Zomato Reviews Scraper

Collect publicly available restaurant details and reviews from Zomato pages.

Input

• restaurantUrls: 1–20 HTTPS restaurant URLs on zomato.com or www.zomato.com
• maxReviewsPerRestaurant: maximum complete unique reviews per restaurant, default 20, range 1–100
• includeSourceEvidence: optional; when enabled, adds a source reference to each review record where available

Example input

{
  "restaurantUrls": [
    "https://www.zomato.com/mumbai/the-big-fish-co-dadar-west/info"
  ],
  "maxReviewsPerRestaurant": 2
}

Output and field semantics

One row per complete public review, with recordType: "zomato_review". Identity, rating, reviewer name, review text, source date, restaurantUrl, resId, source, and scrapedAt are required. Missing optional values are omitted; explicit zero counts and false flags are preserved. Unsupported/incomplete records are not counted toward the accepted total.

• review_text and review_text_short: source strings with HTML tags removed and whitespace collapsed. Short text falls back to full text when absent
• review_text_raw and review_text_short_raw: exact source strings before normalization, when supplied
• date: source timestamp rendered as text, including relative strings such as “2 months ago”; it is not a fabricated ISO publication date
• rating_value: numeric source rating in the range 0–5
• IDs, reviewer profile fields, positive/negative tags, counts, management replies, photos, external links, and display metadata are emitted only where supported by the response. Relative URLs are resolved against Zomato's public origin
• Photo references resolve to actual IMAGES entity objects. Unresolved references are omitted
• scrapedAt: UTC response capture time, not the review publication date
• sourceEvidence: optional {key, sha256, page} identifying the response that supplied the review

Run summary

Run-level status and counts are saved separately from review records in the default key-value store.

Failure and cost bounds

Restaurants are processed one at a time. The review limit caps pagination, which stops earlier when the source has no more complete reviews. Incomplete reviews are omitted; runs report a failure instead of inserting placeholder rows when no complete reviews are available.