AI-Powered Agentic Event Planning with Constraint Satisfaction
A hybrid neuro-symbolic scheduling system that uses backtracking search with AC-3 constraint propagation to dynamically organize event schedules — zero APIs, zero dependencies, runs entirely in the browser.
EventFlow AI is an agentic AI system for event planning that combines:
- Symbolic AI (CSP Solver) — guarantees no hard constraint violations (room conflicts, speaker double-bookings, budget overruns)
- Neural-style Reasoning (Agent Orchestrator) — weighted soft-constraint optimization with chain-of-thought explanations
It handles conferences, hackathons, festivals, workshops, corporate events, and custom formats — all from a single, beautiful neumorphic interface.
┌─────────────────────────────────────────────┐
│ Frontend (Neumorphic UI) │
│ Dashboard │ Wizard │ CRUD Modals │ Chat │
└──────────────────┬──────────────────────────┘
│
┌──────────────────▼──────────────────────────┐
│ Agent Orchestrator │
│ ┌──────────┐ ┌──────────┐ ┌─────────────┐ │
│ │Scheduling│ │ Resource │ │ Attendee │ │
│ │ Planner │ │Allocator │ │ Optimizer │ │
│ └────┬─────┘ └────┬─────┘ └──────┬──────┘ │
└───────┼─────────────┼──────────────┼────────┘
│ │ │
┌───────▼─────────────▼──────────────▼────────┐
│ CSP Solver Engine │
│ Backtracking + AC-3 + MRV + LCV │
│ Hard Constraints │ Soft Optimization │
│ State Manager │ Conflict Resolver │
└─────────────────────────────────────────────┘
No installation. No build step. No backend.
# Clone the repo
git clone https://github.com/ArokyaMatthew/Eventflow.ai.git
# Open in browser
# Just double-click index.html, or:
start index.html # Windows
open index.html # macOS
xdg-open index.html # LinuxDefault login: admin@nexus.ai / nexus2026
Or deploy to GitHub Pages — it works out of the box since it's pure static files.
| Feature | Description |
|---|---|
| CSP Solver | Backtracking with AC-3 arc consistency and MRV heuristic |
| Capacity-Aware LCV | Matches speakers to right-sized venues, minimizing waste |
| Budget Constraint | Hard constraint — total cost never exceeds budget |
| Real-Time Re-optimization | Any mutation triggers full CSP re-solve with diff analysis |
| Chain-of-Thought | Every placement decision is explained with reasoning |
| 3-Agent Pipeline | Scheduling Planner → Resource Allocator → Attendee Optimizer |
- ❌ No room double-booking
- ❌ No speaker time conflicts
- ❌ Budget cap enforcement
- ❌ Speaker availability windows
- ❌ Venue availability windows
- ❌ Venue capacity limits
- ⭐ VIP attendee preference satisfaction
- 📊 Topic diversity per time slot
- 💰 Budget efficiency optimization
- 🎯 Speaker-venue capacity matching
- ⏰ Preferred time slot honoring
- 🎪 6 Event Types — Conference, Hackathon, Festival, Workshop, Corporate, Custom
- 🧙 7-Step Creation Wizard — Guided event setup with live preview
- 📊 Live Dashboard — Schedule grid, metrics, AI insights
- 💬 AI Chatbot — Context-aware assistant for all event queries
- ↩️ Undo System — State snapshots with instant rollback
- 📤 Import/Export — Full JSON serialization
- 🔗 Event Sharing — Join links for collaboration
- 🌙 Dark Mode — Full neumorphic dark theme
- 👥 Team Members — Per-speaker team tracking with contact details
- 📦 Resource Tracking — Chairs, lunch, equipment inventory
eventflow-ai/
├── index.html # Main application shell
├── login.html # Authentication page
├── css/
│ └── styles.css # Complete neumorphic design system
├── js/
│ ├── engine.js # CSP Solver, State Manager, Agent Orchestrator
│ ├── ui.js # Dashboard rendering, entity views, CoT panel
│ ├── wizard.js # 7-step event creation wizard
│ ├── chatbot.js # Context-aware AI chatbot
│ └── app.js # Main controller, CRUD operations, event binding
├── sample_data/
│ └── enterprise_tech_summit.json # Enterprise-scale sample (15 speakers, 6 venues, 30 attendees)
├── implementation_plan.md # Technical architecture document
├── LICENSE # MIT License
└── README.md # This file
Import the included enterprise dataset to see the system at scale:
- Click Import in the sidebar
- Select
sample_data/enterprise_tech_summit.json - Open "Global Tech Summit 2026" → Click Manage
This loads: 15 speakers × 6 venues × 18 time slots (3 days) × 30 attendees with $150K budget.
1. BUILD domains: For each speaker, find all valid (venue, timeSlot) pairs
2. AC-3 PRE-PROCESSING: Prune impossible values via arc consistency
3. BACKTRACK with MRV: Pick the speaker with fewest remaining options
4. CAPACITY-AWARE LCV: Sort venue choices by demand fit, then load balance
5. FORWARD CHECK: After each assignment, prune neighbor domains
6. BUDGET CHECK: Only accept complete assignments within budget
7. DIFF: Compare old vs new schedule, generate change explanations
Contributions welcome! Some ideas:
- Real LLM integration for richer agent reasoning
- Drag-and-drop schedule editing
- PDF/Excel export
- Multi-user real-time collaboration
- Soft constraint relaxation with user approval
- Calendar integration (Google Calendar, Outlook)
MIT — see LICENSE for details.
Built with ❤️ and pure JavaScript — no frameworks, no APIs, no nonsense.