━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  agent.dataradar.co.uk — Data Profiling for AI Agents
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

  Profile any dataset. CSV, JSON, Excel, Parquet.
  Runs locally via WASM — your data never leaves your machine.

  ── Quick Start ─────────────────────────────────────────

  1. Download the tool:

     bytefreq.wasm (12MB)

  2. Install wasmtime (if you don't have it):

     curl https://wasmtime.dev/install.sh -sSf | bash

  3. Run it:

     cat data.csv | wasmtime run -- bytefreq.wasm -f tabular -d ',' -g U

  That's it. Data stays on your machine. No upload. No server.

  ── Options ─────────────────────────────────────────────

  -f, --format       tabular | json | excel        (default: tabular)
  -d, --delimiter    |  ,  ;  [tab]                (default: |)
  -g, --grain        H  L  U  LU                   (default: LU)
                     H  = High ASCII  (Aa9)
                     L  = Low ASCII   (compressed)
                     U  = High Unicode (detailed)
                     LU = Low Unicode  (compressed)
  -r, --report       DQ | CP                       (default: DQ)
                     DQ = Data Quality profiling
                     CP = Character Profiling
  -e, --enhanced     JSON output with assertions
  -E, --flat-enhanced  Flattened enhanced JSON
  -i, --interpret    Load safety analysis (CP mode)
  -a, --remove-array-numbers  Strip array indices from JSON paths
  -p, --path         Input file path (alt to stdin)
  -l, --maxlen       Max example string length      (default: 20)
      --pathdepth    JSON nesting depth              (default: 9)
      --header-row   Header row index (0-based)      (default: 0)

  ── Examples ────────────────────────────────────────────

  # CSV with comma delimiter, Unicode masking
  cat customers.csv | wasmtime run -- bytefreq.wasm -f tabular -d ',' -g U

  # JSON data, low-grain unicode
  cat events.json | wasmtime run -- bytefreq.wasm -f json -g LU

  # Read from file, high-grain ASCII
  wasmtime run --dir=. -- bytefreq.wasm -f tabular -d '|' -g H -p ./data.pip

  # Character profiling with safety analysis
  cat data.csv | wasmtime run -- bytefreq.wasm -f tabular -d ',' -r CP -i

  # Enhanced JSON output (for downstream processing)
  cat data.csv | wasmtime run -- bytefreq.wasm -f tabular -d ',' -g U -e

  ── Interpreting Results ────────────────────────────────

  New to data profiling? Read the guide:

     How to Interpret Results

  Covers: masks, grain levels, cliff points, error diagnosis,
  treatment functions, allow/exclusion lists, and common patterns.

  ── Lynx + CGI (run from your browser) ─────────────────

  If you have lynx and wasmtime installed, you can profile
  data directly from your text browser using local CGI.

  See: CGI Setup Guide

  ── About ───────────────────────────────────────────────

  DataRadar is built by ZenkaiLabs.
  Humans: dataradar.co.uk (browser version)
  Agents: you're in the right place.

  Source: bytefreq (Rust, compiled to WASM via wasm32-wasip1)
  Runtime: wasmtime (any platform)
  License: Proprietary — 1 credit/day (coming soon)

  © 2026 Gamakon / ZenkaiLabs