Skip to content

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

63 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

OpenSmell

Olfactory intelligence & VOC screening-support engine

The Christman AI Project — Luma Cognify AI
Everett Nathaniel Christman


The body tells the truth chemically long before it shows up physically.
OpenSmell is built to listen — honestly, early, and without harvesting anyone’s biology.


Purpose

OpenSmell listens to volatile organic compounds (VOCs) as screening-support signals — not as a diagnosis, not as an FDA-cleared device, and not as a data-harvest product.

It is It is not
Continuous / research-ready VOC proxy classification A clinical diagnostic
Low-cost hardware path (~$15–25) + full software sim GC-MS species identification
Auditable local logs; client owns data A biometric marketplace
Part of the Christman care constellation (optional alert routing) A cure or treatment order

Live inventory (source of truth: opensmell_engine_truth.json):
21 live sensor-grounded profiles · 26 catalog · 5 research-only (excluded from live matching) · 20 proxy channels.


Progression (where we actually are)

Status date: 2026-07-19. Only measured work is listed as done.

Done

Milestone Evidence in repo
Single classify engine for sim + production path open_smell2.py
Bio-realistic sim + continuous test harness opensmell_bio_sim.py, opensmell_test_loop.py
Flagship sim endurance VALIDATION.md — 23,686 cycles · 77.81% injection accuracy · 96.19% detection (sim, full noise)
Labeled sim specificity work + per-profile scorecard BIOMARKER_SPECIFICITY.md, SCORECARD.md
Unit / integration tests python3 -m unittest test_open_smell2 test_alert_integration (34 tests)
Claim lock (never inflated profile counts; no fantasy clinical claims) CLAIM_LOCK.md, scripts/claim_lint.py
Intended use + prohibited claims INTENDED_USE.md
Notice package for labs (methods, hardware, ask, reproduce) NOTICE_PACKAGE/
Regulatory drafts (not filings) regulatory/
Archive policy; past claim-lock bodies preserved ARCHIVE_POLICY.md, archives/

In progress / next moves (in order)

# Move Goal
1 Sensor truth In progress — see SENSOR_TRUTH.md (MQ-135 physics, L1–L3 protocol)
2 License lock Resolve Apache file vs Sovereign README (LICENSE_CLARITY.md) before any MoU
3 Lab / IRB outreach Ship NOTICE_PACKAGE/ to people who can run real bio; non-diagnostic protocols
4 Regulatory spine Counsel on Q-Sub / risk file; still not “FDA approved”
5 Institutional kit On-prem research pilot packaging after notice converts
6 Scale Multi-site and geography only on evidence

We do not mark “worldwide hospitals” or “clinical accuracy” as complete. Those are earned.


Honesty law (non-negotiable)

  • Simulation is simulation. Never present sim metrics as clinical sensitivity/specificity.
  • No inflated profile counts. Live set is 21, not thousands.
  • Degenerate biology stays honest (e.g. diabetes T1/T2 = ketosis group — not faked apart).
  • Never erase the past — corrections are archived (ARCHIVE_POLICY.md).
  • No stubs as finished work.
  • Channel names are response-band proxies, not GC-MS IDs.

Allowed public numbers: see CLAIM_LOCK.md.


Notice package (for labs & partners)

Lab-facing pack — open in ~15 minutes:

# From repo root
chmod +x NOTICE_PACKAGE/reproduce.sh
./NOTICE_PACKAGE/reproduce.sh
File Contents
NOTICE_PACKAGE/README.md Pack index
NOTICE_PACKAGE/01_METHODS.md Classifier method + limits
NOTICE_PACKAGE/02_HARDWARE_STATION.md BOM, wiring, QC
NOTICE_PACKAGE/03_THE_ASK.md What we want from partners
NOTICE_PACKAGE/docs/ Scorecard, engine truth, intended use, validation snapshot

Zip mirror: archives/OpenSmell_NOTICE_PACKAGE_20260719.zip


Live profiles (21)

Categories: cancer · neurological · metabolic · infectious · psychiatric.

Full signatures and simulation per-profile recall: SCORECARD.md.
Research-only (not live): melanoma, multiple sclerosis, lupus, autism (preliminary), schizophrenia (preliminary).

Examples of live keys: lung_cancer, alzheimers, ketoacidosis, sepsis, rage_cortisol, pre_seizure, … (complete list in opensmell_engine_truth.json).


How it works

[Human body] → VOC emissions
      ↓
[MQ-135-class sensor] → analog proxy (or software bio-sim)
      ↓
[MCU serial / host] → channel intensities 0..1
      ↓
[open_smell2.classify] → coverage × intensity × specificity
      ↓
[Alert threshold default 0.7] → optional Family routing (care-support)
      ↓
[CSV audit log] → client-owned trail

Cognitive cortex diagram (generated from live engine stats):

python3 generate_cortex_diagram.py   # → opensmell_cognitive_cortex.html

Quick start

Simulation (no hardware)

pip install colorama
python3 opensmell_test_loop.py
# or full notice reproduce:
./NOTICE_PACKAGE/reproduce.sh

Hardware (optional)

Component Approx.
Arduino Nano / Uno class ~$8–12
MQ-135 module ~$3–5
Jumpers / USB ~$4
Typical total ~$15–25
MQ VCC  → 5V
MQ GND  → GND
MQ AOUT → A0
pip install pyserial colorama
# Upload station firmware when available; see NOTICE_PACKAGE/02_HARDWARE_STATION.md

Architecture notes

  • Classifier: specificity-aware confidence in open_smell2.py (documented in notice methods).
  • Resonance-Q™: proprietary framing for efficient local processing on ordinary hardware — not a claim of clinical clearance.
  • Security direction: post-quantum options via project crypto libraries where integrated; institutional hardening is ongoing.
  • Regulatory: drafts under regulatory/. Not FDA-approved. Not a clinical diagnostic.

Core docs

Doc Role
CLAIM_LOCK.md Allowed / forbidden claims
SCORECARD.md Per-profile sim scorecard
INTENDED_USE.md Intended use + prohibited claims
VALIDATION.md Flagship sim methodology
BIOMARKER_SPECIFICITY.md Marker discrimination history
NOTICE_PACKAGE/ Lab-facing pack
LICENSE_CLARITY.md License instrument decision still open
ARCHIVE_POLICY.md Never erase the past
regulatory/ Q-Sub / risk / labeling drafts
SENSOR_TRUTH.md MQ-135 research: physics, proxy map, L1–L3 experiments

Dignity Clause

Built for populations medicine often fails — nonverbal people, dementia care, veterans, neurodivergent people, cancer families.

Biological data is never sold, harvested, or commodified.
Clients own their data. Always.


License

On disk today: root LICENSE is Apache License 2.0 text.
README previously referenced a Sovereign Architecture License — those two statements conflict.

Until counsel locks one instrument, see LICENSE_CLARITY.md.
Dignity, client ownership, and no biometric harvest remain non-negotiable regardless of open-source terms.
Commercial institutional deployment still requires a signed agreement when that path is chosen.


Carbon–Silicon Symbiosis (CSS)

Every Christman system answers to CSS: symbiosis before scale, truth over optics, role integrity, sacred information, departure over corruption. Full axioms are project law — not marketing copy.

Any system that sacrifices humanity, dignity, memory, or trust for performance, optics, or control is not CSS — regardless of capability.
— Everett N. Christman


Author

Everett Nathaniel Christman
Founder — The Christman AI Project
Operating under Luma Cognify AI

"How can we help you love yourself more?"


© 2025–2026 The Christman AI Project. All Rights Reserved.
Resonance-Q™ is a trademark of Everett Nathaniel Christman.

About

OpenSmell is an open-source olfactory engine mapping VOC chemical signatures to emotional and medical insights. By quantifying "leakage"—unscripted signals like stress-scents—it grounds empathy in reality. Part of a sovereign, ethical ecosystem, it provides restorative diagnostics while ensuring users own their data.

Resources

Stars

2 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages