vtaircrafts.in

Every aircraft on an Indian scheduled or non-scheduled operator permit, as a searchable site and an open dataset, rebuilt monthly from the DGCA's own lists.

Live: https://vtaircrafts.in · Data: data/latest/ · Licence: code MIT, data CC BY 4.0

Why

India's regulator has not published a machine-readable aircraft register since 2019. What it does publish, every month, are two PDFs: the list of scheduled operators and the list of non-scheduled operators, each with every registration, model and seat count under that permit. This project turns those PDFs into a versioned dataset and a site you can actually browse.

Coverage is commercial aviation only: roughly 1,300 of about 2,300 manned aircraft on the register. Privately owned aircraft, flying schools and state government fleets have no public list.

What you get

• Fleet explorer with search by registration, hex, operator, type and model, faceted filters, a dense log view and a photo grid view. State lives in the URL.
• Aircraft pages for every tail: type, ICAO designator, Mode S hex, seats, permit number and validity, serial, build year and age, engines, owner and lessor, source page, and a photo with credit.
• Archive pages for about 2,200 registrations no longer on a permit (Jet Airways, Kingfisher, Go First, Vistara and others), from the registration reports and old lists.
• Operator pages with type mix, fleet age, owners, full fleet lists and former aircraft, plus pages for past operators.
• Type, owner and lessor, flying school and stats pages derived from the same data.
• Change log going back to 2005: exact-dated registrations, de-registrations and ownership changes from DGCA's own reports, plus additions, removals and operator moves between 73 archived snapshots of the two operator lists.
• The dataset in data/latest/ in this repository, plus a GitHub Release per snapshot. The site itself does not serve downloads.

How it works

DGCA PDFs ──▶ pdftotext -bbox-layout ──▶ positional parsers ──▶ normalise ──▶ enrich ──▶ snapshot JSON/CSV ──▶ Next.js static site
                     ▲                                                            ▲
   Wayback captures ─┘                                    DGCA registration reports
   of the same lists                                      (exact dates per airframe)

1. pipeline/fetch.ts downloads both PDFs, reads the "updated as on" date, and stores them unmodified under data/raw/<date>/ with hashes.
2. pipeline/parse.ts extracts word bounding boxes with poppler and rebuilds the tables. It detects columns from the header row, strips the first operator's header that the spreadsheet export stamps on every page, re-joins registrations split across lines, and validates every model group against the count DGCA prints beside it. Mismatches are recorded, never silently fixed.
3. pipeline/images.ts looks up a Wikimedia Commons category per registration and records the chosen photo with its author and licence.
4. pipeline/history.ts builds the historical record, and is run by hand rather than monthly because its inputs never change:
   • it enumerates every Wayback Machine capture of the two operator lists (across all the URLs they have lived at since 2004) with the CDX API, downloads each unique digest, rejects anything that is not a PDF, parses it with the same parsers, reads the "as on" date printed on page 1, and writes data/parsed/history/<list>/<date>.json;
   • it downloads and parses DGCA's registration and de-registration reports (October 2009 to date) and change-of-ownership reports (October 2009 to August 2013) into data/parsed/reports/. These are the only public DGCA documents that carry an exact date per airframe.
5. pipeline/build.ts normalises operator names, classifies models into types, adds hex codes and ICAO designators from the tar1090 aircraft database, picks representative photos for tails without one (same type and operator, else same type), builds the event log from the whole snapshot chain plus the reports, and writes the snapshot plus a diff against the previous one.
6. pipeline/training.ts parses DGCA's list of flying training organisations into training.json, and pipeline/extras.ts writes the archive of off-permit registrations, past operators and per-registration enrichment (engine from the model number, build year and serial from VRS standing-data, CC0).
7. The Next.js app is a static export that reads data/latest/ at build time.

A GitHub Actions workflow runs fetch → parse → images → build on the 3rd of every month and commits the new snapshot; the new snapshot simply extends the chain, so the event log grows without re-reading the archive. Deploys happen from main.

Run it locally

brew install poppler          # pdftotext
npm install
npm run pipeline:history                       # once: archived snapshots + DGCA reports
npm run pipeline:parse data/raw/2026-08-31
npm run pipeline:images data/parsed/2026-08-31
npm run pipeline:build 2026-08-31 -- --previous 2024-06-10
npm run pipeline:training
npm run pipeline:extras
npm run dev

pipeline:history is idempotent and skips anything already on disk; add --no-fetch to re-parse without touching the network.

Data notes

• Registrations are unique across the dataset. The two DGCA lists do not overlap.
• Dates. An event is either exactly dated or interval dated, never guessed. registered, deregistered and owner-change carry a date straight from a DGCA report. added, removed and moved carry from/to: the two snapshot dates the change happened between. owner-change also carries from/to: the 2009-2013 ownership reports print each aircraft's registration date rather than the date of the change, so the event is dated to the calendar month of the report section it appears in. Where a snapshot PDF prints no "as on" date, the Wayback capture date is used and the parse records a warning.
• firstSeen is the earliest snapshot in the whole chain in which a registration appears, archived captures included, so it can predate this repo's own first fetch. It is when DGCA first listed the tail on a commercial permit, not when the aircraft was registered; history.registeredOn is the registration date when a report supplies one.
• Reused registrations. DGCA reissues tails, so a report row is only attached to today's aircraft when the type in the report is the same kind of machine as the type flying now. Where they disagree the fields stay null rather than describing a different airframe.
• Operator identity is resolved through identifyOperator, so a rename (AIX Connect → Air India Express, Air India Charters → Air India Express, Airline Allied Services → Alliance Air) is not reported as a fleet-wide transfer. The older exports also print the operator column narrower and truncate long names; an id that only extends another and never appears alongside it in the same list on the same day is treated as the same company. And a tail that reads as operator A, then B, then A across three consecutive snapshots is treated as one misread snapshot rather than two transfers, so neither move is reported.
• Coverage gaps. The Archive holds no usable capture of either list between February 2020 and January 2023: for that period the site kept the PDFs behind a JavaScript viewer, and only the wrapper HTML was archived. Registration and de-registration reports cover October 2009 to date except July to December 2021, which DGCA published as PDFs with no text layer. The current year's pair (2026 on) is a year-to-date file DGCA re-publishes in place; the monthly workflow re-fetches it with npx tsx pipeline/history.ts --current. Ownership-change reports stop at August 2013.
• seats is the number DGCA prints; freighters and aerial-work aircraft carry role instead. Some rows read 222/232 for mixed configurations; the first number is used.
• Where DGCA's stated count for a model group disagrees with the registrations it lists, both numbers are kept and the discrepancy is recorded under issues in meta.json. It is not shown on the site.
• Photos are hotlinked from Wikimedia Commons and credited per image. A tail with no photograph of its own shows a labelled representative photo: the same type in the same operator's livery where one exists, else the same type elsewhere. Photographs of military, police or agency aircraft, accidents, seizures, drawings and museum pieces are never used (pipeline/lib/photoRules.ts).

Contributing

Type classification rules live in pipeline/lib/normalize.ts; operator display names and websites in the same file. Pull requests that fix a misread model, add a missing alias, or improve the parser on a new DGCA layout are welcome. Run npm run lint && npm run typecheck before opening one.