilotkit*/               # CopilotKit chat endpoints (3 variants)
│   │   ├── forecast-scenarios/        # Forecast scenario persistence
│   │   ├── news/                      # News management API
│   │   ├── suggestions/               # AI suggestion engine
│   │   └── strategy/                  # Strategy listing & execution
│   ├── auth/                          # Auth callback handlers
│   └── signin/                        # Authentication UI
├── components/
│   ├── digital-twin/                  # Twin visualization components
│   │   ├── canvas/                    # React Flow canvas components
│   │   ├── display/                   # Node display components
│   │   ├── forms/                     # Node/edge creation forms
│   │   ├── layout/                    # Layout components
│   │   └── utils/                     # Graph utility functions
│   ├── simulation/                    # Simulation UI (16 components)
│   │   ├── simulation-page.tsx        # Main simulation orchestrator
│   │   ├── professional-template-selection.tsx
│   │   ├── enhanced-scenario-configuration-form.tsx
│   │   ├── forecast-scenarios.tsx     # AI forecast visualization
│   │   ├── FinalizeStrategyPanel.tsx  # Strategy approval UI
│   │   └── ImplementationRoadmapPanel.tsx
│   ├── news-room/                     # News Room components
│   │   ├── timeline-event-card.tsx    # Individual event rendering
│   │   ├── alert-details-sheet.tsx    # Full alert detail modal
│   │   └── news-room-header.tsx       # Severity filter controls
│   ├── dashboard/                     # Dashboard widgets
│   │   ├── notification-feed/         # Live notification feed
│   │   └── dashboard-page.tsx         # Main dashboard view
│   ├── risk-prediction/              # Risk prediction components
│   ├── orchestrator/                  # Agent orchestration UI
│   ├── copilot/                       # CopilotKit chat interface
│   ├── strategy-dashboard.tsx         # Strategy management view
│   ├── cascading-failure-map.tsx      # Failure propagation visualization
│   ├── node-impact-grid.tsx           # Node impact assessment grid
│   └── ui/                            # shadcn/ui component library
├── lib/
│   ├── adk/                           # Google ADK integration layer
│   │   ├── core/
│   │   │   ├── trace.ts               # Agent tracing & observability
│   │   │   └── session.ts             # Session state management
│   │   ├── mcp/                       # Model Context Protocol adapters
│   │   └── types.ts                   # ADK type definitions
│   ├── ai-config.ts                   # Centralized AI model & key configuration
│   ├── audit-logger.ts                # Audit logging pipeline (fire-and-forget)
│   ├── supabase/                      # Supabase client setup (anon + service role)
│   ├── clients/                       # External API client wrappers
│   ├── stores/                        # Zustand state management
│   ├── monitoring.ts                  # Sentry + custom observability
│   ├── digitalTwinStore.ts            # Digital twin state store
│   ├── template-selector.ts           # Simulation template engine
│   ├── seed-data.ts                   # Demo data seeding
│   ├── fixed-tavily.ts                # Tavily API wrapper with error handling
│   ├── zod-patch.ts                   # Zod 3.25 ↔ ADK compatibility shim
│   └── validation/                    # Input validation schemas
├── types/
│   └── supply-chain.ts                # Core TypeScript type definitions
├── constants/                         # Application constants
├── hooks/                             # Custom React hooks
├── utils/                             # Shared utility functions
├── scripts/                           # Build & deployment scripts
├── supabase/                          # Supabase migrations
├── public/                            # Static assets
├── Dockerfile                         # Multi-stage Docker build for Cloud Run
├── instrumentation.ts                 # Sentry server-side instrumentation
└── instrumentation-client.ts          # Sentry client-side instrumentation

🛠️ Tech Stack

Category	Technology
Framework	Next.js 16.2.4 (App Router, Turbopack)
Language	TypeScript 5
AI Models	Google Gemini 2.5 Flash (with 1.5 Flash quota fallbacks)
Agent Framework	Google ADK (`@google/adk` v1.2.0) for autonomous agent orchestration
AI Chat Interface	CopilotKit (`@copilotkit/react-core`, `@copilotkit/runtime`)

🗄️ Database Schema

PRISM uses Supabase PostgreSQL with the following key tables:

Table	Purpose
`supply_chains`	User-created supply chain definitions
`nodes`	Supply chain nodes (suppliers, factories, warehouses, etc.)
`edges`	Connections between nodes (routes, transport modes, costs)
`notifications`	All alerts, news, and intelligence events (live news, weather, threats)
`forecasts`	AI-generated forecast data with risk scores, weather, news, and market data
`weather_intelligence`	Live weather data for all node coordinates and transit midpoints
`audit_logs`	Complete audit trail of all agent and user actions
`agent_traces`	Agent execution observability (session, duration, success/failure)
`sessions`	ADK session state persistence for agent memory
`agent_queue`	Agent-to-Agent (A2A) communication queue
`pending_approvals`	Human-in-the-loop strategy approval workflow

All tables enforce Row Level Security (RLS) so users can only access their own data.

⚡ Getting Started

Prerequisites

Node.js ≥ 22
Python ≥ 3.10 (for the ML risk prediction API)
pnpm 11+ (recommended)
Docker (for Cloud Run deployment)
API keys for required services (see Environment Variables)

Installation

# Clone the repository
git clone https://github.com/Prashant-thakur77/prism.ai.git
cd prism.ai

# Install dependencies
pnpm install

# Start the development server (Turbopack)
pnpm dev

ML Risk Prediction API

# Navigate to the ML directory
cd ../Market-Supply

# Install Python dependencies
pip install xgboost scikit-learn pandas numpy fastapi uvicorn python-multipart joblib shap

# Train the model (first time only, ~7s on GPU)
python ml/train.py

# Start the FastAPI server
uvicorn api.main:app --reload --port 8001

The Next.js app at http://localhost:3000 and the ML API at http://localhost:8001 must run simultaneously.

Environment Setup

Create a .env file in the root directory:

# ─── Supabase ────────────────────────────────────────
NEXT_PUBLIC_SUPABASE_URL=your_supabase_url
NEXT_PUBLIC_SUPABASE_ANON_KEY=your_anon_key
SUPABASE_URL=your_supabase_url
SUPABASE_SERVICE_ROLE_KEY=your_service_role_key

# ─── AI Services (Required) ─────────────────────────
GOOGLE_GENERATIVE_AI_API_KEY=your_gemini_api_key
GOOGLE_API_KEY=your_google_api_key

# ─── Multi-Key Quota Management (Optional) ──────────
GOOGLE_API_KEY_ORCHESTRATOR=your_key
GOOGLE_API_KEY_AGENTS=your_key
GOOGLE_API_KEY_DIGITAL_TWIN=your_key
GOOGLE_API_KEY_SUGGESTIONS=your_key

# ─── Intelligence APIs ──────────────────────────────
TAVILY_API_KEY=your_tavily_key
OPENWEATHER_API_KEY=your_openweather_key

# ─── Memory ─────────────────────────────────────────
MEM0_API_KEY=your_mem0_key

# ─── Caching ────────────────────────────────────────
UPSTASH_REDIS_URL=your_redis_url
UPSTASH_REDIS_TOKEN=your_redis_token

# ─── Monitoring (Optional) ──────────────────────────
SENTRY_DSN=your_sentry_dsn
NEXT_PUBLIC_SENTRY_DSN=your_sentry_dsn

# ─── CopilotKit ─────────────────────────────────────
NEXT_PUBLIC_COPILOTKIT_ENABLED=true

# ─── ML Risk Prediction API ─────────────────────────
RISK_MODEL_API_URL=http://localhost:8001

🐳 Deployment (Google Cloud Run)

PRISM uses a multi-stage Docker build optimized for production:

# Build the Docker image
docker build -t prism-ai .

# Run locally
docker run -p 3000:3000 --env-file .env prism-ai

# Deploy to Google Cloud Run
gcloud run deploy prism-ai \
  --source . \
  --region us-central1 \
  --allow-unauthenticated \
  --set-env-vars "NODE_ENV=production"

The Dockerfile uses:

Node 22 Alpine base image for minimal footprint
Multi-stage build (deps → builder → runner) for optimized image size
Standalone output mode from Next.js for minimal production bundle
Non-root user (nextjs:nodejs) for security

📡 API Reference

Agent Endpoints

Method	Endpoint	Description
`GET`	`/api/agent/automated-alerts`	Autonomous threat scan for a specific supply chain
`POST`	`/api/agent/weather-intelligence`	Weather scan across all nodes & transit midpoints
`GET`	`/api/agent/news-polling`	Fetch and deduplicate breaking supply chain news
`POST`	`/api/agent/forecast`	Generate predictive supply chain forecasts
`GET`	`/api/agent/forecast`	Retrieve cached forecasts
`POST`	`/api/agent/orchestrator`	Multi-agent coordinated analysis
`POST`	`/api/agent/impact`	Quantitative risk impact assessment
`POST`	`/api/agent/scenario`	Disruption scenario generation
`POST`	`/api/agent/strategy`	Mitigation strategy planning
`POST`	`/api/agent/strategy/finalize`	Finalize and approve a strategy
`POST`	`/api/agent/strategy-execution`	Track strategy execution progress
`POST`	`/api/agent/route-optimization`	Calculate alternate routes around failed nodes
`POST`	`/api/agent/info`	Intelligence gathering & node analysis
`POST`	`/api/agent/live-intelligence`	Real-time intelligence for specific regions
`POST`	`/api/agent/news-simulation`	Generate news impact simulation

Other Endpoints

Method	Endpoint	Description
`GET`	`/api/audit-logs`	Fetch audit logs for a user
`GET`	`/api/forecast-scenarios`	Retrieve persisted forecast scenarios
`POST`	`/api/suggestions`	AI-generated supply chain suggestions
`POST`	`/api/copilotkit`	CopilotKit chat endpoint
`POST`	`/api/copilotkit-digital-twin`	Digital twin-specific copilot

🔄 Autonomous Agent Lifecycle

┌────────────────────────────────────────────────── ────┐
│                   Frontend (Dashboard)                │
│                                                       │
│  ┌────────── ───┐ ┌──────────┐ ┌───────────────────┐  │
│  │ Notification │ │ News Room│ │  Weather Timeline │  │
│  │    Feed      │ │ Timeline │ │   (Segregated)    │  │
│  └──────┬───────┘ └────┬─────┘ └────────┬──────────┘  │
│         │              │                │             │
└─────────┼──────────────┼────────────────┼─────────────┘
          │              │                │
    Every 30s       Every 2min       Every 3hrs
          │              │                │
          ▼              ▼                ▼
┌──────────────┐ ┌──────────────┐ ┌───────────────── ─┐
│  Automated   │ │ News Polling │ │    Weather        │
│  Alerts Agent│ │    Agent     │ │ Intelligence Agent│
│  (Tavily +   │ │ (Tavily raw  │ │ (OpenWeather +    │
│   Gemini AI) │ │  dedup)      │ │  Gemini AI)       │
└──────┬───────┘ └──────┬───────┘ └────────┬──────────┘
       │                │                  │
       ▼                ▼                  ▼
┌────────────────────────────────────────────────┐
│              Supabase PostgreSQL               │
│  notifications │ weather_intelligence │ audit  │
└────────────────────────────────────────────────┘

✅ Roadmap

Built with ❤️ by the PRISM Team

Name		Name	Last commit message	Last commit date
.cursor/rules		.cursor/rules
.vscode		.vscode
__tests__		__tests__
app		app
components		components
constants		constants
docs		docs
hooks		hooks
lib		lib
public		public
scripts		scripts
types		types
utils/functions		utils/functions
.dockerignore		.dockerignore
.gitignore		.gitignore
.hintrc		.hintrc
Dockerfile		Dockerfile
PRODUCTION_AGENT_README.md		PRODUCTION_AGENT_README.md
README.md		README.md
SIMULATION_REDESIGN_COMPLETE.md		SIMULATION_REDESIGN_COMPLETE.md
components.json		components.json
data-analysis.md		data-analysis.md
instrumentation-client.ts		instrumentation-client.ts
instrumentation.ts		instrumentation.ts
list-supabase-tables.ts		list-supabase-tables.ts
next.config.mjs		next.config.mjs
notepad.txt		notepad.txt
package-lock.json		package-lock.json
package.json		package.json
pnpm-lock.yaml		pnpm-lock.yaml
pnpm-workspace.yaml		pnpm-workspace.yaml
postcss.config.mjs		postcss.config.mjs
proposed_actions.md		proposed_actions.md
sentry.edge.config.ts		sentry.edge.config.ts
sentry.server.config.ts		sentry.server.config.ts
supabase_adk_tables.sql		supabase_adk_tables.sql
supabase_forecasts_table.sql		supabase_forecasts_table.sql
tailwind.config.ts		tailwind.config.ts
test-strategy.cjs		test-strategy.cjs
tsconfig.json		tsconfig.json

Navigation Menu

Repository navigation

About

Repository files navigation

🔷 PRISM.ai

✨ Overview

🚀 Key Features

🗺️ Interactive Digital Twin & Copilot

🤖 Autonomous Multi-Agent AI System (Google ADK)

Agent Infrastructure

🔬 Simulation & Chaos Engine

📰 News Room & Timeline

📊 Dashboard & Analytics

🧠 Memory & Trend Analysis

🏗️ Architecture

🛠️ Tech Stack

🗄️ Database Schema

⚡ Getting Started

Prerequisites

Installation

ML Risk Prediction API

Environment Setup

🐳 Deployment (Google Cloud Run)

📡 API Reference

Agent Endpoints

Other Endpoints

🔄 Autonomous Agent Lifecycle

✅ Roadmap

About

Releases

Packages

Contributors

Languages

Latest commit

History

Folders and files

Repository files navigation

🔷 PRISM.ai

✨ Overview

🚀 Key Features

🗺️ Interactive Digital Twin & Copilot

🤖 Autonomous Multi-Agent AI System (Google ADK)

Agent Infrastructure

🔬 Simulation & Chaos Engine

📰 News Room & Timeline

📊 Dashboard & Analytics

🧠 Memory & Trend Analysis

🏗️ Architecture

🛠️ Tech Stack

🗄️ Database Schema

⚡ Getting Started

Prerequisites

Installation

ML Risk Prediction API

Environment Setup

🐳 Deployment (Google Cloud Run)

📡 API Reference

Agent Endpoints

Other Endpoints

🔄 Autonomous Agent Lifecycle

✅ Roadmap

About

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages