This repository now has a direct operational tool:
python -m omnia_radar_auditor.cli --input examples/sample_radar_stream.jsonl --out-dir report
It solves a concrete problem:
given streams of structural observations over time,
detect normal signal,
drift,
anomaly,
and alert conditions.
In short:
streams / observations over time -> signal / drift / anomaly / alert report
Systems often fail gradually before they fail visibly.
OMNIA-RADAR turns temporal structural movement into a reproducible audit:
group observations by stream_id
order each stream by time or step
build a baseline window
measure deviation from baseline
measure step-to-step shock
classify each observation as signal, drift, anomaly, or alert
emit a reproducible certificate
optionally fail CI when alerts appear
Clone the repository:
git clone https://github.com/Tuttotorna/OMNIA-RADAR.git
cd OMNIA-RADAR
Install locally:
pip install -e .
The auditor only uses the Python standard library.
Run the sample audit:
python -m omnia_radar_auditor.cli --input examples/sample_radar_stream.jsonl --out-dir report
Run and fail if alert is detected:
python -m omnia_radar_auditor.cli --input examples/sample_radar_stream.jsonl --out-dir report --fail-on-alert
Run and fail if anomaly or alert is detected:
python -m omnia_radar_auditor.cli --input examples/sample_radar_stream.jsonl --out-dir report --fail-on-anomaly
The auditor accepts JSONL.
Required fields:
stream_id
step
value
Optional fields:
timestamp
domain
metric
note
Example:
{"stream_id":"omega_stream","step":1,"value":0.91}
{"stream_id":"omega_stream","step":2,"value":0.90}
{"stream_id":"omega_stream","step":3,"value":0.40}
Classification rule:
signal = stable movement inside baseline tolerance
drift = gradual deviation from baseline
anomaly = strong deviation or shock
alert = critical deviation or repeated anomaly
The auditor writes:
report.json
report.csv
report.html
drift_events.jsonl
anomaly_events.jsonl
alert_events.jsonl
certificate.json
Meaning:
report.json
Full structured radar analysis.
report.csv
Spreadsheet-friendly event summary.
report.html
Human-readable radar report.
drift_events.jsonl
One JSON object per drift event.
anomaly_events.jsonl
One JSON object per anomaly event.
alert_events.jsonl
One JSON object per alert event.
certificate.json
Reproducibility certificate with thresholds, counts, and boundary statement.
Fail when alert appears:
python -m omnia_radar_auditor.cli --input examples/sample_radar_stream.jsonl --out-dir report --fail-on-alert
Fail when anomaly or alert appears:
python -m omnia_radar_auditor.cli --input examples/sample_radar_stream.jsonl --out-dir report --fail-on-anomaly
Exit codes:
0 = analysis completed without selected blocking condition
2 = anomaly detected under --fail-on-anomaly
3 = alert detected under --fail-on-alert or --fail-on-anomaly
4 = invalid input or measurement error
This is not prediction.
It does not infer future truth.
It does not decide what action to take.
It measures structural movement inside the supplied stream boundary.
The boundary is explicit:
measurement only;
radar alert means structural deviation inside supplied streams,
not semantic danger or future prediction.
The rest of this repository documents the radar concept:
temporal structural monitoring
signal movement
drift
anomaly
alert
baseline deviation
shock detection
measurement boundary
The code above is the operational entrypoint.
The repository below is the derivation path.
Repository: Tuttotorna/OMNIA-RADAR Release tag: v2026.05.21 Release commit: f0c14e8 Release DOI: 10.5281/zenodo.20322686
Boundary:
measurement != validation validation != orchestration orchestration != decision decision != measurement
Zenodo DOI badge for this repository.
Repository: Tuttotorna/OMNIA-RADAR GitHub repository id: 1143359510 Release tag: v2026.05.21 Latest release DOI: 10.5281/zenodo.20322686
Release DOI: 10.5281/zenodo.19829096
GitHub release: OMNIA-RADAR v1.0.0 release
From a clean environment:
git clone [OMNIA-RADAR.git](https://github.com/Tuttotorna/OMNIA-RADAR.git)
cd OMNIA-RADAR
python -m pip install -e .
pytest
If example scripts are available, run the smallest demonstration after tests pass.
The goal is to see the radar path:
candidate trace
-> signal scan
-> anomaly / persistence / drift
-> measurement candidate
-> external validation
OMNIA-RADAR scans candidate traces, outputs, representations, or sequences for structural signals.
It can help detect:
- anomaly;
- persistence;
- drift;
- weak structural signal;
- candidate instability;
- candidate patterns worth measuring later.
Public compression:
RADAR detects.
OMNIA measures.
VALIDATION tests artifacts.
OMNIA-RADAR does not:
- infer semantic truth;
- decide correctness;
- replace OMNIA measurement;
- replace OMNIA-VALIDATION;
- perform security scanning;
- perform cryptographic attacks;
- recover keys;
- prove physical truth;
- convert detection into final decision.
The final decision remains external.
Not every trace deserves full measurement.
OMNIA-RADAR detects structural candidates.
OMNIA decides nothing; RADAR decides even less.
RADAR is an early-warning layer, not a verdict layer.
Every serious OMNIA-RADAR result should make clear:
| Component | Meaning |
|---|---|
| candidate | The trace, output, representation, or sequence being scanned |
| scan mode | What kind of signal scan is applied |
| anomaly | Whether unexpected structure appears |
| persistence | Whether signal survives basic variation |
| drift | Whether signal moves across observations |
| candidate signal | What should be sent to full measurement |
| limitation | What the detection does not prove |
| external validation | How the result should be validated later |
Recommended result vocabulary:
no_signal
candidate
anomaly
persistent
drifting
inconclusive
Meaning:
- no_signal: no relevant structural signal detected under the declared scan;
- candidate: a possible signal exists and should be measured;
- anomaly: unexpected structure appears;
- persistent: signal remains visible across declared checks;
- drifting: signal moves across observations;
- inconclusive: the scan is insufficient or ambiguous.
- docs/QUICKSTART_RADAR.md
- docs/RADAR_OVERVIEW.md
- docs/SIGNAL_CONTRACT.md
- docs/ANOMALY_PERSISTENCE_DRIFT.md
- docs/DETECTION_BOUNDARY.md
- docs/RADAR_MANIFEST.json
For the full ecosystem map, start here:
For public validation artifacts, start here:
For core structural measurement, start here:
not a truth oracle
not a semantic judge
Decision remains external
structure != meaning
| Repository | Role |
|---|---|
| lon-mirror | Canonical public entry point |
| OMNIA-VALIDATION | Public validation showroom |
| OMNIA | Core structural measurement engine |
| OMNIABASE | Representation invariance foundation |
| omnia-limit | Stop / continue boundary layer |
| OMNIA-RADAR | Structural signal detection layer |
| OMNIA-INVARIANCE | Structural invariance layer |
| OMNIA-CONSTANT | Structural constant candidate layer |
| OMNIAMIND | Structural cognition orchestration layer |
| OMNIA-THREE-BODY | Dynamic divergence stress test |
| OMNIA-SECURITY | Bounded structural security diagnostics |
| OMNIA-CRYPTO | Bounded structural crypto diagnostics |
measurement != inference != decision
MIT.
measurement != inference != decision
This repository is part of the MB-X.01 / OMNIA ecosystem. Its outputs must be read as structural measurement, validation, detection, orchestration or adapter artifacts according to the repository role. They are not autonomous semantic truth claims and they do not make external decisions.
This repository is one bounded measurement role inside Structural Observability.
Role:
stream drift, anomaly, and alert layer
Boundary:
Alert means structural deviation inside supplied streams. It is not prediction or final danger judgment.
Structural Observability foundation:
- lon-mirror: https://github.com/Tuttotorna/lon-mirror
- Foundation release: https://github.com/Tuttotorna/lon-mirror/releases/tag/v0.2.2
- DOI: https://doi.org/10.5281/zenodo.20379374
Role document: