amalshaji/cafeblr/add-to-cafeblr

Add the cafe from the tweet URL in $ARGUMENTS to data/cafes.json and raise a PR.

Requirements: a clone of amalshaji/cafeblr (clone it if not already inside one), bun, and an authenticated gh CLI.

1. Start from latest main

Ids must be computed against the newest data, never a stale checkout:

git fetch origin && git checkout main && git pull

2. Fetch the tweet

Extract the numeric status id from the URL (x.com/<user>/status/<id> or twitter.com/...), then:

bun -e '
import { fetchTweet } from "./scripts/lib/tweet.ts";
const t = await fetchTweet("TWEET_ID");
console.log(JSON.stringify({
  user: t?.user?.screen_name, likes: t?.favorite_count, text: t?.text,
  media: t?.mediaDetails?.map(m => ({ type: m.type, url: m.media_url_https,
    mp4: m.video_info?.variants?.filter(v => v.content_type === "video/mp4")
      .sort((a, b) => (b.bitrate ?? 0) - (a.bitrate ?? 0))[0]?.url })),
}, null, 2));'

3. Guardrails — stop and tell the user if any fail

• Tweet missing or deleted (adapter returns null).
• No photo — every entry needs an image; a text-only tweet cannot be added.

When multiple tweet media items exist, inspect them and prefer a food/drink photo first; if none show food or drink, use the best available tweet image.

• Duplicate — the source URL (or the same cafe) is already in data/cafes.json.
• Sourcing policy — the author must be a real person sharing a find. Reject brand

accounts, the cafe's own account, news aggregators, and marketing handles.

4. Build the entry

• name / area — from the tweet text (📍 lines are common). If the name is not in the

tweet, check the author's own follow-up replies in the thread (browser). If still unknown, ask the user rather than guessing. Area is a neighbourhood ("Indiranagar", "HSR Layout"), not an address.

• knownFor — one line paraphrasing what the post itself praises. Never invent

qualities the tweet doesn't mention.

• image — inspect the tweet media and choose a photo that clearly shows the

food/drink item first. If no media shows food or drink, use the best available tweet photo (or the video's poster). video — best-bitrate mp4 URL if the media is a video, else null.

• mapsUrl — Google Maps link if the tweet/thread contains one (check

entities.urls[].expanded_url), else null.

• googleRating — Google Maps rating for the exact place if verified, else null.
• source — canonical https://x.com/<user>/status/<id>. author — @<user>.
• postedAt — the tweet page datePublished ISO timestamp. Use the browser-rendered X page metadata, not a guessed date.
• likes — favorite_count from the fetch. addedAt — today, YYYY-MM-DD.
• id — max + 1 over the file from step 1; entries stay in id order (schema and rules:

src/lib/schema.ts, README "Add a cafe").

Append to data/cafes.json (2-space indent, key order matching existing entries).

5. Validate

bun run validate
BASE_REF=origin/main bun scripts/validate.ts   # live-checks the new entry

Both must pass before pushing.

Then geocode the new entry for the homepage map (fetches only ids missing from data/geo.json):

bun scripts/geocode.ts

If it can't resolve the entry (opaque Google Maps data=!4m2… URLs need a browser), open the mapsUrl in the browser and copy the @lat,lng from the resolved address bar into data/geo.json under the new id. A cafe missing from geo.json just won't appear on the map — don't block the PR on it; mention it in the PR body instead.

6. Raise the PR

git checkout -b add-cafe-<kebab-name>
git add data/cafes.json data/geo.json
git commit -m "data: add <Name> (<Area>)"
git push -u origin add-cafe-<kebab-name>
gh pr create --title "Add <Name> (<Area>)" --body "..."

PR body: cafe details table (name, area, source link, likes) plus a Missing data section listing anything unknown (usually mapsUrl). If mapsUrl is null, also leave a PR comment listing the missing Maps links so a human can fill them in. No AI/assistant attribution in commits or the PR.

Finish by reporting the PR URL.