"Climate tools don't have a data problem. They have a decision problem."
ClimateOS is a constraint-aware decision engine built to help individuals move from awareness to prioritized action. Instead of just showing you what you already did wrong (tracking), it simulates future impact and returns a ranked, resource-aware action playbook.
Most climate products fall into two failure modes:
- Carbon Trackers: Dashboards that emphasize past mistakes.
- Generic AI Wrappers: Unranked tips without context or constraints.
ClimateOS answers the only question that matters: Given my life, my budget, and my time — what should I do next?
The defining feature of ClimateOS is its Hybrid Inference Pipeline. Pure LLMs are prone to "carbon hallucinations" (inconsistent math), while pure heuristic systems lack contextual reasoning. We split the labor:
All emissions are computed with fixed factors before the AI ever sees the data.
- Reproducibility: Same inputs always produce the same baseline.
- Explainability: Every number has a traceable, non-stochastic source.
- Math Safety: Gemini never performs arithmetic; it only reasons about the results.
Gemini receives the computed data and performs four specific high-level tasks:
- Ranking: Selects top actions by impact-to-effort ratio.
- Constraint Filtering: Removes options outside the user’s budget or time window.
- Tradeoff Analysis: Surfaces real downsides (e.g., upfront costs vs. long-term savings).
- Rejection Reasoning: Explains why certain typical "green" tips were excluded for your specific case.
The same heuristic functions from the backend run in the browser. Slider changes trigger useMemo recalculations in sub-100ms, allowing for zero-latency exploration.
- The Hero Action: A single, highest-ROI change called out separately. No decision fatigue.
- Constraint-Aware Planning: Inputs for Budget (Low/Med/High) and Time (Low/Med/High) filter the AI's reasoning.
- Real-Time Simulator: Adjust your commute, diet, or energy mix and see your score transition (e.g., 42 → 86) instantly.
- Collective Impact Engine: Scales personal choices to a city-wide frame ("If 10,000 people did this...").
- Shareable Impact Card: Export a high-fidelity PNG of your plan via
html-to-image.
- Frontend: Next.js 15 (App Router), React 19, Framer Motion
- Styling: Tailwind CSS (Modern/Brutalist Aesthetic)
- AI Layer: Google Gemini 2.0 Flash
- Validation: Zod (Strict schema enforcement on both Input and AI Output)
- Persistence: LocalStorage (No database overhead, instant state)
- Export: html-to-image
- Node.js 18+
- A Google AI Studio API Key (for Gemini)
-
Clone the repository:
git clone https://github.com/[your-username]/climateos.git cd climateos -
Install dependencies:
pnpm install
-
Setup environment variables: Create a
.env.localfile in the root:GEMINI_API_KEY=your_api_key_here AUTH0_SECRET=your_secret_here AUTH0_BASE_URL=http://localhost:3000 AUTH0_ISSUER_BASE_URL=your_issuer_base_url_here AUTH0_CLIENT_ID=your_client_id_here AUTH0_CLIENT_SECRET=your_client_secret_here
-
Run the development server:
pnpm run dev
Open http://localhost:3000 to start your plan.
- Persistent Backend: Migration to Postgres/Supabase for long-term action tracking.
- Geo-Specific Factors: Integration with Electricity Maps API for real-time grid intensity.
- Habit Loops: Weekly check-ins tied to your Hero Action.
- Dynamic OG Images: Server-side generation for even better social sharing.
We use Zod to enforce a strict contract between the AI and the UI. If Gemini returns malformed JSON or invalid types, the system automatically catches the error and falls back to a purely deterministic action generator, ensuring the app never breaks for the user.