PROVENANCE AND LICENSING STATUS

The browser implementation and independent synthetic test fixtures were newly
written. Protocol interpretation was informed by Texas Instruments mmWave
Applications / Industrial Visualizer files: gui_parser.py, parseFrame.py,
parseTLVs.py, tlv_defines.py, gui_common.py, demo_defines.py, and the
people_tracking.py and vital_signs.py demo modules.

The supplied source subset contained no license file or copyright/license
headers. This does not establish permission to redistribute or relicense any
Texas Instruments material or derivative works. This source release includes
no TI Python files, firmware, vendor configuration files, vendor images,
private input manifests, personal recordings, or private logs.

Original material the licensor has authority to license is covered by LICENSE,
the Radar Studio Chromebook Source-Available No-Resale License, added to main
on 2026-10-10. It permits personal/internal-business use, paid teaching and
free redistribution subject to its terms, and restricts resale and paid access.
It is source-available, not OSI-approved open source. Third-party rights remain
affected only by their own applicable terms, not by this project license.
No statement here resolves possible third-party rights in protocol-derived
work. Obtain and review the applicable TI package license for vendor material.

No third-party JS/CSS dependencies are downloaded by the app. No endorsement
by Texas Instruments, Google, Silicon Labs, or any medical authority is implied.

Version 0.3 also inspected the official Radar Toolbox 4.00.00.05 firmware C
transmitters, common TLV structs, gesture/ground-speed visualizer definitions
and demo guides. PROTOCOL_EVIDENCE.md and DEMO_CATALOG.md identify these files.
Only metadata and independently authored implementations/fixtures are included;
no TI code, firmware bytes or configuration contents were copied into this
release. Matching imported file hashes establishes artifact identity, not a
right to redistribute the imported material or hardware compatibility.

V0.4 adds first-party decoders informed by official Area/Doors, Level and CPD wire sources. Vendor code/firmware/config bytes remain excluded. See LEGACY_PROTOCOL_EVIDENCE.md. No hardware or real-browser verification is implied.

V0.5 ORIGINAL MOTION APPLICATION AND CAPTURE

The motion-signature feature extraction, calibration/ROI handling, labelled
session workflow, standardized nearest-centroid browser classifier, evaluation,
local dataset handling and UART recorder are newly written project components.
They do not copy a TI activity-classifier algorithm, trained activity model or
TI demo application implementation. This release still uses the existing
protocol decoders and compatible TI firmware output for sensor acquisition;
that dependency is not presented as original radar firmware.

No original on-board firmware build, custom BIN, raw-ADC acquisition pipeline
or full micro-Doppler pipeline is delivered here. A future original firmware
pipeline may use properly licensed TI low-level/SDK components after a separate
engineering and licensing review. The roadmap in MOTION_LAB.md is future work,
not a claim of implementation, compilation or hardware validation.

Motion calibration and activity labels are user-observed/measured experimental
inputs. Imported provenance is self-reported and untrusted; the import operation
cannot establish capture authenticity, physical calibration, session independence
or participant consent. Synthetic fixtures are original software-test data, not
real-human validation data. No real activity accuracy claim is made.

Local captures may contain personal activity and metadata. They are not bundled
as public example data or automatically uploaded. Captured-dataset license
status remains pending, and the software LICENSE does not independently confer
rights to publish participant records or to redistribute vendor material.
The exact project LICENSE and all earlier third-party exclusions remain intact.

V0.6 offline learning additions
Original project code: offline_features.js, offline_model.js, offline_review.js,
offline_ui.js, offline.html, tools/offline-ml.js and their synthetic tests. The
model is a deterministic CPU standardized nearest-centroid baseline learned from
user-approved labels; no TI classifier, third-party model weights, neural-network
service or copied activity-detection implementation is included. This software
release retains the existing TI-derived protocol interface, distinct from future
original sensing firmware. No captured user data or trained model is bundled.
