real-fake-data.com
Go to DashboardGet API key

Changelog

Our product is growing for You

A curated, month-by-month history of what shipped — new generators, capabilities added across the whole catalogue, and the platform milestones that made them possible.

  1. New generator

    Plates from anywhere — one endpoint, 29 countries, one vocabulary

    The 29 per-country plate endpoints now have a blend: /v1/vehicle-registration draws a country per record and returns a plate in that country’s real format — its mask, its national alphabet, its region codes, its reserved series. ?countries=pl,de,it narrows the pool exactly as it does on the other multi-country endpoints. The part that took actual work is type. Every country names its plate kinds the way its own law does, and those names do not line up: Italy’s standard is the United States’ passenger, Hungary’s oldtimer is Ireland’s vintage is Finland’s museum, Lithuania’s ev is Germany’s electric. Filtering on the raw union of all 29 would have produced something that looks like it works while quietly meaning "not the US and not Canada" for standard, and missing five countries for historic. So the cross-country type speaks a shared vocabulary of 17 categories — standard, custom, motorcycle, moped, military, police, diplomatic, government, commercial, taxi, trailer, historic, electric, temporary, export, dealer, other — and every record reports both: the category you asked for and the type its own country actually calls it. Asking for historic reaches all fifteen countries that issue a veteran plate, under five different names. A type also narrows the pool to the countries that issue it, so ?type=taxi draws from the seven that do rather than failing on the other 22 — and pairing it with a countries list that issues nothing of the kind returns a 400 naming the countries that can, instead of quietly substituting another class. Each record also carries its region spelled out where the country has one (Bayern, mazowieckie), which matters most for the US and Canada, whose plates encode no jurisdiction at all. For a country’s full record — era, district, individual part — and its own format, use that country’s endpoint; those are unchanged.

  2. Across generators Enhancement

    A seed you can pin — dictionaries grow without moving your data

    Until now, adding a single surname to a dictionary reshuffled almost everything: names were drawn by position in a list, so one insertion shifted every position after it and ~95% of pinned seeds returned something different. That made enriching the catalogue a breaking change, so in practice it never happened. Generators now draw by scoring each entry against the seed rather than indexing — the entry that scores highest wins, and a newcomer only wins where it genuinely scores highest. Adding 10 surnames to a pool of 480 now moves 2.1% of draws instead of 95%, which is the mathematical floor: new entries have to be reachable by someone, and nothing can do better than their proportional share. Reordering a dictionary file is now completely free. In exchange, this release changes the value every seed returns once — a deliberate, one-time reset so that it never has to happen again for a dictionary addition. Applies to person names across all 28 languages, company names, offerings, junk words, email nouns and domains.

  3. Enhancement

    Polish addresses: a frozen catalogue by default, the live register on request

    A pinned seed on /v1/pl/address now reproduces the same address indefinitely. The endpoint serves a curated catalogue — a frozen extract of the national register with ~9,300 places, ~25,000 streets and ~329,000 addresses — instead of drawing from the live PRG dataset, which shifts whenever GUGiK renames a street or adds a building. Every one of Poland’s 2,479 gminas is represented, so teryt filtering still resolves at voivodeship, powiat and gmina level, and every address is real: a real street, in a real place, with a real postal code. Villages with no street name are in there too, exactly as the register has them. If you need the full breadth or today’s data — an address that resolves on a map right now, or a locality the catalogue does not carry — pass live=true and you get the whole register as before, with the honest caveat that a seed is not guaranteed to reproduce across releases. Existing callers who want the previous behaviour should add live=true.

  4. Milestone

    A VS Code extension — data without leaving the editor

    The third addon, after Playwright and the MCP server, and the first one you see rather than call: przeslijmi.real-fake-data on the Visual Studio Marketplace. Press Ctrl+Shift+P, pick from all 326 generators, and the record lands at your cursor — insertion follows multiple cursors, so three cursors and a batch of three give each its own value. The generator list is read from the API at runtime rather than baked into the extension, so a generator shipped here appears there with no update to install. Presets pin a generator, its parameters, and the one field worth inserting — a preset for de/vehicle-registration with district: M and extract: "value" gives you M AB 1234 rather than the whole five-field record — and living in .vscode/settings.json they travel with the repo and bind to a key. Generator defaults are the same shape applied as policy: any.offering capped to countries: "pl,de,fr" holds on every call, however the generator is reached. The other half is compose: describe the shape your app expects in a .rfd.json file and press Reseed, and the filled dataset opens beside it — edit, reseed, look again. Those files get completion for the whole $-grammar (every generator id, and the actual response fields of whichever one you picked, both read from the API) and validation as you type, so an unknown generator or an undeclared $alias is underlined before you spend a token on it. Because compose bills per draw and nested counts multiply, a $count over 100 is flagged and a reseed estimated above 1,000 draws asks first — and Reseed Preview caps every count at 2, so checking that a shape is right costs a fraction of running it. No account needed: without an API key everything runs on the anonymous lane.

  5. Enhancement

    The multi-country endpoints now span 29 countries, not 27

    /v1/person-name, /v1/company-name and /v1/offering drew from the 27 EU member states only — the United States and Canada were reachable one country at a time, but invisible in the blend that /v1/email had already covered. All four now span the same 29. Nothing new had to be built: the American and Canadian generators, name pools and company data were already shipping; the three aggregate endpoints simply were not wired to them. ?countries=us,ca now narrows the pool the same way ?countries=pl,sk,it always has, and an unqualified call reaches all 29. Canadian records arrive bilingual, as they do on /v1/ca/* — names from a weighted anglophone and French-Canadian mix, legal forms including Ltée and ULC. One thing to know before you upgrade a fixture: /v1/offering is no longer single-currency. American offerings are quoted in USD and Canadian ones in CAD, and prices are not rate-converted, so a batch drawn without countries now mixes EUR, USD and CAD. Every record has always carried its own currency field — read it rather than assuming euros, or pin countries to keep a test in one currency.

  6. Milestone

    For educators — free credits for your students

    Teaching QA, testing, or backend development? We now issue a batch of one-time codes for your course, and the offer has its own card on the pricing page rather than a footnote. You tell us the institution, the course and roughly how many students; we send back a list of codes, one per student. Each student redeems on their own account — there is no shared class login, so their work stays theirs after the course ends, and the credits do not expire once redeemed, including after they graduate. A code works exactly once, so a leaked handout costs one grant and nothing more, and codes themselves lapse a year after issue so an old slide cannot be redeemed indefinitely. A student may redeem more than one, so a short workshop now does not rule out a full-semester grant later. No payment details at any point — not to sign up, not to redeem. This is the same product, not a teaching edition: every generator, every country, edge, extreme, invalid and compose all included, because paying here only ever buys volume.

  7. Milestone New generator

    Canada — eleven generators, bilingual throughout

    Eleven new generators at /v1/ca/*: SIN, Business Number, transit number, bank account, vehicle registration, whole person and company, plus person-name, company-name, email and offering. Canada inverts several of the American expectations, and the API reflects it rather than copying the US shapes across. Where the SSN and EIN carry no checksum at all, the SIN and Business Number are both Luhn-checked — genuinely verifiable, so invalid=true breaks real arithmetic instead of merely violating a range. Where the US has no national account-number standard, Canada has CPA Standard 006 fixing the account number at 7 or 12 digits, making /v1/ca/bank-account the better standardised of the two. The transit number returns both real renderings — XXXXX-YYY as printed on a cheque, 0YYYXXXXX for electronic transfer — whose field order reverses, because a system that accepts one routinely chokes on the other. The Business Number treats the CRA program account correctly: 123456789RC0001 and 123456789RT0001 are the same company filing corporate tax and GST/HST, so the bare nine digits are the default and code treating all fifteen characters as the identity will double-count. Everything is bilingual: names draw from a weighted anglophone and French-Canadian mix (Quebec’s Tremblay and Gagnon descend from a few thousand 17th-century settlers and are concentrated in a way no French list reflects), legal forms include Ltée and the unlimited-liability ULC beside Inc. and Ltd., plates cover all 13 provinces and territories in each jurisdiction’s own format, and QC, Quebec and Québec all resolve to the same place. Offerings are quoted in CAD.

  8. Milestone New generator

    The United States — the first country outside the EU

    Eleven new generators at /v1/us/*: SSN, EIN, ABA routing number, bank account, vehicle registration, whole person and company, plus person-name, company-name, email and offering. Three of them work unlike anything in the European catalogue, and the API says so rather than papering over it. A Danish CPR or Polish PESEL encodes the birth date and sex — an SSN encodes neither, nor (since the SSA moved to randomised assignment in 2011) a state of issue, so /v1/us/ssn offers no sex or age filters at all, and /v1/us/person accepts them but draws the birth date rather than decoding it. /v1/us/company carries only an EIN, because US incorporation is a state matter with no federal registry — there is no American equivalent of the CVR or KRS. And /v1/us/bank-account is honest about a split: the ABA routing number is a real standard with a real 3-7-1 mod-10 checksum (and realBank=true returns a genuine published one from a registry of 40 institutions), while the account number has no national standard whatsoever — length and structure are each bank’s own choice — so it is a plausible digit string of a length that bank really issues, and nothing more. Plates cover all 51 jurisdictions in each state’s own serial format, with the state reported rather than parsed out of the characters (a US plate doesn’t encode it), and Wyoming’s real county codes 1–23 rather than free digits. US offerings are quoted in USD.

  1. Milestone

    Seed a database — a whole dataset in one request

    Every other endpoint returns one field. POST /v1/compose returns your whole database: not a value at a time, but the entire nested shape your app expects — users with their orders, orders with their line items, foreign keys that line up — ready to load into Postgres, a db.json, or a mock server. You send a JSON skeleton and mark each slot with a $-source ($generator.pl.person, $generator.any.uuid, a literal, a nested object); everything else passes through. $count turns any node into an array (a fixed number, or a random [2, 5] that rolls per parent, so 100 orders each get their own 2–5 items). Declaring an entity once under $generators or top-level shared keeps several fields describing the same person — a buyer whose name, PESEL, and email agree — where shared is drawn once for the whole response (the one admin who created every row) and a local entity is drawn fresh per record. Dates relate to each other: $date computes an offset or a random range from an anchor, and $anchor / $parent.createdAt measure a child from its own parent, so every order lands after the customer who placed it and every shipment after its order. The whole dataset is reproducible from one seed, and the output format is pluggable — JSON today, with SQL and CSV designed for. See the “Seed a database” guide under the Data menu for the full grammar; reachable from the /docs playground, the MCP server, and the Playwright addon (compose()).

  2. New generator

    Weighted enum and object pickers — bring your own distribution

    Two new locale-agnostic generators for the categorical fields no registry issues but every dataset needs — status, tier, plan, a skewed cohort. /v1/enum draws a member from a weighted map you supply (choices={"gold":1,"silver":4,"bronze":15} → drawn 1/20, 4/20, 15/20), and /v1/object does the same over whole JSON objects (choices=[{"object":{"tier":"gold"},"weight":3},…]), returning the chosen object verbatim. Weights are relative — they need not sum to 1, so probabilities and integer counts are the same thing — and the response echoes the drawn value’s normalized probability. Pass the choices as URL-encoded JSON (capped at 16 KB). Output is reproducible from the seed, and all three testing modes apply: edge inverts the distribution (the rarest member becomes the most likely, a zero-weight member rises to the top), extreme wraps the value in a hostile encoding, and invalid returns a value that is not in your set. Live in the /docs playground and the MCP server.

  3. New generator

    Product & service offerings for all 27 EU countries

    A whole new data type: the offering — a realistic product or service a business sells, with a name, a price, and the industry it belongs to. Every EU country gets a /v1/<cc>/offering endpoint, and /v1/offering draws across the union — pass countries=pl,de,it to mix, omit for all 27, and each record reports its country. The name is localised into that country’s language (multilingual states blend theirs — Belgium across Dutch/French/German, Luxembourg across French/German/Luxembourgish, Finland Finnish/Swedish, Ireland English/Irish, Malta Maltese/English), and the industry label comes back localised too. Draw the whole catalogue or narrow it: pin an industry by NACE code, restrict to type=product or type=service, or full-text filter by offeringName or industryName (?offeringName=fibre). Prices are the real touch — each is a single plausible amount snapped onto a sensible grid for its size (a €25 haircut, a €1,499.99 fibre install, a €209,500 construction job), never raw machine noise. As everywhere, output is reproducible from the seed, and edge (the price/format corners), extreme (a hostile-encoded name), and invalid (a deliberately broken record) all apply. Live in the /docs playground and the MCP server.

  4. New generator

    Synthetic ID generators — UUID, ULID, Nano ID, ObjectId, and sequences

    Five new locale-agnostic generators for the technical identifiers every record needs but no national registry issues — the primary keys. /v1/uuid produces a UUID, random (version=4, the default) or time-ordered (version=7, RFC 9562); /v1/ulid a 26-character Crockford-Base32 ULID; /v1/nanoid a compact URL-safe Nano ID, with size and a custom alphabet; /v1/object-id a well-formed 24-character MongoDB ObjectId; and /v1/sequence an auto-increment integer — the classic SQL primary key — where a single call returns start (default 1) and a batch (count) returns the run start, start + step, …. Like everything else they are deterministic from the seed, so a seeded UUID is reproducible — better than a real random one for fixtures. Any embedded timestamp (UUID v7, ULID, ObjectId) is derived from the seed, not the wall clock, keeping output stable. They fill the gap when you seed a whole record and need an id column to sit beside the name, address, and national number.

  5. New generator

    IBANs for all 27 EU countries

    The IBAN generator went pan-European: every EU country now has its own /v1/<cc>/iban endpoint producing a bank account number in that country’s exact ISO 13616 format — Germany’s 22-character DE…, France’s 27-character FR…, Italy’s IT… with its leading CIN letter, Spain’s ES… with the two dígitos de control, and so on. Every IBAN carries correct mod-97 check digits and, where the national account number embeds its own internal check digit, that too is computed the real way (Belgium’s mod-97 suffix, France’s clé RIB, Italy’s CIN, Spain’s DC, Portugal’s and Slovenia’s ISO 7064 checks, Finland’s Luhn, Estonia’s 7-3-1, Hungary’s two check digits, Croatia’s per-field MOD 11-10, and the Czech/Slovak self-checking account) — so the numbers pass strict per-country validators, not just the outer checksum. The embedded bank identifier is drawn from a real registry per country (Germany’s Bankleitzahl, France’s code banque, the Netherlands’ four-letter BIC prefix, …), so you can pin the issuing bank by bankCode or bankName, or let it be chosen at random. As before, invalid=true deliberately breaks the checksum for testing validators, and edge and extreme both apply.

  6. New generator

    Vehicle registration plates for all 27 EU countries

    The vehicle-registration generator went pan-European: every EU country now has its own /v1/<cc>/vehicle-registration endpoint producing realistic plates in that country’s exact current format — Germany’s district-coded M-AB 1234, France’s AB-965-KL, Italy’s AK 514 RH, Ireland’s year-and-county 262-D-12345, the Netherlands’ sidecode PDA-01-D, and so on. Where a country encodes a region in the plate, the code is drawn from the real registry (Germany’s Unterscheidungszeichen, Slovakia’s district codes, Romania’s 42 counties, Ireland’s counties, Poland’s wyróżniki, …), so the regional part is genuine and filterable. Each country offers the plate classes whose format actually differs — standard plus, where they exist, custom/vanity, military, diplomatic, police, temporary/dealer/export, motorcycle, historic, taxi, and electric — and, where the old format still on the road looks different, an era switch (current / legacy / both) that reaches back to it (Spain’s province plates, France’s FNI, Hungary’s 3-letter series, Ireland’s two-digit-year plates, …). Bulgaria and Greece can render in their native Cyrillic/Greek glyphs on request. Poland additionally gained motorcycle and reduced-size (US/JP-import) variants. Plates carry no checksum, so there is no “invalid” mode — but edge and extreme both apply.

  7. Across generators Enhancement

    New extreme trigger — correct data in a deliberately hostile encoding

    Every generator that already took edge now also accepts extreme=true: it returns correct data presented in the most disturbing way a real system ever sees it. The value is wrapped with one class per record — untrimmed whitespace (including non-breaking spaces and a trailing newline), invisible and zero-width characters with a leading byte-order mark, homoglyph letters (Cyrillic/Greek/fullwidth lookalikes that read as Latin), or a right-to-left override / stacked combining marks — so a batch (count) rotates across all four. It’s built to answer “how should my system behave here?”: does your pipeline trim, normalise (NFKC), and compare values safely, or does a zero-width space break an equality check and a homoglyph slip past a filter? Only the human-facing string is mangled and it stays recoverable: identifiers (PESEL, NIP, IBAN, CPR, codice fiscale, BSN, …) keep their digits ASCII — homoglyphs are excluded — so they still parse and checksum after normalisation; names, companies, emails, and whole people/companies mangle the name (or assembled value) only, leaving initials, legalForm, birth dates, and national identifiers clean. The lorem and custom-regex endpoints don’t take it (it would break their length and pattern-match contracts). Output stays reproducible from the seed; the flag is live in the /docs playground, the MCP server, and the Playwright addon.

    Read the full story →
  1. New generator Enhancement

    Localised email generators for all 27 EU countries

    Email generation goes country-aware: every member state gets a /v1/<country>/email endpoint, and /v1/email now draws across the EU — pass countries=pl,de,it to mix, omit for all 27, and each record reports its country. Names come from each country’s real first-name and surname banks, romanised to ASCII for non-Latin scripts so Bulgarian and Greek names become deliverable local parts. A new corporate domain category builds the address on a company-derived domain (anna.schmidt@mueller-bau.de), drawing the brand from that country’s company-name generator and the suffix from its ccTLDs; two new local-part patterns weave a company token in (first.company, company.first), alongside regional providers (wp.pl, web.de, …) and global webmail. Output stays reproducible from the seed; the new endpoints are live in the /docs playground and the MCP server.

  2. New generator

    Full-company generators for all 27 EU countries

    The one-call “whole company” idea behind /v1/pl/company now spans the EU: every member state gets a /v1/<country>/company endpoint that returns a consistent synthetic company — a full trading name with its legal form, plus the matching national identifiers — in a single request. Each bundles that country’s real registers: Denmark’s CVR, France’s SIREN, Sweden’s organisationsnummer, Italy’s partita IVA, the Netherlands’ KvK + RSIN + btw-id, Germany’s Handelsregister + USt-IdNr + Wirtschafts-IdNr, Austria’s Firmenbuchnummer + UID + Steuernummer, and so on — one to three numbers depending on what the country issues. The name and identifiers are drawn from one seeded sequence, so a given seed always yields the same company; strategy and legalForm shape the name exactly as on /v1/<country>/company-name, invalid=true corrupts every checksummed identifier while leaving the name (and any checksum-less number) intact, and edge=true sweeps the rare corners. Each costs the same as any other single-record call, regardless of the sub-draws inside. The new endpoints are live in the /docs playground and the MCP server.

  3. New generator

    Custom-regex generator — random data matching your own pattern

    A new /v1/custom-regex endpoint inverts a regular expression you supply into a random string that matches it — for seeding data whose format the catalogue doesn’t model: in-house serial numbers, SKUs, ticket ids, order references. Pass pattern (the regex in source form, e.g. [A-Z]{2}-\d{6}) and get back a conforming value; maxRepetition caps how far unbounded quantifiers (*, +, {n,}) expand. Patterns are validated up front — back-references and look-around assertions are rejected (they can’t be honoured when generating a match), as is any pattern whose worst-case expansion is too large. Reserved characters must be percent-encoded in the query (+ as %2B). Output stays reproducible from the seed. Requires a free account — so a submitted pattern is attributable — but never payment.

  4. New generator

    Full-person generators for all 27 EU countries

    The one-call “whole person” idea behind /v1/pl/person now spans the EU: every member state gets a /v1/<country>/person endpoint that returns a mutually consistent synthetic person — given name(s), surname, initials, birth date, and the matching national number — in a single request, so the name, sex, and identity always agree. Each is wired to that country’s real identifier: Denmark’s CPR, France’s NIR, Italy’s Codice Fiscale, Sweden’s personnummer, Germany’s Steuer-ID, Spain’s DNI, the Netherlands’ BSN, and so on. Where the number encodes a birth date and sex (CPR, EGN, isikukood, CNP, EMŠO, …) it is minted first and those facts flow into the name — so a female record always carries the matching parity and birth date. Italy’s Codice Fiscale, which encodes the name itself, is built the other way round: the name is generated first and the code derived from it, letters and all. Where the number is non-semantic (BSN, DNI, Steuer-ID, PPSN, TIC, Maltese ID, Cartão de Cidadão, personas kods) the birth date is drawn independently within the same age/birth-date filters and sex shapes the name. Every endpoint takes the familiar sex and age/birth-date filters, invalid=true (a deliberately wrong checksum, name and date intact), edge=true, and caseStrict=false to mangle the name casing for testing; each costs the same as any other single-record call, regardless of the sub-draws inside, and stays reproducible from the seed. The new endpoints are live in the /docs playground and the MCP server.

  5. Enhancement

    Mix both standards in one call for the time-versioned ID generators

    The national-ID generators that run two coexisting standards now take standard=both alongside the existing single-standard choices: Ireland’s PPSN (/v1/ie/ppsn) and VAT (/v1/ie/vat) — pre-2013 vs modern; the Netherlands’ btw-id (/v1/nl/btw-id) — legacy vs modern; Latvia’s personas kods (/v1/lv/personas-kods) — legacy vs modern; and Hungary’s személyi azonosító (/v1/hu/szemelyi-azonosito) — pre-1997 vs modern. With both, each record independently draws one of the two standards, so a single batch realistically mixes the old and new forms instead of being uniform. The drawn standard is resolved before generation and reported back where the generator already returns it, so values are never ambiguous. Output stays reproducible from the seed.

  6. Enhancement

    Historic (vintage) plates for the Polish vehicle-registration generator

    The /v1/pl/vehicle-registration generator gains a fifth plate kind: type=historic mints a tablica zabytkowa — the black-on-yellow plate carried by registered vintage vehicles. Its right-hand historic-vehicle symbol shortens the number to five characters, so behind a genuine area code (wyróżnik, drawn from the official registry exactly like a standard plate) sits a short individual part of digits, optionally with one letter — BSI 12A, DJ 15M, IM 62C. Like standard plates it honours voivodeship and county, reports the real region back, and joins the edge=true corner set. Output stays reproducible from the seed.

  7. New generator

    National ID and tax numbers for all 27 EU countries

    Real Fake Data now mints the national identification and tax numbers of every EU member state — 66 new generators, the PESEL/NIP idea taken Europe-wide. Personal numbers (/v1/de/steuer-id, /v1/fr/nir, /v1/it/codice-fiscale, /v1/se/personnummer, /v1/es/dni, …) and company/VAT numbers (/v1/fr/siren, /v1/de/ust-idnr, /v1/nl/btw-id, /v1/it/partita-iva, …) for all 27 countries, each carrying a correct, country-specific checksum — Luhn, mod-11, ISO 7064, Verhoeff, mod-97, the lot — so they pass real validators. Date-encoded personal numbers accept the familiar sex and age/birth-date filters and report the encoded birth date back; company and VAT numbers take format (national vs the EU vat form, e.g. HR… → HR…, Greece’s EL prefix); every generator supports invalid=true (a deliberately wrong checksum, for testing your validators) and edge=true (rare-but-valid corners). Where a country runs two standards the generator lets you pick: Denmark’s pre/post-2007 CPR checksum, the Netherlands’ 2020 btw-id, Latvia’s 2017 personal-code format, Ireland’s 2013 PPSN. Croatia’s single OIB serves both people and companies; legacy numbers (Croatian JMBG) are available too. Everything stays reproducible from the seed, and the new endpoints are live in the /docs playground and the MCP server.

  8. New generator

    Company-name generators for all 27 EU countries

    Company-name generation now spans the EU: new company-name endpoints for every member state at /v1/<country>/company-name, plus a multi-country /v1/company-name?countries=de,fr,it that blends them in one call (each record reports its country). Every country produces names across four naming families — coined portmanteaus (Polbud, Eurosysteme), owner-surname firms with the local partnership form (Kowalski i Syn, Müller & Söhne, Rossi & Figli, García e Hijos), descriptive trading names, and modern tech brands — each closed with that country’s real legal form in its authentic local script: German GmbH / AG, French SARL / SAS, Italian S.r.l. / S.p.A., Spanish S.L., Greek Α.Ε. / Ε.Π.Ε. in Greek letters, Bulgarian ООД / АД in Cyrillic, and so on. The /docs playground enumerates each country’s actual legal forms, so requesting one that doesn’t exist there is rejected. strategy, legalForm, and edge controls work everywhere, founder surnames are shared with the person-name generators, and output stays reproducible from the seed.

  9. New generator

    Mixed-country person-name generator

    A new /v1/person-name endpoint returns names drawn from a mix of EU countries in one call. Pass countries as a comma-separated list of ISO codes — ?countries=pl,sk,it — and each record is generated by one of them, picked at random, so a single batch can blend Polish, Slovak, and Italian names, each in its own local form (script, surname inflection, ordering). Every record reports the country it came from. Omit countries to draw from all 27 EU countries at once. The same sex, edge, and caseStrict controls apply, and output stays reproducible from the seed — handy for seeding a realistic, pan-European set of test users in one request.

  10. New generator

    Person-name generators for 26 EU countries

    Person-name generation now spans the whole EU: new person-name endpoints for all 26 other member states — Austria, Belgium, Bulgaria, Croatia, Cyprus, Czechia, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, Netherlands, Portugal, Romania, Slovakia, Slovenia, Spain, and Sweden — each at /v1/<country>/person-name. Every one uses authentic local given names and surnames in the country’s own script and diacritics: Greek and Cypriot names in Greek letters, Bulgarian in Cyrillic, German umlauts, Polish and Czech diacritics. The local conventions are built in, not bolted on — surnames inflect to the chosen sex where the language does (Czech/Slovak -ová, Lithuanian -aitė, Latvian -a, Greek genitive, Bulgarian -ова, Irish Ní/Nic); Spain and Portugal carry two surnames; Hungary writes the surname first; France hyphenates compound given names (Jean-Pierre); Denmark leans on double given names (Hans Christian); and Dutch/German/Italian nobiliary particles (van der, von, De) are handled correctly, including in the initials. The same sex, edge, and caseStrict controls work everywhere, and every result is reproducible from its seed.

  11. New generator

    Standalone Polish person-name generator

    A new pl/person-name endpoint generates just a realistic Polish name — given name(s), surname, and initials — without minting an identity around it. Often you only need a believable name to label a row, seed a UI, or fill a form, and pulling a full PESEL and birth date for it is overkill; now you can ask for the name alone. It carries the same realism as the full person: the surname is inflected to the chosen sex (Kowalski / Kowalska), names can run to a second given name (Anna Maria) or a double-barrelled surname (Kowalska-Nowak), edge=true pushes those rarer shapes plus very short names (Jan Lis), and caseStrict=false mangles the casing for testing case-insensitive matching. The full pl/person generator is now built on top of it — same names, same seeds — and still returns the name together with the birth date and a matching, checksum-valid PESEL.

  12. Enhancement

    Richer Polish names and casing control

    The Polish person generator now produces realistic name variety — second given names (Anna Maria), double-barrelled surnames (Kowalska-Nowak), and very short names (Jan Lis), all far more likely under edge. A new caseStrict=false switch deliberately mangles the casing of name and surname (all-lower, all-upper, or random like kArOl nOwAkOwSki) for testing case-insensitive matching and storage.

  13. Milestone

    Public website and live playground

    Launched the marketing site with a spec-driven live Playground, an API reference generated straight from the OpenAPI spec, pricing, and self-contained GDPR cookie consent. The product now has a public face people can try before they buy.

  14. Across generators

    Edge-case mode for every generator

    Rolled an edge parameter across all fifteen generators in one sweep: boundary values, unusual-but-valid records, and deliberately awkward inputs on demand. Customers can now test how their systems handle the hard cases, not just the happy path.

    Read the full story →
  15. New generator

    Five Polish identity and registry generators

    Added Polish ID card, passport, KRS, land-register, and driving-licence number generators in one batch — five new document types, each valid by format.

  16. New generator

    Lorem Ipsum generator

    A locale-agnostic Lorem Ipsum generator for filler text, exposed over REST and as a Playwright fixture.

  17. Across generators

    KRS numbers and the P.S.A. legal form for companies

    The Polish company generator now emits KRS registration numbers and can produce the P.S.A. (simple joint-stock) legal form.

  18. Milestone

    Seeded test suites across every generator

    Added seeded invariant test suites covering every generator, so each one is checked for valid output and deterministic, reproducible results on every change.

  19. Across generators

    Bulk generation with count

    Every endpoint can now return many records in a single call. Seed a whole table or fixture set in one request instead of looping over the API.

  20. New generator

    Email addresses

    A locale-agnostic email generator, exposed over REST and as a Playwright fixture.

  21. New generator

    Composite Polish company generator

    A full Polish company generator that assembles a coherent record — name, NIP, REGON, and address — available over REST and as a Playwright fixture.

  22. New generator

    Polish person and company-name generators

    New Polish person and company-name generators — coherent people and believable Polish business names.

  23. Milestone

    Playwright addon launched

    Shipped the @przeslijmi/real-fake-data-playwright addon — seeded, valid fixtures pulled straight into Playwright tests, one line per record. No HTTP glue required.

  24. New generator

    Polish REGON and vehicle-registration generators

    New Polish REGON numbers (with correct checksums) and Polish vehicle registration plates.

  25. Milestone

    Self-service API docs

    Interactive OpenAPI docs went live at /docs, generated from the same Zod schemas that validate every request — so the docs can never drift from the API.

  26. Across generators

    Invalid-data testing

    An invalid parameter across the API returns deliberately invalid records — wrong checksums, malformed fields — so customers can test their rejection paths, not just the happy path.

  27. New generator

    Polish tax and banking identifiers

    Polish NIP, IBAN (with bank selection by code or name), and PESEL generators — every one with correct checksums, so they pass real validation just like production data.

    Read the full story →
  1. Milestone

    Polish addresses at national scale

    Moved the Polish address generator onto the PRG dataset — 8.57 million real Polish address points with valid postal codes — behind a memory-bounded cache so it runs comfortably on modest hosting.

  2. Milestone New generator

    Platform launch and first generator

    The hexagonal core, deterministic PRNG, generator registry, and REST API went live, shipping with the first generator: Polish addresses built from real TERYT cities and streets.