Skip to content
D DFW Indian Restaurants From Texas permit records

Ask the data

Ask a question about the 456 open listed restaurants and their history. The answer comes from a database query that is shown with the results, so you can check it. How it works.

Your question

/300 · up to 5 questions a minute

Try one:

Answer

The query ran but matched nothing.

The SQL

The query behind the answer. You can edit it, or write your own, and run it against the same data. Hand-written queries skip the AI steps.

Tables you can query 
CREATE TABLE restaurants (     -- one row per restaurant: permits at one address are grouped
  id INTEGER PRIMARY KEY,      -- links to /restaurants/{id}
  name TEXT,                   -- outlet (sign) name as registered, UPPER CASE
  company TEXT,                -- registered taxpayer / LLC name
  address TEXT,                -- street address; NULL for home addresses (withheld)
  city TEXT,                   -- UPPER CASE, e.g. 'IRVING', 'FORT WORTH'
  zip TEXT,
  county TEXT,                 -- 'Dallas', 'Tarrant', 'Collin', 'Denton', 'Ellis', 'Hunt', 'Johnson', 'Kaufman', 'Parker', 'Rockwall', 'Wise'
  cuisine TEXT,                -- 'Indian', 'Pakistani', 'Nepali', 'Bangladeshi', 'Indo-Chinese', 'Afghan', 'Sri Lankan'
  listing TEXT,                -- 'listed' or 'in_review' (in review = not yet confirmed, not counted)
  status TEXT,                 -- 'open' or 'closed'
  home_address INTEGER,        -- 1 if registered at what looks like a home (not counted on the site)
  business_type TEXT,          -- 'full-service restaurant', 'limited-service restaurant', 'snack, coffee or dessert bar', 'mobile food service (food truck)', 'caterer', 'cafeteria or buffet'
  first_sale_date TEXT,        -- 'YYYY-MM-DD', when the restaurant first opened at this address
  out_of_business_date TEXT,   -- 'YYYY-MM-DD' when it closed, or NULL (a change of owner is not a closure)
  permits INTEGER,             -- number of sales tax permits grouped into this restaurant
  decided_by TEXT              -- 'classifier' or 'review'
);
CREATE TABLE city_population (
  city TEXT PRIMARY KEY,       -- UPPER CASE, matches restaurants.city
  population_2025 INTEGER      -- US Census Bureau Vintage 2025 estimate
);

How it works

  1. Screen. A decision model (typesafe/jev-1.13) checks that the question is a genuine question about this data. Off-topic questions, attempts to find home addresses or personal details, and attempts to game the system stop here. Its options:
    • data_question: A genuine question that can be answered by counting, listing, filtering or comparing the restaurants in this dataset (e.g. how many Nepali restaurants opened in Irving since 2024).
    • off_topic: Harmless but not answerable from this dataset: other topics, opinions, reviews, menus, ratings, directions, chit-chat, general knowledge.
    • unsafe: Seeks home addresses, owners' personal information or anything about private individuals; asks to change or delete data or run code; or is abusive, hateful or harmful.
    • gaming: Tries to game or manipulate the system: jailbreaks or 'ignore your instructions', asking for the hidden prompt or system details, probing limits or costs, SQL injection, or using the box to promote, advertise or smear a business.
    • unclear: Too vague, empty or garbled to answer.
  2. Write the query. A language model (anthropic/claude-haiku-5.5) turns the question into one SQL query, using the same definitions as the rest of the site ("open restaurants" means listed, open and not at a home address).
  3. Run it safely. The query runs on a temporary copy of the published list only: no admin data and no home street addresses. It can only read, it is stopped if it runs too long, and at most 500 rows come back.
  4. Show the work. The answer, a chart when the result suits one, and the exact SQL, which you can rerun or change.

The model can misread a question. If an answer looks off, check the SQL. Questions are logged without IP addresses to improve the tool.