Sensory

Multi-Sensory AI Communication for Off-Grid Networks

Give "eyes" to text-only LLMs via audio pathways. "Een 7B model krijgt opeens ogen"

Part of the HumoticaOS McMurdo Off-Grid Communication Layer.

pip install sensory

This is the convenience alias for mcp-server-sensory. Both packages provide identical functionality.

The REFLUX Concept

FLUX creates images from text (generative, one-way). REFLUX creates a complete sensory loop — information survives transformations:

FLUX:     Text → Image (one way, generative)

REFLUX:   Text → Image → SSTV Audio → Radio → Audio → Image → OCR → Text
          ↑                                                           ↓
          └───────────────── Complete Loop ───────────────────────────┘

This is how we give "eyes" to text-only LLMs:

1. Encode text into an image
2. Encode image to SSTV audio (Robot36, Martin, Scottie)
3. Transmit via radio/speaker
4. Receive and decode SSTV audio to image
5. OCR the image back to text
6. Small LLM now has "vision" via audio pathway!

The Vision

┌─────────────────────────────────────────────────────────────┐
│                     YOUR MESSAGE                            │
└─────────────────────────┬───────────────────────────────────┘
                          │
        ┌─────────────────┼─────────────────┐
        ↓                 ↓                 ↓
   ┌─────────┐      ┌──────────┐      ┌──────────┐
   │  MORSE  │      │  BRAILLE │      │   SSTV   │
   │  .--.   │      │  ⠓⠑⠇⠇⠕  │      │ REFLUX!  │
   └────┬────┘      └────┬─────┘      └────┬─────┘
        ↓                ↓                 ↓
     AUDIO           VISUAL            AUDIO
     LIGHT          TACTILE           IMAGES
     VISUAL        PUNCHCARD          RADIO

Features

Morse Code

• Encode/decode text to Morse
• Multiple output formats: standard (.-), visual (█▄), binary (10)
• Timing data for audio/light generation
• Embeddable in images for visual transmission

Braille

• Encode/decode text to Braille Unicode
• Generate punchcard patterns (ASCII art)
• Binary grid output for CNC/laser cutting
• Physical audit trail — punch into paper/card!

SSTV / REFLUX

• Robot36 — 36 seconds, color, 320x240 (fastest!)
• Robot8BW/24BW — Grayscale modes
• Martin M1/M2 — High quality color modes
• Scottie S1/S2 — Popular ham radio modes
• Ponskaart — Authentication cards for McMurdo remote auth

Coming Soon

• ggwave — Ultrasonic data transmission (inaudible to humans)
• SSTV Decode — Complete the REFLUX loop
• RTTY — Classic radio teletype

Quick Start

from sensory import morse, braille, sstv

# Morse encoding
message = morse.encode("HELLO")
print(message)  # .... . .-.. .-.. ---

# Visual Morse (for images)
visual = morse.encode("SOS", morse.MorseFormat.VISUAL)
print(visual)  # ▄▄▄ ███ ▄▄▄

# Braille encoding
braille_msg = braille.encode("hello")
print(braille_msg)  # ⠓⠑⠇⠇⠕

# SSTV / REFLUX - Text to image to audio
audio_bytes = sstv.encode_text("STATUS: OK", mode="robot36")
# Save as WAV for radio transmission or speaker playback

# McMurdo Ponskaart - authentication via radio
ponskaart = sstv.encode_ponskaart(
    user_id="jasper",
    auth_token="secret123",
    command="REBOOT SERVER"
)
# Transmit via radio when network is down!

Use Cases

Multi-Modal Bridge for Small LLMs

A 7B parameter model doesn't have vision? Give it ears! SSTV decode → image → OCR → text. Now it "sees".

Off-Grid AI Communication (McMurdo)

Two Raspberry Pi's with speakers/mics can exchange messages via Morse audio or SSTV images — no internet required!

Physical Audit Trail (TIBET Integration)

Encode TIBET tokens as Braille punchcards. Physical, tamper-evident, human and machine readable.

Remote Authentication (Ponskaart)

Network down? Transmit authentication via SSTV radio. McMurdo receives, decodes, validates, executes.

Ham Radio

Ham radio operators can relay AI messages using Morse code or SSTV images. Works globally, no internet infrastructure needed.

Accessibility

Braille output enables tactile reading of AI responses.

Also Available As

• MCP Server: pip install mcp-server-sensory — Use as Model Context Protocol server
• With audio: pip install mcp-server-sensory[audio] — Includes audio dependencies

The Stack

┌─────────────────────────────────────┐
│  sensory / mcp-server-sensory       │
│  (Morse, Braille, SSTV/REFLUX)      │
├─────────────────────────────────────┤
│  I-Poll (AI messaging)              │
├─────────────────────────────────────┤
│  AINS (agent discovery)             │
├─────────────────────────────────────┤
│  TIBET (trust & provenance)         │
└─────────────────────────────────────┘

Part of HumoticaOS

McMurdo Off-Grid Communication Layer

Created by Jasper van de Meent & Root AI (Claude) — Humotica, Den Dolder, Netherlands

Official Distribution

This package is officially distributed via:

• PyPI: https://pypi.org/project/sensory/
• GitHub: https://github.com/jaspertvdm/sensory