Skip to content
BranchesTags
 
 

Repository files navigation

🏭 Multi-Agent Industrial Geolocation Engine

Built for the GoogleX Hackathon in collaboration with DeepStation

Live Demo deployed on Cloud Run

An AI-powered multi-agent platform that helps investors and entrepreneurs identify the best industrial parks across India by combining real-time web scraping, Google Maps intelligence, government scheme matching, and Gemini AI deep research — all in a single unified pipeline.

🏗️ Architecture

flowchart LR
    A["👤 User Input<br/>Sector · State ·<br/>Land · Budget"] --> B["Step 2<br/>Query Engine<br/>Filter 4200+ parks"]
    B --> C["🤖 Scraper Agent<br/>Geocode · Logistics<br/>· Gemini Research"]
    C --> D["💾 MongoDB<br/>Store Enriched Data"]
    D --> E["🤖 Ranking Agent<br/>Score (100pts) ·<br/>Deep AI Research"]
    E --> F["🤖 Scheme Agent<br/>Gemini + Google Search<br/>Central + State Schemes"]
    F --> G["📊 Top 10 Results<br/>Score · Breakdown ·<br/>AI Recommendation"]
    G --> H["📄 PDF / Excel Export<br/>+ ROI Calculator"]

    style C fill:#3b82f6,color:#fff
    style E fill:#3b82f6,color:#fff
    style F fill:#3b82f6,color:#fff
    style H fill:#ec4899,color:#fff
Loading

Agent Detail View

flowchart TD
    subgraph SA["🤖 Scraper Agent"]
        SA1["Google Maps Geocoding<br/>Find Place → Place Details"]
        SA2["Google Maps Places<br/>Highway · Railway · Airport · Port"]
        SA3["Gemini AI Research<br/>Water · Raw Materials · Incentives"]
        SA1 --> SA2 --> SA3
    end

    subgraph RA["🤖 Ranking Agent"]
        RA1["Multi-Criteria Scoring<br/>7 categories = 100 pts"]
        RA2["Deep Research via Gemini<br/>Why suitable · Why attractive"]
        RA1 --> RA2
    end

    subgraph SchA["🤖 Scheme Agent"]
        SchA1["Gemini + Google Search Grounding"]
        SchA2["Central & State Schemes"]
        SchA3["Subsidy Stack Calculation"]
        SchA1 --> SchA2 --> SchA3
    end

    subgraph POST["📊 Post-Pipeline"]
        P1["Live AI Recommendations"]
        P2["ROI Calculator"]
        P3["PDF & Excel Export"]
    end

    SA --> RA --> SchA --> POST

    style SA fill:#3b82f620,stroke:#3b82f6
    style RA fill:#3b82f620,stroke:#3b82f6
    style SchA fill:#3b82f620,stroke:#3b82f6
    style POST fill:#ec489920,stroke:#ec4899
Loading

✨ Key Features

Feature Description
Multi-Agent Pipeline 3 specialized AI agents (Scraper, Ranking, Scheme) orchestrated via SSE streaming
4,200+ Industrial Parks Comprehensive dataset covering all Indian states and union territories
Precise Geocoding Google Maps Find Place → Place Details → Geocoding API pipeline
Multi-Criteria Scoring 7-category weighted scoring engine (Sector, Land, Logistics, Water, Incentives, Plug&Play, Raw Materials)
Deep AI Research Per-park Gemini analysis with unique insights and recommendations
Government Schemes Gemini + Google Search grounding for real, active central & state schemes
ROI Calculator AI-powered investment return projections (break-even, NPV, payback)
PDF/Excel Export Professional report generation with Gemini-crafted executive summaries
Live AI Recommendations Async per-card Gemini recommendations loaded after results render
Interactive Maps Color-coded pins (green/yellow/red by score) with Google Maps integration

🛠️ Tech Stack

Layer Technology
Backend Python 3.11+ · Flask · SSE Streaming
AI Engine Google Gemini API (v1beta REST) · google-genai SDK
Maps & Location Google Maps Platform (Places, Geocoding, Distance Matrix, JS API)
Database MongoDB (with in-memory fallback)
Report Generation ReportLab (PDF) · openpyxl (Excel)
Frontend Vanilla HTML/CSS/JS · Google Maps JavaScript API

🚀 Quick Start

Prerequisites

⚠️ Google Maps API requires these APIs enabled in Cloud Console:

  • Places API (New), Geocoding API, Distance Matrix API, Maps JavaScript API

Option A: Run Locally (Python)

# Clone the repository
git clone https://github.com/mandeepsinh-parmar/GoogleX_Hackathon.git
cd GoogleX_Hackathon

# Set up environment
python -m venv .venv
source .venv/bin/activate  # Windows: .venv\Scripts\activate
pip install -r requirements.txt

# Configure keys
cp .env.example .env
# Edit .env with your Google API keys

# Run the app
python app.py

Open http://localhost:5000

Option B: Run Locally (Docker)

docker build -t industrial-finder .
docker run -p 8080:8080 --env-file .env industrial-finder

Open http://localhost:8080

☁️ Deployment (Google Cloud Run)

The application is containerized and ready for serverless deployment on Google Cloud Run.

# 1. Install Google Cloud CLI and authenticate
gcloud auth login
gcloud config set project YOUR_PROJECT_ID

# 2. Deploy directly from source
gcloud run deploy startupadvisor \
  --source . \
  --region asia-south1 \
  --allow-unauthenticated \
  --memory 512Mi \
  --set-env-vars "GOOGLE_API_KEY=YOUR_GEMINI_KEY,GOOGLE_MAPS_API_KEY=YOUR_MAPS_KEY"

For detailed deployment instructions, see DEPLOY.md.

📁 Project Structure

GoogleX_Hackathon/
├── app.py                      # Flask backend — SSE pipeline orchestrator
├── requirements.txt            # Python dependencies
├── .env.example                # Environment variable template
│
├── agents/
│   ├── scraper_agent.py        # 🤖 Agent 1: Geocoding + Logistics + Gemini Research
│   ├── ranking_agent.py        # 🤖 Agent 2: Multi-criteria scoring + Deep Research
│   └── scheme_agent.py         # 🤖 Agent 3: Government scheme matching via Gemini
│
├── tools/
│   ├── location_tools.py       # Park query engine (4200+ parks) + geocoding
│   ├── scheme_tools.py         # Scheme matching + subsidy estimation
│   ├── scoring_tools.py        # Weighted location scoring + state ranking
│   └── export_tools.py         # PDF (ReportLab) + Excel (openpyxl) generation
│
├── db/
│   └── mongo_client.py         # MongoDB client with session management
│
├── data/
│   └── iilb_parks.json         # Dataset: 4,200+ industrial parks across India
│
├── templates/
│   └── index.html              # Frontend: Wizard UI + Google Maps + Results
│
└── docs/
    └── ARCHITECTURE.md         # Detailed architecture documentation

🔌 API Endpoints

Method Endpoint Description
GET / Main application UI
POST /api/find-parks Step 2: Filter parks by sector, state, land
POST /api/run-pipeline Steps 3–7: Full SSE pipeline (scrape → rank → schemes)
GET /api/results/<id> Fetch stored results by session ID
POST /api/ai-recommendation Generate unique AI recommendation for a park
POST /api/roi-calculator AI-powered ROI calculation for a park
POST /api/export/pdf Download professional PDF report
POST /api/export/excel Download Excel data export
POST /api/chat Direct Gemini Q&A
GET /api/health Health check

🔄 Pipeline Flow

User Input → Query 4,200+ Parks → Scraper Agent (Geocode + Logistics + Research)
    → MongoDB Storage → Ranking Agent (Score + Deep Research)
    → Scheme Agent (Central + State Schemes) → Top 10 Results
    → [Async] AI Recommendations → [On-Demand] ROI Calculator → [On-Demand] PDF/Excel Export

Scoring Breakdown (100 points)

Category Max Points How It's Scored
Sector Match 20 Exact match vs. mixed-use
Available Land 20 Meets or exceeds requirement
Logistics 20 Highway + Railway + Airport + Port distances
Water Supply 10 Availability assessment
Incentives 15 Number and relevance of park incentives
Plug & Play 5 Ready-to-move infrastructure
Raw Materials 10 Regional availability

🤝 Team

Built with ❤️ for the GoogleX Hackathon in collaboration with DeepStation.

📄 License

MIT License — see LICENSE for details.

About

No description, website, or topics provided.

Resources

Readme
License
Activity

Stars

0 stars

Watchers

0 watching

Forks

0 forks
Report repository

Releases

Packages

Contributors

Languages