Clear, evidence-backed answers to all your climate change questions.

Earthweinherit is a RAG-powered Q&A application that lets you ask anything about climate change and get responses grounded in scientific reports from IPCC, UNEP, and other authoritative sources — not hallucinations. Every answer is traceable back to its source.

Watch on YouTube →

Features

• RAG-based Q&A — answers are generated from semantically retrieved chunks of real scientific reports, not from model memory alone
• Source citations — every response is annotated with the report name, URL, and source logo so you can verify claims
• Live climate charts — visualizes real CO₂ levels, temperature anomalies, and sea ice data using Recharts
• Persistent chat history — conversations are saved per user (cookie-based, no sign-up required) with a sidebar for previous sessions
• Streaming responses — answers stream in token by token via the Vercel AI SDK
• Curated resources — a handpicked reading list linking to IPCC reports, Our World in Data, NASA, WMO, and more

---

How It Works

User question
     │
     ▼
Embed with text-embedding-3-small (768d)
     │
     ▼
Vector similarity search in Supabase (pgvector)
     │
     ▼
Top-k chunks retrieved (threshold: 0.3, k: 7)
     │
     ▼
GPT-4o-mini generates a grounded answer
with citations streamed back to the client

The knowledge base is built offline by running a PDF ingestion pipeline that loads scientific reports, splits them into 256-token chunks (with 20-token overlap), embeds each chunk, and stores them in Supabase with their source metadata.

---

Tech Stack

Layer	Technology
Framework	Next.js 15 (App Router, Turbopack)
Language	TypeScript
UI	Tailwind CSS, shadcn/ui, Framer Motion
AI	OpenAI GPT-4o-mini + text-embedding-3-small
AI SDK	Vercel AI SDK (streaming)
Database	Supabase (PostgreSQL + pgvector)
Data ingestion	LangChain (PDF loader, token text splitter)
Charts	Recharts
Package manager	pnpm

---

Getting Started

Prerequisites

• Node.js 18+
• pnpm
• A Supabase project with pgvector enabled
• An OpenAI API key

Setup

# Clone the repo
git clone https://github.com/divyamdotfoo/earthweinherit.git
cd earthweinherit

# Install dependencies
pnpm install

# Set up environment variables
cp .env.example .env
# Fill in SUPABASE_URL, SUPABASE_KEY, and OPENAI in .env

# Run the dev server
pnpm dev

Open http://localhost:3000.

Building the Knowledge Base

To ingest a new scientific report (PDF) into the vector store:

pnpm embed:text path/to/report.pdf

To embed images from a report:

pnpm embed:img path/to/report.pdf

To test a query against the vector store directly:

pnpm query "What are the tipping points in the climate system?"

---

Acknowledgements

Knowledge base sourced from:

• IPCC Sixth Assessment Report
• UNEP Strategy 2022–2025
• Other UNEP and WMO publications