ClimaChain

Climate intelligence dashboard for African nations - temperature, precipitation, CO₂ emissions and vegetation (NDVI) signals paired with policy briefs that are computed deterministically from the loaded data, then optionally refined by DeepSeek for tone.

A Neuravox Ltd platform.

Stack

• Framework: Next.js 15 (App Router)
• Hosting: Cloudflare Workers via @opennextjs/cloudflare - single-Worker bundle with nodejs_compat
• UI: Tailwind v4, Radix primitives, shadcn-style components
• Typography: Public Sans across all levels (institutional clarity, high-glare legibility)
• Charts: Recharts
• AI: DeepSeek via the Vercel AI SDK (OpenAI-compatible endpoint at https://api.deepseek.com/v1)
• Data: World Bank Indicators API (live), deterministically reconstructed from CCKP / CMIP6 baselines when the live source is unavailable. NDVI uses regional vegetation baselines and is ready for Copernicus DataSpace integration.

Getting started

npm install
cp .dev.vars.example .dev.vars   # optional: add DEEPSEEK_API_KEY
npm run dev

Visit http://localhost:3000.

The AI briefs work without a key - the route synthesises a 4-5 sentence data-grounded policy memo from the loaded series. Adding a DEEPSEEK_API_KEY only enables a DeepSeek pass to tighten the prose. Get a key at https://platform.deepseek.com.

Deploying to Cloudflare

# build the Worker bundle
npm run preview          # builds + runs locally on Wrangler
# or
npm run deploy           # builds + deploys to Cloudflare

For repo-driven deploys via the Cloudflare dashboard:

• Framework preset: Next.js
• Build command: npx opennextjs-cloudflare build
• Output directory: .open-next
• Compatibility flags: nodejs_compat
• Secrets: npx wrangler secret put DEEPSEEK_API_KEY (optional)

Design system

Implements the Arid Health Management design system ("Geology of Care"): warm light-first "Salt Sand" canvas (#fff8f5), terracotta primary (#98421f), deep-sage secondary (#396660) and ochre/sand tertiary. Modern-corporate with a tactile influence — flat surfaces, hard tonal layers and directional shadows (no ambient blur, no glow), 0.25rem radii, 8px spacing rhythm. See app/globals.css for the full token system.

Project layout

app/
  api/
    ai/insights/      Policy brief synthesis + DeepSeek refinement
    climate/
      temperature/    Annual temperature series
      rainfall/       Annual precipitation series
      co2/            CO₂ emissions (kt)
      ndvi/           Vegetation health
  layout.tsx          Root layout, fonts + theme provider
  page.tsx            Dashboard shell
components/           UI components (shadcn-style)
hooks/                React hooks
lib/
  climate-api.ts      Shared data layer (used by routes + client)
  countries.ts        Country list + climatological baselines
  insight-synthesis.ts  Deterministic policy-brief generator
  utils.ts            cn() helper

Data quality

The World Bank does not currently expose annual surface-temperature observations through its public REST API. ClimaChain reconstructs an internally consistent temperature record by anchoring each country to its long-term mean (CCKP normal, 1991-2020) and modulating with anchor years from the agricultural-land indicator. Every series carries a quality field (live | estimated | unavailable) and a human-readable source so downstream consumers can attribute correctly.

For projections through 2050, ClimaChain uses CMIP6 ensemble warming priors at the country level. Replace with bespoke RCP scenarios when needed.

License

Apache-2.0 — Neuravox Ltd, 2026.