# Show HN: IP/fal – An IP checker that reimagines your location with AI

> Source: <https://github.com/kiliczsh/ip-fal>
> Published: 2026-09-02 07:49:25+00:00

An open-source, geo-aware IP page powered by Cloudflare Workers and fal.ai. Each visitor sees the public IP used for the current connection, Cloudflare's approximate location metadata, and an AI-generated view informed by local place, season, and time.

Production example: [https://ip.adsparty.com](https://ip.adsparty.com)

- Cloudflare edge geolocation with no separate geolocation database
- Location- and time-aware fal.ai image generation
- Responsive 16:9 desktop and 9:16 mobile images
- Edge caching in 15-minute location/time buckets
- Per-IP generation rate limiting and short-lived signed requests
- Client-side visit history stored only in the visitor's browser
- Privacy mode for screenshots
- Credit-free local previews with five location fixtures

```
.
├── .github/                 GitHub Actions and contribution templates
├── docs/assets/             Documentation images
├── fixtures/locations/      Simulated Cloudflare request profiles
├── scripts/                 Local developer tools
├── src/
│   ├── assets/              Generated static assets
│   ├── page.ts              Server-rendered HTML, CSS, and browser behavior
│   └── worker.ts            Worker routes, security, cache, and fal.ai calls
├── tests/                   Unit tests
├── wrangler.jsonc           Safe default configuration for forks
└── wrangler.production.example.jsonc
```

- Node.js 20 or newer
- A Cloudflare account
- A fal.ai API key for real image generation

```
npm install
cp .dev.vars.example .dev.vars
npm run preview -- istanbul
```

Open [http://127.0.0.1:8787](http://127.0.0.1:8787). Mock preview mode never calls fal.ai and does not
consume credits.

Available fixtures: `istanbul`

, `tokyo`

, `new-york`

, `sydney`

, and `reykjavik`

.

Run a fixture with real fal.ai generation:

```
npm run preview -- tokyo --live
```

This requires `FAL_KEY`

in `.dev.vars`

.

Public configuration lives in `wrangler.jsonc`

:

| Variable | Required | Purpose |
|---|---|---|
`FAL_MODEL_ID` |
No | fal.ai model; defaults to `fal-ai/flux-2/turbo` |
`APP_HOST` |
Production | Restricts generation to the configured hostname |
`WEB_ANALYTICS_TOKEN` |
No | Enables Cloudflare Web Analytics |

Secrets must never be committed:

| Secret | Purpose |
|---|---|
`FAL_KEY` |
Server-side fal.ai credentials |
`GENERATION_SIGNING_KEY` |
Signs short-lived, IP-bound generation tokens |

```
npm run typecheck
npm test
npm run check
```

`npm run check`

runs TypeScript, unit tests, and a Wrangler dry run. The same
command runs in GitHub Actions for pushes and pull requests.

The default configuration deploys to a personal `workers.dev`

hostname:

```
npx wrangler secret put FAL_KEY
npx wrangler secret put GENERATION_SIGNING_KEY
npm run deploy
```

For a custom domain, create an ignored maintainer configuration:

```
cp wrangler.production.example.jsonc wrangler.production.jsonc
```

Update its account ID, Worker name, custom domain, `APP_HOST`

, and rate-limit
namespace, then add secrets and deploy:

```
npx wrangler secret put FAL_KEY --config wrangler.production.jsonc
npx wrangler secret put GENERATION_SIGNING_KEY --config wrangler.production.jsonc
npm run deploy:production
```

`GET /geo-image-data`

accepts only requests carrying a short-lived HMAC token
issued by the page and bound to the connecting IP. Same-origin checks, hostname
restriction, edge caching, and a Cloudflare rate-limit binding reduce blind
hotlinking and unexpected fal.ai spend. This is cost control, not authentication.

Visit history stays in `localStorage`

; IP addresses are deliberately excluded.
Cloudflare provides only the IP family used for the current HTTP connection, so
the other family may display as not detected.

See [SECURITY.md](/kiliczsh/ip-fal/blob/main/SECURITY.md) for responsible disclosure.

Contributions are welcome. Read [CONTRIBUTING.md](/kiliczsh/ip-fal/blob/main/CONTRIBUTING.md) before opening
a pull request.