LEAF Web - Climate AI Assistant

LEAF is an advanced AI assistant specialized in climate change and environmental sustainability. Built with Nuxt.js and powered by Google's Gemini AI models, LEAF provides comprehensive, evidence-based responses to climate-related queries while maintaining a conversational experience.

Features

🌱 Climate-Focused AI

• Specialized in climate change, environmental sustainability, and related topics
• Trained on extensive climate datasets for accurate and relevant responses
• Smart query routing to determine the most appropriate response strategy

🔍 Multi-Source Intelligence

• Web Search Integration: Fetches the latest information from trusted climate sources using Tavily API
• Vector Database: Accesses pre-indexed climate documentation for consistent, reliable information
• PDF Processing: Extract and analyze information from uploaded PDF documents

📊 Rich Response Format

• Markdown formatting with headers, lists, and emphasis
• Image integration with strategic placement and descriptive captions
• Hyperlinked sources for easy reference
• Comprehensive citation of information sources

💬 Conversation Memory

• Maintains context across conversation threads
• References previous exchanges for more coherent interactions
• Allows for follow-up questions and deep-dive discussions

Technology Stack

• Frontend: Nuxt.js, Vue.js
• Backend: Server API routes with H3
• Authentication: JWT-based authentication
• Database: Prisma ORM with relational database
• AI: Google Generative AI (Gemini models)
• Vector Database: Qdrant for semantic search capabilities
• Web Search: Tavily API for retrieving up-to-date information
• Document Processing: LangChain for PDF handling and text extraction

System Architecture

LEAF employs a sophisticated multi-step processing pipeline:

1. Query Analysis: Determines if a query is climate-related or conversational
2. Query Routing: Routes to appropriate processing strategy based on query type
3. Research Collection: For research queries, gathers information from:
   • Web search results (prioritizing recent and reliable sources)
   • Vector database of climate knowledge
   • Uploaded PDF documents (when provided)
4. Response Generation: Synthesizes information into comprehensive, conversational responses
5. Source Citation: Automatically formats and includes references to information sources

Getting Started

Prerequisites

• Node.js 16+ and npm/yarn/pnpm
• API keys for:
  • Google Generative AI (Gemini)
  • Tavily (for web search)
  • Qdrant (for vector database)
• PostgreSQL or other database supported by Prisma

Installation

1. Clone the repository:

git clone https://github.com/yourusername/LEAF_WEB_JS.git
cd LEAF_WEB_JS

2. Install dependencies:

# npm
npm install

# pnpm
pnpm install

# yarn
yarn install

3. Set up environment variables in a .env file:

# Database
DATABASE_URL="postgresql://username:password@localhost:5432/leaf_db"

# API Keys
GEMINI_API_KEY="your-gemini-api-key"
TAVILY_API_KEY="your-tavily-api-key"
QDRANT_URL="your-qdrant-instance-url"
QDRANT_API_KEY="your-qdrant-api-key"
FIREBASE_PROJECT_ID="your-firebase-project-id"
FIREBASE_CLIENT_EMAIL="your-firebase-client-email"
FIREBASE_PRIVATE_KEY="your-firebase-private-key"

# Authentication
NUXT_JWT_SECRET="your-jwt-secret"

4. Set up the database:

npx prisma migrate dev

5. Start the development server:

# npm
npm run dev

# pnpm
pnpm dev

# yarn
yarn dev

Deployment

Build the application for production:

# npm
npm run build

# pnpm
pnpm build

# yarn
yarn build

For more information on deploying Nuxt applications, refer to the Nuxt deployment documentation.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

Acknowledgements

• Google for the Generative AI API
• Tavily for the search API
• Qdrant for the vector database
• All the contributors to the climate data sources used in this project