Cafe BLR ☕

Nice cafes in Bengaluru, found by the internet. Every cafe on the site comes from a post on X that people actually loved — each card links back to the original post (the receipts) and to Google Maps directions.

Live site: deployed on Cloudflare Workers from main.

How it works

data/cafes.json is the entire database — a single JSON array.
The Astro site prerenders every cafe at build time; search and area filters run client-side with a few lines of vanilla JS.
CI validates every change to the data; Cloudflare Workers Builds deploys main.

Add a cafe

Suggest one

Open a cafe submission issue and paste the X/Twitter post URL. That's it.

Submissions are labeled cafe-submission. A maintainer or agent will review the post, fetch the media/metadata, and add the cafe if it fits the directory rules.

Process a submission with the skill

Install the add-to-cafeblr skill into your agent (Claude Code, Cursor, Codex, …):

npx skills add amalshaji/cafeblr

Then hand your agent the submitted tweet link:

/add-to-cafeblr https://x.com/someone/status/1234567890

It fetches the tweet, builds the JSON entry with the next id, validates, and raises the PR — flagging anything it couldn't find (usually the Maps link).

Direct PRs

Maintainers can also open a pull request that appends one entry to data/cafes.json:

{
  "id": 14,
  "name": "Cafe Name",
  "area": "Indiranagar",
  "knownFor": "One line on what the internet loves about it",
  "image": "https://pbs.twimg.com/media/….jpg",
  "video": null,
  "mapsUrl": null,
  "source": "https://x.com/someone/status/1234567890",
  "postedAt": "2026-07-06T12:34:56.000Z",
  "author": "@someone",
  "likes": 123,
  "addedAt": "2026-07-06",
  "googleRating": null
}

Rules (CI enforces them):

id — next integer (max + 1), entries stay in id order.
source — the X post the cafe was found in; must be unique across the file.
postedAt — the source post timestamp from the tweet page, as an ISO datetime.
image — the post's photo (pbs.twimg.com URL). If the post has multiple media items, inspect them and prefer a food/drink image first; if none show food or drink, use the best available tweet image. video — optional mp4 URL.
mapsUrl — optional Google Maps link; when null the site generates a directions search from name + area.
googleRating — optional Google Maps rating for the exact place, or null if unknown.
likes — the post's like count when you added it.
The post must exist and the image must load — CI re-fetches both for new entries only, so old entries never break the build if a tweet disappears.

Check locally before pushing:

bun install
bun run validate        # schema + ids + duplicates (+ live checks for new entries in CI)
bun scripts/validate.ts --live   # force live checks on every entry

Develop

bun install
bunx astro dev --background   # dev server at localhost:4321
bun run build                 # static build into dist/

Deploy

Cloudflare Workers with static assets (wrangler.jsonc points at dist/). The repo is connected to Workers Builds — every push to main builds (bun run build) and deploys; PRs get preview URLs.

SEO canonical URLs default to https://cafeblr.com. If the production origin changes, set PUBLIC_SITE_URL in the build environment.

Notes

Photos belong to their posters on X; cards always credit the author and link to the original post. Takedown requests: open an issue.
Data source is X posts fetched via Twitter's public syndication API at curation time — no X API keys needed anywhere.

Cafeblr

README