God's Eye Predictor

A web lab that picks six numbers from a pool of 1–42, 1–45, 1–49, 1–55, or 1–58. It uses the correlated-field math from the God's Eye random-number lab and can tilt that field with about 10 years of historical draws from Google Sheets.

This is not a forecast of the next official ticket. Independent draws cannot be predicted. The app samples a history-biased field (FBM value noise, recency, co-occurrence) and shows the six highest spaced peaks.

The original design mockup is kept as reference under Random Number Prediction UI/ and is not the running app.

Run locally

npm install
npm run dev

Open http://localhost:5173/. Restart the dev server after you change .env.

npm run build    # production files in dist/
npm run preview  # serve the build locally

Historical draws (Google Sheets)

Create a .env file in the project root (see .env.example). One URL per game:

VITE_SHEET_42=https://docs.google.com/spreadsheets/d/YOUR_ID/edit?usp=sharing
VITE_SHEET_45=...
VITE_SHEET_49=...
VITE_SHEET_55=...
VITE_SHEET_58=...

Share/edit links are converted to CSV automatically. Each sheet must be Anyone with the link can view. Expected columns (names can vary): a date, and six numbers, often as 28-39-29-06-32-41.

Until URLs are set, the app still runs with a uniform prior (no history tilt).

Do not commit .env. For Netlify, add the same VITE_SHEET_* variables in the site settings, then rebuild. The build writes /_sheets/{42|45|49|55|58} proxy rules into dist/_redirects so the browser does not need CORS on Google.

How to use the UI

Control	Meaning
Game (top of the result card)	Preset 6/42 … 6/58. Sets range 1…N and loads that sheet.
Get today's numbers / New numbers	Draws the set. In Daily mode it turns into Saved for today once used.
Copy	Copies the six numbers to the clipboard.
Daily / Manual	Daily: one set per game per calendar day, saved in this browser until midnight. Manual: draw whenever you want.
Range (Tune the field)	Editable min/max. A custom range drops official-game history.
Detail	FBM octaves (1 = smooth, 4–6 = more bumps).
Hills	FBM scale: how many hills are stretched across the pool.
History weight (α)	Mix of live FBM vs 10-year frequency. 0 = field only; default 0.35.
Seed	Optional. Applied the next time you generate.
Data / header pill	Row count and last fetch. “Cached” means sessionStorage, not a new Google hit.

The six cards are the current picks, sorted ascending, with a peak-strength bar. The bar chart is the live field (gold = picks; muted bars = historical frequency). Orbit view is the same pool as points of light. Recent official draws are the last eight results from the sheet; numbers that also appear in your set are gold.

Design and motion

The look follows godsviewai.com: neutral near-black surfaces, warm off-white ink, 2px corners, gold for highlights, and a labelled icon rail. Tokens live at the top of src/style.css.

• Type: Switzer (Fontshare, ITF Free Font License, free for commercial use) for text, IBM Plex Mono (OFL) for numbers and data only.
• Icons: Lucide, copied from Iconify into the SVG sprite in index.html. Colour and stroke come from CSS.
• Charts: live SVG (src/ui/charts.js). Mark colours are CSS classes (.c-bar, .c-dot, …) driven by the --chart-* tokens.
• Motion: GSAP 3 only (src/ui/motion.js): ScrollSmoother, a pinned hero, SplitText heading reveals, per-section scroll entrances, and the pick-card reveal. With prefers-reduced-motion, smoothing, pinning and movement are off, state changes fade, and the field holds a still frame.
• Hero icon: set LORDICON_SRC in src/config.js to a Lordicon CDN link to play it once on load (the footer adds the required credit). Empty = the built-in SVG eye draws itself.
• Particle reveal: Canvas UI's effect is installed through the shadcn registry (src/components/canvasui/) but off (PARTICLE_REVEAL in src/config.js). It needs Chrome's experimental HTML-in-Canvas API.

Bottom metrics describe the lab field, not winning odds:

• Peak separation — average gap between the six picks.
• Field energy — how peaky the live surface is.
• Drift rate — how fast the FBM walks in time (not draw frequency).
• Calibration — history prior if a sheet loaded, else uniform prior.

Algorithm (short)

For each candidate k in 1…N:

1. Sample 4-octave value-noise FBM across the pool (Lab 2), with slow time drift and a small Ornstein–Uhlenbeck wander (Lab 3).
2. Mix in normalized hit frequency from the sheet (α).
3. Small overdue tilt for numbers not seen recently.
4. After the first peak, remaining scores get a pair boost from historical co-occurrence (Lab 1 idea, discrete).
5. Take six peaks with a spacing filter so they are not all on one hill.

Score: (1 − α) · FBM + α · freq + β · overdue + γ · pairBoost.

Project layout

index.html          # page shell
src/main.js         # boot, generate, animation
src/config.js       # games + sheet URL helper
src/engine/         # FBM, LCG / Box–Muller / OU, peak pick
src/data/sheets.js  # CSV fetch + parse + stats
src/data/daily.js   # daily lock in localStorage
src/ui/             # DOM, SVG charts, GSAP motion
src/components/     # Canvas UI particle reveal (shadcn registry)
Dockerfile          # Cloud Run image (nginx + sheet proxy)
cloudbuild.yaml     # build image and deploy Cloud Run
netlify.toml        # optional Netlify static build

Deploy on Google Cloud (Cloud Run)

The app is a static site plus a same-origin /_sheets/{42|45|49|55|58} proxy so the browser can load Google CSV without CORS. Cloud Run is the intended GCP target for project gods-eye-random.

One-time setup

In Google Cloud Console (project Gods Eye Random / gods-eye-random):

1. Enable Cloud Run, Cloud Build, and Artifact Registry (or Container Registry).
2. Install the Google Cloud CLI and log in:

gcloud auth login
gcloud config set project gods-eye-random

Deploy from this folder

gcloud run deploy gods-eye-predictor \
  --source . \
  --region asia-southeast1 \
  --allow-unauthenticated \
  --port 8080

Then attach the sheet URLs (same values as local .env; do not put them in git):

gcloud run services update gods-eye-predictor \
  --region asia-southeast1 \
  --set-env-vars ^;^VITE_SHEET_42=YOUR_42_URL;VITE_SHEET_45=YOUR_45_URL;VITE_SHEET_49=YOUR_49_URL;VITE_SHEET_55=YOUR_55_URL;VITE_SHEET_58=YOUR_58_URL

On Linux/macOS use commas instead of ^;^ / ;:

gcloud run services update gods-eye-predictor \
  --region asia-southeast1 \
  --set-env-vars VITE_SHEET_42=YOUR_42_URL,VITE_SHEET_45=YOUR_45_URL,VITE_SHEET_49=YOUR_49_URL,VITE_SHEET_55=YOUR_55_URL,VITE_SHEET_58=YOUR_58_URL

Cloud Run prints a https://…run.app URL. History should show history prior after a refresh.

You can also set those variables in Cloud Run → gods-eye-predictor → Edit & deploy new revision → Variables & secrets.

Deploy from GitHub

Connect the Gods-Eye-Random-Numbers repo to Cloud Build with cloudbuild.yaml (region asia-southeast1, service gods-eye-predictor). After the first image deploy, set VITE_SHEET_* on the Cloud Run service as above. Later builds keep those env vars unless you replace them.

Netlify (optional)

Build command npm run build, publish directory dist. Set VITE_SHEET_42 … VITE_SHEET_58 in the host environment so the client and /_sheets/* redirects work in production.