BuildSafe Copilot

AI-powered construction compliance with Elastic hybrid search and Vertex AI grounding.

Overview

BuildSafe Copilot is a comprehensive construction compliance assistant that leverages Google Cloud's Vertex AI and Elastic's hybrid search capabilities to provide accurate, grounded answers about construction safety regulations and generate compliance documents.

Architecture

┌─────────────────┐    ┌─────────────────┐    ┌─────────────────┐
│   Next.js       │    │   FastAPI       │    │   Elastic       │
│   Frontend      │◄──►│   Backend       │◄──►│   Cloud         │
│   (Chat UI)     │    │   (Cloud Run)   │    │   (Serverless)  │
└─────────────────┘    └─────────────────┘    └─────────────────┘
                                │
                                ▼
                       ┌─────────────────┐
                       │   Google Cloud  │
                       │   - Vertex AI   │
                       │   - Gemini 1.5  │
                       │   - Embeddings  │
                       │   - GCS Bucket  │
                       └─────────────────┘

Tech Stack

Frontend

• Next.js 14 with App Router
• Tailwind CSS for styling
• shadcn/ui for components
• TypeScript for type safety

Backend

• FastAPI with Python 3.11
• Google Cloud Run for hosting
• Vertex AI for LLM and embeddings
• Elastic Cloud for hybrid search

Data Storage

• Elastic Cloud Serverless for document storage and search
• Google Cloud Storage for generated PDFs and ZIP files

Features

Core Capabilities

• Hybrid Search: Combines BM25 text search with vector similarity search
• Grounded AI: All answers are backed by cited sources from compliance documents
• Document Generation: Creates Job Safety Analysis (JSA) PDFs
• Audit Packs: Generates ZIP files with relevant compliance evidence

Supported Document Types

• Compliance Documents (cc_docs): Safety regulations, standards, and procedures
• Training Materials (cc_training): Educational content and certification requirements
• Incident Reports (cc_incidents): Safety incidents and corrective actions

Quick Start

Prerequisites

• Node.js 18+ and pnpm
• Python 3.11+
• Google Cloud Project with Vertex AI enabled
• Elastic Cloud account

1. Clone and Install

git clone <repository-url>
cd BuildSafeCopilot
pnpm install
cd app/backend && pip install -r requirements.txt

2. Environment Setup

Copy and configure environment files:

cp app/backend/.env.example app/backend/.env
cp app/frontend/.env.example app/frontend/.env.local

3. Elastic Setup

# Create index template and indices
python scripts/elastic/setup_indices.py

# Ingest seed data
python scripts/embed_and_ingest.py --index cc_docs --file scripts/seed/seed_cc_docs.jsonl
python scripts/embed_and_ingest.py --index cc_training --file scripts/seed/seed_cc_training.jsonl
python scripts/embed_and_ingest.py --index cc_incidents --file scripts/seed/seed_cc_incidents.jsonl

4. Run Locally

# Backend (Terminal 1)
cd app/backend
uvicorn main:app --host 0.0.0.0 --port 8080

# Frontend (Terminal 2)
cd app/frontend
pnpm dev

Visit http://localhost:3000 to use the application.

Deployment

Google Cloud Run

# Deploy backend
cd app/backend
gcloud run deploy buildsafe-backend --source . --platform managed --region us-central1

# Deploy frontend
cd app/frontend
gcloud run deploy buildsafe-frontend --source . --platform managed --region us-central1

Terraform Infrastructure

cd infra/terraform
terraform init
terraform plan
terraform apply

API Endpoints

Search & Query

• POST /search - Raw hybrid search results
• POST /ask - AI-powered compliance answers with citations

Document Generation

• POST /action/jsa - Generate Job Safety Analysis PDF
• POST /action/audit-pack - Create compliance audit pack ZIP

Data Schema

Document Structure

{
  "id": "unique-document-id",
  "title": "Document Title",
  "source": "OSHA|CSA|Provincial",
  "doctype": "regulation|standard|procedure",
  "trade": "electrical|plumbing|general",
  "jurisdiction": "SK|ON|NS|Federal",
  "language": "en|fr",
  "published_at": "2024-01-01T00:00:00Z",
  "url": "https://source-url.com",
  "content": "Full document text...",
  "embedding": [0.1, 0.2, ...] // 768-dimensional vector
}

Compliance & Licensing

• License: Apache 2.0
• Data: All seed data is synthetic and created for demonstration purposes
• No Copyrighted Content: No third-party copyrighted materials are included
• Hackathon Compliance: Uses only Google Cloud AI services and Elastic Cloud

Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Add tests if applicable
5. Submit a pull request

Support

For questions or issues, please open a GitHub issue or contact the development team.

---

Built for AI Accelerate Hackathon 2025