PredictPal is a guided time-series forecasting app for hackathon demos and rapid prototyping. It takes users from raw data to trained forecasts, then to shareable notebook-style story posts.

• Upload target and optional driver datasets.
• Configure preprocessing and modeling with a guided 5-step flow.
• Train baseline + multivariate models and generate forecast artifacts.
• Review model quality and forecast outputs in Step 4 visual analysis.
• Build and publish story posts in Step 5 (including anonymous/local demo publishing).
• Browse posts in Explore with search and category filters.

• Frontend: Next.js 16, React 19, Tailwind CSS 4, Zustand, Recharts, Lucide
• Backend: FastAPI, Pandas, scikit-learn, skforecast
• Optional integrations: Supabase and Gemini/OpenAI-style AI helpers

backend/
  app/
    main.py
    api/endpoints.py
    core/
    outputs/
frontend/
  src/
    app/
      page.tsx
      create/page.tsx
      explore/page.tsx
      explore/[storyId]/page.tsx
      about/page.tsx
      login/page.tsx
      terms/page.tsx
    components/
      steps/
      story/
    lib/
      api.ts
      store.ts
      debugStories.ts
      localStories.ts
README.md
supabase_schema.sql

• Node.js 20+
• npm 10+
• Python 3.10 to 3.12

1. Clone and open the project:

git clone <repo-url>
cd LeedsHack2026

2. Create Python environment and install backend dependencies:

python -m venv venv
# Windows
.\venv\Scripts\activate
# macOS/Linux
source venv/bin/activate

pip install -r backend/requirements.txt

3. Create backend env file:

copy backend\.env.example backend\.env

Linux/macOS:

cp backend/.env.example backend/.env

4. Start backend:

cd backend
python -m uvicorn app.main:app --reload --port 8000

5. Start frontend in a second terminal:

cd frontend
npm install
npm run dev

• Frontend: http://localhost:3000
• Backend API: http://localhost:8000/api
• Backend docs: http://localhost:8000/docs
• Health check: http://localhost:8000/health

Backend (backend/.env):

SUPABASE_URL=your-supabase-url
SUPABASE_KEY=your-supabase-anon-key
OPENAI_API_KEY=your-openai-key

Frontend (frontend/.env.local, optional):

NEXT_PUBLIC_API_URL=http://localhost:8000/api

If NEXT_PUBLIC_API_URL is not set, frontend defaults to http://localhost:8000/api.

1. Get Started
2. Process Data
3. Train & Forecast
4. Analysis & Results
5. Publish Story

• Use the bug button on /create to open Debug State.
• Debug State supports step-jump controls.
• Step 5 debug pinning now lives there:
• Pin Step 5 publish as persistent debug sample in Explore
• This flag controls whether published stories are also stored as pinned debug samples.

• Backend stories are in-memory by default.
• The app also keeps local persisted stories in browser storage for demo resilience:
• Anonymous/local published stories can still appear and open in Explore.
• Pinned debug samples can persist in the browser.
• Explore feed merges backend, local, and sample stories with dedupe.

Backend:

python -m uvicorn app.main:app --reload --port 8000
python -m py_compile app/api/endpoints.py app/core/training.py

Frontend:

npm run dev
npm run build
npm run lint
npx tsc --noEmit

• If backend restarts, in-memory backend projects reset.
• Local persisted stories/debug samples in browser can still power Explore demos.
• For clean demo setup, clear browser local storage between runs if needed.