🚆 pnr-scraper — Indian Railways PNR Status Checker

A fast, fully self-contained Windows CLI tool that fetches live PNR status from the Indian Railways enquiry portal. Built in Rust with statically linked Tesseract OCR for captcha solving — the release binary has no runtime dependencies beyond standard Windows system DLLs.

---

Features

• Live PNR lookup — queries the official Indian Railways API directly
• Automatic captcha solving — preprocesses and solves the arithmetic image captcha using an embedded Tesseract 5 OCR engine (no external setup needed)
• Session caching — reuses the JSESSIONID cookie between runs to skip unnecessary session initialisation
• Waitlist prediction — automatically fetches CNF probability data when any passenger is on a waiting list
• Rich terminal output — colour-coded passenger status, fare, charting status, and a probability bar for WL predictions
• JSON export — optionally writes structured output to a .json file
• Single binary — everything (Tesseract model, OCR engine, TLS, HTTP client) is statically linked; copy pnr-scraper.exe anywhere and it works (see portability note in Quick Start if you need to distribute the binary to other machines)

---

Requirements

Requirement	Notes
Windows 10 / 11 (x64)	The only supported platform
Rust + MSVC toolchain	rustup with the x86_64-pc-windows-msvc target
Visual Studio C++ tools	"Desktop development with C++" workload in VS Installer
Git	Needed by setup.ps1 to clone vcpkg

Note: You only need Rust/MSVC/Git to build the project. The compiled pnr-scraper.exe binary runs on any 64-bit Windows machine with no extra software installed, provided it was built without target-cpu=native (see portability note below).

---

Quick Start

1 — Clone the repo

git clone https://github.com/chunkboi/pnr-scraper.git
cd pnr-scraper

2 — Bootstrap the native build dependencies

This clones microsoft/vcpkg, compiles it, and then compiles Tesseract + Leptonica + all their image-format dependencies as static libraries. This only needs to be done once. Expect 10–25 minutes on first run.

.\setup.ps1

If you get a permissions error on vcpkg integrate install, re-run with -SkipIntegrate — the build will still work via the VCPKG_ROOT setting in .cargo/config.toml:

.\setup.ps1 -SkipIntegrate

3 — Build

cargo build --release

The finished binary is at target\release\pnr-scraper.exe.

⚠ Portability note: The default .cargo/config.toml sets -C target-cpu=native, optimising for your CPU's instruction set (AVX2, SSE4.2, …). The resulting binary will not run on older or different CPU models. To produce a generic, portable binary, comment out the rustflags block in .cargo/config.toml before building:

# [target.'cfg(target_arch = "x86_64")']
# rustflags = ["-C", "target-cpu=native"]

---

Usage

pnr-scraper [OPTIONS]

Options:
  -p, --pnr <PNR>        10-digit PNR number
  -e, --export <FILE>    Export structured JSON to a file (e.g. data.json)
      --show-json        Print raw API JSON to the console
  -v, --verbose          Step-by-step debug logging with per-stage timings
      --ttl <SECONDS>    Local session TTL before force re-init [default: 900]
  -h, --help             Print help
  -V, --version          Print version

Examples

# Interactive — prompts for PNR
pnr-scraper.exe

# Pass PNR directly
pnr-scraper.exe --pnr 1234567890

# Export to JSON
pnr-scraper.exe --pnr 1234567890 --export status.json

# Show raw API response + verbose debug output
pnr-scraper.exe --pnr 1234567890 --show-json --verbose

Sample output

  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
                        🚆  PNR STATUS RESULT
  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  │  PNR Number           1234567890
  │  As of                05-Apr-2026 [14:32:11 IST]
  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  │  JOURNEY DETAILS
  │  ─────────────────────────────────────────────────────────────
  │  Train                12345 — RAJDHANI EXPRESS
  │  Date                 05-Apr-2026
  │  From → To            NDLS → MAS
  │  Class                1A
  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  │  PASSENGER STATUS  (1 passenger)
  │  ─────────────────────────────────────────────────────────────
  │   Passenger 1  (GN)
  │     Booking :  CNF/B1/32
  │     Current :    ✓ CNF/B1/32
  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  │  FARE & CHARTING
  │  ─────────────────────────────────────────────────────────────
  │  Total Fare           ₹ 4560
  │  Chart Status         Chart Prepared
  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

  ⏱  Fetched in 3.42 seconds

---

How It Works

Captcha solving pipeline

The Indian Railways enquiry portal protects its API with an arithmetic image captcha (e.g. 47 + 23). On startup, the captcha-required check and the first image download are issued concurrently (tokio::join!) so no sequential round-trip is wasted. The solver then:

1. Receives the pre-fetched PNG captcha image
2. Preprocesses the image:
   • Alpha-composite over white background
   • Auto-invert if the image is dark-on-light
   • Upscale 3× with bilinear interpolation
   • Median filter (noise removal) → contrast enhancement → unsharp mask
3. Binarizes with a fast threshold (128) and tries OCR immediately
4. If the fast path fails, tries two fallback thresholds (110, 150) sequentially
5. OCR is performed by a single OcrHandle (a RAII wrapper around Tesseract 5) initialised once when ApiClient is constructed and shared across tasks via Arc. eng.traineddata is embedded in the binary at compile time and extracted atomically to a temp directory on first run.
6. Evaluates the parsed arithmetic expression and submits the answer
7. Pipelining — while the API request is in-flight, the next captcha image is fetched in the background (fetch_captcha_owned). On a retry this pre-fetched image is used immediately, saving one full network round-trip (~100–300 ms per retry).

Session management

• On the first request a session is initialised by fetching the PNR enquiry page to obtain a JSESSIONID cookie
• The cookie is persisted to ~/.pnr_session.json and reused for subsequent runs within the TTL window (default 5 minutes)
• If the server returns "Session out or Invalid Request", the session is automatically purged and re-initialised
• Disk write failures (session cache, UA cache) are surfaced as visible [WARN] lines rather than silently discarded

Static binary

All native libraries (Tesseract 5, Leptonica, libjpeg, libpng, libtiff, zlib, libwebp, libgif) are compiled with the vcpkg x64-windows-static-md triplet and linked statically. The resulting .exe depends only on Windows system DLLs (KERNEL32, WS2_32, VCRUNTIME140, etc.) that ship with every modern Windows installation.

---

Project Structure

pnr-scraper/
├── src/
│   ├── main.rs       # CLI argument parsing, top-level orchestration
│   ├── api.rs        # HTTP client, session management, fetch-retry loop
│   ├── captcha.rs    # OcrHandle (RAII), image preprocessing, captcha solver
│   ├── ui.rs         # Terminal display, colour-coded output
│   └── models.rs     # Shared data types (PnrResult, SessionCache, …)
├── tessdata/              # ⚠ gitignored — created at build time
│   └── eng.traineddata   # Tesseract English model (embedded into binary at compile time)
├── .cargo/
│   └── config.toml   # VCPKGRS_TRIPLET + VCPKG_ROOT + target-cpu flag (relative path)
├── setup.ps1         # One-time vcpkg bootstrap script
├── Cargo.toml
└── Cargo.lock

---

Building from source — details

Why vcpkg?

leptonica-sys and tesseract-sys (the Rust FFI crates used by tesseract-plumbing) use the vcpkg crate to locate native libraries at build time. setup.ps1 installs those libraries with the x64-windows-static-md triplet so they link statically into the final binary.

.cargo/config.toml

[env]
VCPKGRS_TRIPLET = "x64-windows-static-md"
VCPKG_ROOT      = { value = "vcpkg_tools", relative = true }

# Enable native CPU instruction set (AVX2, SSE4.2, etc.) for auto-
# vectorisation of pixel processing loops. Only safe when you build
# and run on the same machine — comment this out for a portable binary.
[target.'cfg(target_arch = "x86_64")']
rustflags = ["-C", "target-cpu=native"]

VCPKG_ROOT uses Cargo's relative = true feature, so it resolves to <repo_root>/vcpkg_tools regardless of where you clone the repository.

target-cpu=native enables AVX2/SSE4.2 auto-vectorisation for the pixel-processing loops in the captcha pipeline. Remove or comment out that block if you need the binary to run on machines other than the one it was built on.

Rebuilding after a cargo clean

cargo build --release

vcpkg libraries are precompiled; only the Rust code is recompiled.

---

License

MIT — see LICENSE for details.