73-wonderphi-phone

Description

Intellectual–property status: All rights, title, and interest in and to WonderPhi Phone — including its source code, plan–file formats, recipe library structure, and documentation — are the exclusive property of Christopher Gabriel Brown, protected by U.S. copyright law (17 U.S.C. § 102(a)(1)) and applicable international treaties. WonderPhi Phone is a companion product to WonderPhi Compute , which carries patent–pending status under the inventor’s USPTO application portfolio (41 applications filed: 28 utility, 6 design, 7 provisional). The WonderPhi name, the constitutional commandments text, the plan.wpc.txt format, and the recipe library structure are the author’s intellectual property.

Overview

WonderPhi Phone v1.0.0 is a library of 28 batch recipes for telephony operations that require zero third–party APIs. It handles voicemail transcription, call–recording processing, CDR (Call Detail Record) parsing, phone–number normalisation, Do–Not–Call scrubbing, and the innovative “sudo–SMS” queue–to–file pattern for operator–delivered messaging — all without ever contacting Twilio, Plivo, Bandwidth, or any other paid gateway.

The library does not run anything itself. It writes a plain–text plan file ( .plan.wpc.txt ) that WonderPhi Compute then executes, dispatching every task across every available CPU core (and GPU, when present). The relationship is simple: WonderPhi Phone writes the plan; WonderPhi Compute runs the plan.

WonderPhi Phone is written in pure Python with zero external dependencies — standard library only. It makes no network connections and includes no telemetry. It never modifies wpc.exe ; the 10 Commandments that govern WonderPhi Compute remain intact at all times.

The “sudo–SMS” Pattern

SMS and voicemail–drop recipes write one JSON queue record per phone number to a local folder. Your operator (or a downstream script you write) delivers them manually from that queue. Phone number is identity — that identity stays with you, not a paid gateway. No Twilio. No monthly bill scaling with volume. A typical queue record looks like this:

{ "phone": "+15555550100", "message": "Your appointment is confirmed for tomorrow at 2pm.", "queued_at": 1721600400.123 }

Recipe Input Model Requires on PATH Description

whisper-vm files whisper + CUDA Transcribe each voicemail with Whisper medium model on GPU (set $WPC_GPU=1 )

whisper-vm-cpu files whisper (CPU only) Transcribe each voicemail on CPU only using the Whisper base model

whisper-vm-tiny files whisper Transcribe each voicemail with Whisper tiny model (fastest; lowest accuracy)

whisper-vm-srt files whisper Transcribe each recording with timestamped SRT output (courtroom–ready)

Audio Processing (7 recipes)

Recipe Input Model Requires on PATH Description Output

loudness-normalize files ffmpeg EBU R128 loudness–normalize each recording (broadcast–safe archival); parameters: I=−16, LRA=11, TP=−1.5 .mp3 (LAME V2)

silence-trim files ffmpeg Trim silence from head and tail of each recording (voicemail cleanup); threshold −40 dB, minimum silence 0.5 s .mp3 (LAME V2)

audio-to-mp3 files ffmpeg Convert each recording to standard MP3 (V2) for archival .mp3

audio-to-opus files ffmpeg Convert each recording to Opus at 32 kbps (long–term storage; approximately 60% smaller than MP3) .opus

audio-mono-8k files ffmpeg Downsample each recording to 8 kHz mono PCM (PSTN–quality; STT–ready); codec: pcm_s16le .wav

audio-spectrogram files ffmpeg Emit a spectrogram PNG per recording (visual audit / anomaly detection); image size 1024×512 .png

audio-waveform files ffmpeg Emit a waveform PNG per recording (silence / clipping audit); image size 1200×200, mono channel layout .png

Fax Processing (2 recipes)

Recipe Input Model Requires on PATH Description Output

fax-tif-to-pdf files ImageMagick + Ghostscript Convert each fax TIFF (including multi–page) to PDF .pdf

fax-ocr-searchable files tesseract OCR each fax TIFF into a searchable text–layer PDF (English language) .pdf

CDR Parsing (2 recipes)

Recipe Input Model Requires on PATH Description Output

cdr-generic-parse files python (stdlib) Parse each CDR CSV file; emit a JSON summary containing total calls, total minutes, and unique–caller count .summary.json

cdr-toll-fraud-scan files python (stdlib) Flag international and premium–rate destinations per CDR for toll–fraud detection (non–domestic = any destination not starting with +1 or 1) .fraud.txt

Phone Number Operations (4 recipes)

Recipe Input Model Requires on PATH Description

e164-normalize lines python (stdlib) Normalize each phone number to E.164 format (+1 followed by the last 10 digits); invalid numbers are flagged

dnc-scrub lines python (stdlib) + DNC list via --tag Flag each number against a local Do–Not–Call list (path provided via --tag ); outputs OK or DNC per number

cnam-lookup-file lines python (stdlib) + cnam.csv via --tag Local caller–ID name lookup from a CSV map (format: phone,name); returns UNKNOWN for unmatched numbers

dial-plan-validate lines python (stdlib) Validate each number against the US dial plan (10–digit NANP); checks that the leading digit is not 0 or 1

SMS Queue — sudo–SMS Pattern (4 recipes)

Recipe Input Model Requires on PATH Description

sms-queue-write lines python (stdlib) Write one SMS–queue JSON record per number (message text via --tag ); operator delivers manually

sms-appointment-reminder-queue lines python (stdlib) Queue an appointment–reminder SMS per number ( --tag = template string, e.g. “Your appt with Dr. X is tomorrow at 2pm.”)

sms-otp-queue lines python (stdlib) Queue a one–time–passcode SMS per number ( --tag = the OTP code)

sms-broadcast-queue lines python (stdlib) Queue the same broadcast message to every number in the list ( --tag = message text)

Voice Queue (2 recipes)

Recipe Input Model Requires on PATH Description

voicemail-drop-queue lines python (stdlib) + audio file via --tag Queue one voicemail–drop record per number; the JSON record includes phone, audio path (from --tag ), action type, and timestamp

reminder-call-queue lines python (stdlib) Queue a reminder–call task per number ( --tag = script text); the JSON record includes phone, action type, script, and timestamp

Compliance & Archival (3 recipes)

Recipe Input Model Requires on PATH Description Output

recording-hash-sidecar files certutil (built–in on Windows) SHA–256 hash sidecar per recording for tamper–evident regulatory archival .sha256

recording-encrypt files 7z (7–Zip) AES–256 password–encrypt each recording with 7z (password via --tag ); header encryption enabled ( -mhe=on ) .7z

transcript-redact-pii files python (stdlib) Regex–redact Social Security Numbers (###–##–####), credit–card numbers (####–####–####–####), and phone–number patterns from each transcript; outputs redacted text with markers [SSN–REDACTED], [CC–REDACTED], [PHONE–REDACTED] .redacted.txt

Supported Input Formats

Category File Extensions Scanned

Audio recordings *.wav , *.mp3 , *.m4a , *.aac , *.flac , *.ogg , *.opus , *.amr

Fax images *.tif , *.tiff

Call Detail Records *.csv , *.cdr , *.log

Transcripts (for PII redaction) *.txt , *.srt

Input Models

Model How It Works

files Recursively scans a folder for files matching the recipe’s extension list; each matched file becomes one task in the plan

lines Reads a text file with one target per line (an E.164 phone number, a file path, etc.); blank lines and lines starting with # are skipped

python wpp-recipe.py --list python wpp-recipe.py [options]

Flag Description

--list List all 28 recipes with their categories, input models, dependencies, and descriptions, then exit

--out FILE Where to write the generated plan file (default: .plan.wpc.txt )

--outdir DIR Where output files go (default: wpp-out )

--tag VALUE Value substituted for {tag} in each recipe template — serves as message text, DNC list path, audio path, encryption password, etc., depending on the recipe

--stdout Print the generated plan to standard output instead of writing a file

--check Dry run: report how many inputs matched without writing anything

External Tool Dependencies

WonderPhi Phone itself has zero external Python dependencies — it uses only the Python standard library. However, the plans it generates invoke the following external tools, which must be present on the operator’s machine:

Tool Used By Purpose

whisper (OpenAI Whisper) whisper–vm, whisper–vm–cpu, whisper–vm–tiny, whisper–vm–srt Speech–to–text transcription (models: medium, base, tiny)

ffmpeg loudness–normalize, silence–trim, audio–to–mp3, audio–to–opus, audio–mono–8k, audio–spectrogram, audio–waveform Audio processing, format conversion, spectrogram and waveform generation

tesseract fax–ocr–searchable Optical character recognition for fax TIFF images

ImageMagick (+ Ghostscript) fax–tif–to–pdf Fax TIFF to PDF conversion

certutil recording–hash–sidecar SHA–256 hashing (built–in on Windows)

7z (7–Zip) recording–encrypt AES–256 archive encryption with header encryption

System Requirements

Requirement Details

Python 3.8 or later (standard library only; no pip install required)

WonderPhi Compute Required to execute the generated .plan.wpc.txt files ( wpc.exe )

Operating System Windows (plan files are cmd.exe –safe; certutil is Windows built–in)

GPU (optional) CUDA–capable GPU recommended for whisper-vm recipe; CPU–only alternatives provided

Usage Examples

Task Command

Transcribe 5,000 voicemails on GPU python wpp-recipe.py whisper-vm .\voicemails

Normalize recordings to broadcast–safe loudness python wpp-recipe.py loudness-normalize .\call_recordings

Downsample recordings for speech–to–text python wpp-recipe.py audio-mono-8k .\recordings

Visual anomaly audit via spectrograms python wpp-recipe.py audio-spectrogram .\suspicious_calls

Parse CDR files into JSON summaries python wpp-recipe.py cdr-generic-parse .\cdrs

Scan CDRs for toll–fraud patterns python wpp-recipe.py cdr-toll-fraud-scan .\cdrs

Normalize a phone list to E.164 python wpp-recipe.py e164-normalize .\raw_phones.txt

DNC–scrub against a local list python wpp-recipe.py dnc-scrub .\prospects.txt --tag dnc_list.txt

Queue SMS appointment reminders python wpp-recipe.py sms-appointment-reminder-queue .\appts.txt --tag "Your appt with Dr. X is tomorrow at 2pm."

SHA–256 every recording for regulatory archival python wpp-recipe.py recording-hash-sidecar .\call_recordings

AES–256 encrypt a recording archive python wpp-recipe.py recording-encrypt .\call_recordings --tag "YourPassphrase"

Redact SSN, CC, and phone patterns from transcripts python wpp-recipe.py transcript-redact-pii .\transcripts

How It Works — The Plan–File Pipeline

- Generate the plan: Run python wpp-recipe.py . WonderPhi Phone scans your input (a folder of files or a text file of phone numbers) and writes a .plan.wpc.txt file containing one command per matched item.

- Execute the plan: Run wpc.exe . WonderPhi Compute reads the plan and dispatches every command across all available cores, respecting its constitutional commandments.

- Collect results: Output files appear in the --outdir directory (default: wpp-out ). For queue recipes (sudo–SMS, voicemail–drop, reminder–call), JSON records appear in the output directory for your operator to deliver.

The plan file follows the # WPC v1 header format and includes metadata comments documenting the recipe name, description, input model, tag value, and task count.

Key Design Principles

• No network: WonderPhi Phone makes zero network connections. Every operation is local.
• No telemetry: Nothing phones home. Ever.
• No third–party APIs: No Twilio, Plivo, Bandwidth, or any paid telephony gateway.
• No external Python dependencies: Standard library only. No pip install .
• Constitutional integrity: Never modifies wpc.exe . The 10 Commandments stay intact.
• Phone number is identity: Your phone numbers stay with you, not with a provider.
• Separation of concerns: WonderPhi Phone writes plans; WonderPhi Compute executes them.

License Scope Key Terms

Retail Single–User Perpetual, non–exclusive, non–transferable license tied to you, the buyer

Install on any and all machines you personally own or administer. No expiration, no annual renewal, no seat count, no activation, no telemetry. May process any lawful workload (personal or commercial) that produces output for yourself or your employer. Source code may be read, studied, and modified for personal understanding and personal–use modifications.

Industrial (Commercial) Perpetual, non–exclusive, non–transferable license tied to a single legal entity

Install on any number of machines, VMs, containers, or cloud instances operated by the licensed organisation. No seat count, no machine limit, no annual renewal, no activation, no telemetry. May process production, customer–facing, and revenue–generating workloads. Employees, contractors, and consultants may operate under the organisation’s direction. Repackaging, rebranding, or exposing the software as a service to third parties requires a separate OEM license.

Both licenses are governed by the laws of the State of Georgia, United States of America, with exclusive jurisdiction in the state or federal courts sitting in Gwinnett County, Georgia.

What Is in the Package

File Description

wpp-recipe.py The recipe generator — a single Python script (standard library only, zero dependencies) containing all 28 recipe definitions, the file–scanner, the line–reader, the plan emitter, and the command–line interface

README.md Full product documentation including overview, recipe list, usage examples, the sudo–SMS pattern explanation, command–line options reference, and package contents

INDEX.md Complete recipe index table listing all 28 recipes with their category, input model, PATH dependencies, and one–line description

LICENSE WonderPhi Software License v1.0.0 — Retail Single–User terms (9 sections covering grant, restrictions, intellectual property, warranty disclaimer, liability limitation, termination, governing law, and entire–agreement clause)

The Industrial (Commercial) license ( LICENSE-INDUSTRIAL.txt ) is provided separately with the Industrial–tier purchase.

Version History

Version Notes

1.0.0 Initial release — 28 recipes across 7 categories (Transcription, Audio Processing, Fax, CDR Parsing, Phone Numbers, SMS Queue, Voice Queue, Compliance & Archival)

How this is delivered

This package ships on a single Blu-ray disc, posted to the delivery address on your order. The archive on the disc is AES-256 encrypted; the passphrase is sent separately, by email, once the disc is despatched — so the disc alone is of no use to anyone who intercepts it.

Discs are despatched by tracked, signed-for post. Allow 5 business days for mastering and verification before despatch.

© 2026 Christopher Gabriel Brown

Photo