DisclosureOS documentation · v2 Beta
Run the showcase
Start the example viewer locally, check its files and trace a displayed value to its source.
Run the DisclosureOS example case on your own computer, with the same viewers used by the portal. The download includes the application source, standard documents and retained source files. Everything in the case is fictional; it is separate from the ELDÆON dataset.
You need Node.js 22 and npm. Running the viewer does not require the portal repository, an account or unpublished DisclosureOS packages. Internet access is needed to install the application's dependencies.
Run the viewer
Download the complete example and unzip it. The extracted folder is named meridian, the case's original working name. Keep its contents together: the viewer reads source packets beside the viewer/ folder.
From the folder containing meridian, run:
cd meridian/viewer
npm install
npm run check:bundle
npm run devOpen localhost:3000/showcase, or the local address printed by the server if that port is busy. You should see the Casebook entrance. Choose Recordings, then Explore the recording, to reach the research desk.
The source files are served by your local app. Standard, documentation and partner links lead to the public portal. The standalone shell uses system fonts; the research viewers share the portal's implementation.
To check a production build, stop the development server with Ctrl+C, then run npm run build followed by npm start from the same folder.
Read the check result
npm run check:bundle prepares local copies of the retained files, verifies those copies and checks the source library's file lengths and SHA-256 fingerprints. It then reproduces one fixed laboratory calculation. Both checks must pass for the command to exit successfully.
The first result includes these status fields, along with file and collection counts:
{
"sourceIntegrity": "passed",
"canonicalValidation": "not_run",
"scientificValidation": "not_checked"
}The second result reports reproduced: true: three authored copper values, 1.21, 1.25 and 1.29 mg/kg, produce a mean of 1.25 mg/kg. Its uncertainty remains not_calculated. The calculation does not derive the laboratory report's separately declared uncertainty.
| Check | What it establishes |
|---|---|
| File delivery | The prepared copies match the retained files, and indexed sources match their declared fingerprints. |
| Fixed calculation | The included method reproduces the saved expected result from its retained inputs. |
| Standard validation | Not run by this command. Each source packet records its own validator, inputs and source revision. |
| Scientific validity | Not established by matching bytes, rendering a chart or reproducing arithmetic. |
Try a changed file, then recover
Stop the viewer first. From meridian/viewer, add a newline to the prepared copy of the optical CSV. This exercise leaves the retained original intact.
node --input-type=module <<'JS'
import { appendFileSync } from 'node:fs';
appendFileSync('public/examples/meridian/optical.csv', '\n');
JS
node scripts/check-bundle.mjsThe direct check exits with status 1 and reports Prepared source differs: optical.csv. It checks the files already prepared, without copying them again.
Recover with:
npm run check:bundleThe preparation step restores this changed copy from the retained original, and both checks pass again. If you alter an indexed retained file instead, copying it cannot repair its fingerprint mismatch. Use a fresh download to restore that original.
Follow a reading
Open Recordings → Original scalar readings at the research desk. Choose Optical and set the time to 21:01:30 UTC, or open that selection locally.
The displayed value is 39.92°. Use Inspect source row to connect it to the files at the root of meridian/:
| Layer | Where to look |
|---|---|
| Source bytes | optical.csv, row 11 including the header: 2026-08-14T21:01:30.000Z,39.92,deg. |
| Recorded measurement | observation.json, measurement optical-9: 39.92 deg, with declared standard uncertainty 0.2 deg. |
| Measurement connection | measurements.json names the product, optical channel and exact UTC capture time. |
| Capture context | bindings.json connects the observation product to the capture in acquisition.json. |
| Display | view.json derives the application's reading from those documents. The viewer consumes the same projection in meridian.generated.json. |
The three scalar channels retain their own units and samples at ten-second intervals. The display does not smooth them or establish a common physical source. The calibration declarations are fictional and unreviewed.
For a smaller exercise focused on these connections, use Check a measured value.
How the viewer is built
Standard documents hold the research data. A projection makes selected parts convenient to display. Components render that projection and inspect retained source bytes. The projection and URL parameters belong to the application; they are not additional standard contracts.
| Inside the download | Purpose |
|---|---|
observation.json, acquisition.json, bindings.json and related documents | The original field record and its declared connections. |
fixture.mjs and project.mjs | Author the original example and derive its application view. |
viewer/src/lib/showcase/ | Application models, generated projections and source readers. |
viewer/src/components/showcase/ and viewer/src/mission/ | Research-desk components and shared recording visualizers. |
recordings-v1/, reader-r2/, specimens-r2/ and other source folders | Additional recordings, documents, materials and reviews, each with its own source notes. |
viewer/scripts/prepare.mjs | Copies the adjacent source packets into the app's public directory. |
Start by changing presentation code in viewer/ while preserving its data inputs. For your own recording, begin with one retained source file and one traceable measurement, using Equipment and captures and Connect measurements. Then adapt a suitable renderer from the support guide.
The downloaded app is a fixed reference implementation. It does not import arbitrary datasets. A new integration needs its own source adapter, standard documents and checked projection. Keep displayed values derived from those sources, and retain unavailable information rather than copying the fixture's fictional declarations.
Regenerate the original projection
regenerate.mjs validates the authored fixture, writes the original source files and standard documents, then updates view.json, validation.json and the viewer's meridian.generated.json. From meridian/, after building the pinned checkout described in the next section:
node regenerate.mjs /path/to/disclosureos-sourceThis writes files; keep a separate working copy when experimenting. It regenerates the original packet only. Later packets and the source library retain fingerprints of their inputs. Changing those inputs requires rebuilding the affected dependent packets and projections too; running this one script does not update the entire showcase. Restart the viewer after preparing the changed files.
Validate a source packet
The download contains several independently versioned packets. Read the selected folder's README.md and validation.json before rerunning its checks. Use the exact standard-source commit named there, including any sibling packets the builder needs. Installing the published beta or using the latest source revision is not a substitute for that pin.
Run the original field packet's eight checks
Use Node.js 22, Git and Corepack with pnpm. From the folder containing meridian, prepare a separate standard-source checkout:
git clone https://github.com/disclosurefoundation/disclosureos.git disclosureos-source
cd disclosureos-source
git checkout c894a086e8326f048ba1806e7ba9d07556f16489
corepack pnpm install --frozen-lockfile
corepack pnpm build
cd ..
node meridian/verify.mjs disclosureos-sourceThe verifier reads the original downloaded documents and sources, compares them with the authored fixture and runs eight checks: observation, context, entities, history, acquisition, case, research history and measurement bindings. It also checks timestamps, units and values for all 72 scalar samples. An altered or missing original file fails this exercise.
Compare the output with meridian/validation.json. All eight results have success: true, while unchecked scientific and profile stages remain explicit. This receipt does not cover every later packet in the download. For mixed recordings, for example, use the separate recording-bindings-v1 verifier and its own pinned checkout.
Implementation notes
Choose the part you are adapting. These notes retain the source relationships, controls, assumptions and reproduction commands for each workspace. Folder names identify packet revisions, not separate npm releases.
Navigation and context
Casebook and research desk
The showcase opens with four Casebook chapters: the recording, the accounts, the sample and the review. Each chapter opens an existing viewer at the research desk. Browse the full collection skips the introduction and opens the collection catalog. The desk groups workspaces into Recordings, Accounts, Documents, Materials, Places & history, and Reviews & questions. On phones, use browse collections to switch groups.
The desk's local search finds workspace names and descriptions. Source library searches retained source text and metadata. Build with it connects the running example to its download, components and documentation. These are different tasks, so the interface labels them separately.
casebook.tsx, research-desk.tsx and desk-navigation.ts own this editorial
navigation. surface, chapter, deskMode and deskGroup are application URL
state, not standard fields. Existing links with a selected viewer still open that
viewer directly. Returning to the Casebook preserves source selections. The
chapter illustrations are editorial artwork; exact recordings and source files
remain in their respective viewers.
Inside a record
Detail pages use controls suited to their content. An account opens in a reader, with a compact account chooser, text/audio/sketch presentation and a separate comparison task. The specimen page leads with the selected portion, its own results and handling; the collection tree is available on demand and the gallery uses image thumbnails. The field recording keeps its view chooser alongside the sticky playback controls. These controls retain the existing source selections and URL fields. The other recording families, documents and reviews use their own detail controls within the same desk shell.
record-picker.tsx supplies a keyboard-accessible peer chooser. It closes after
selection, outside pointer interaction or Escape; Escape returns focus to the
trigger. The collection tree remains a native disclosure. Collection navigation
belongs to the desk, rather than being repeated above every material viewer.
Collection and availability
Places & history → Collection & availability connects
the retained collection sections to their workspaces and indexed source files.
collection-inventory.tsx loads the fingerprinted collection-v1/index.json
projection on demand. Its builder checks every library file, observation owner,
and acquisition/product join. The file detail verifies selected bytes again.
Browse a section or filter included files by content, declared product kind and explicit deployment site. Metadata search includes instrument references and manifest channel quantities. A shared source file can retain several products: the weather CSV keeps both meters. Identical bytes in separate collections retain their own observation context. Derived products without an acquisition do not inherit a site, time or instrument from upstream inputs.
Inclusion is a delivery fact, not a completeness or renderer-support claim. A raw product in Meridian is still synthetic. The availability view connects exact half-open detector coverage gaps to their count windows, keeps missing exposure separate from zero counts, and quotes the retained format-example declarations for absent, unsupported and altered sources. It also exposes public-only access declarations without supplying omitted source payloads. Unknown outage causes remain unknown; a runtime verification failure does not rewrite source access.
Collection sections are editorial navigation over existing standard documents. They add no canonical case membership; the existing membership declarations remain inspectable in the timeline. Source inspection retains observation IDs, source/product declarations, acquisition context and available edition citations. The same projection and component ship in the standalone download.
URL state uses overviewMode=collection, collectionMode=sections|files|limits,
collectionGroup, collectionQuery, collectionType, collectionKind,
collectionSite, collectionItem and collectionLimit. Back, reload and library
return preserve these selections. Filters do not claim to search numerical values
or decode native formats. File results are rendered in batches of 20.
Site and conditions
Open Places & history → The local site to connect the local plan, accounts and measured station conditions.
The site projection is generated from existing, hash-checked sourcebook files. It retains source records, observation identities and available canonical edition citations from the library. Its six coordinates agree with the two station rows in the spatial packet. Layer selection groups equipment, viewpoints and next-day collection sites; it does not register any independent capture or locate the light.
The plan uses equal east/north scales. All declared up coordinates are zero; the account's lower path, ridge and branch remain narrative descriptions. The sky viewer's ideal flat horizon is a separate camera-display assumption. There is no surveyed terrain or geodetic transform to overlay on a real map.
Expandable passages preserve exact source text and file line numbers. File lines are application reading aids, not newly invented canonical locators. Existing account edition/page citations remain available in the inspector and source library.
The weather table uses weatherSamples and the canonical product-source reader.
Each selected row retains its layout, measurement identity, instrument, unit and
timestamp. At a field cursor between samples, the preceding ten-second sample
is held with its age displayed; it is not interpolated. Both humidity meters stay
separate. Remembered wind, cloud, rain and visibility do not acquire sensor values.
Shared URLs retain overviewMode=site, siteLayer, sitePoint and the field
cursor. Source-library return preserves these selections. Direct record links
select the appropriate account, specimen or recording workspace. The public
download includes the same renderer, builder and retained sources.
Case timeline
Open Places & history → Case chronology to follow 26 selected activities across the recording, written accounts, specimen work, report editions, reviews and case grouping. The display is an application projection of existing canonical fields, not a new chronology contract.
Entity assertions supply account recordedTime, edition issuedAt, specimen
collectionTime, trace recordedTime and preparation/analysis performedAt.
Reviews use evaluatedAt; the case grouping uses recordedAt. These meanings
remain distinct. The field window comes from the observation's eventTime
interval. Artifact generation timestamps are not substituted for source dates.
Day-only declarations remain literal dates, without conversion to midnight UTC. Activities are grouped by day, not scaled onto the four-minute recording. Placement within a day does not assert an order for records without a time. Unknown or redacted dates remain unplaced. The bounded builder refuses multiple temporal assertions or unsupported precision rather than choosing a value. Separate bench captures are not assigned to the field session.
Select an activity to inspect its exact canonical document, JSON pointer and
retained declaration. The input file must pass SHA-256 and length verification
before source inspection, the library handoff or contextual navigation is enabled.
The timeline projection itself is loaded lazily and hash checked; failure offers
retry without substituting a made-up date. overviewMode and timelineEvent
survive sharing, reload, Back and library return.
The grouping entry exposes the retained field and specimen observation references and their membership basis. This does not create corroboration, a common cause, a merged observation, or a new source record. Its onward link opens the existing directed material-to-field review. Other links open the exact account, edition, specimen or review, preserving the independent recording cursor.
The timeline source guide and builder ship with the standalone viewer. The original sourcebook chronology and all underlying standard documents and source bytes remain unchanged. Site, availability and investigation views connect to this timeline without merging their records.
Recordings and source selection
Explore the mixed recordings
The default workspace has four panels: a visible clip, a numerical thermal field, an RF waterfall and weather from two meters. All five instruments are fictional. Focus a panel, move the shared playhead, select a thermal pixel, rectangular region or radio bin, and open its source inspector. Each source retains its own cadence. Thermal and RF samples occur every two seconds; weather every ten seconds. The display holds the most recent sample and reports its sample time. Visible video has twelve frames per second.
The recording manifest declares source hashes, byte lengths, array layout, units, models and limitations. Thermal values come from little-endian float32 arrays, never from image colors. The RF array contains power per bin, not spectral density. Environmental CSV rows retain both meter identities. The visible clip is synthetic, lossy H.264 imagery with a separate frame/time/pointing table.
The viewer verifies numerical source bytes against the manifest before displaying them. The current field loads a 192 KB thermal chunk. Full-recording analysis verifies all twelve chunks (2.304 MB total) and caches that fixed source set. Inspect a pixel to obtain its zero-based frame, row, column and byte offset; inspect a radio bin for its frequency edges and exact array offset. The pixel or region history covers all 120 frames, from 0 through 238 seconds. Video inspection identifies the requested frame interval; use a decoder for exact frame extraction rather than treating browser playback as a scientific decoder guarantee.
The bundle includes recordings-v1/build.mjs, its scenario input and a source guide. Regenerate these sources with node build.mjs --write from that directory, using Node 22 and ffmpeg with libx264. Numerical outputs are deterministic; encoded MP4 bytes may vary across ffmpeg builds, so the manifest records the encoder version and arguments.
Validation scope: the original eight-check receipt covers the original record packet and scalar readings. The mixed recordings now have a separate canonical record packet and source-selection evaluation, described below. The historical recording manifest retains its original wording and bytes. The original artifacts retain their original bytes. No scientific calibration, thermal/visible registration or common physical cause is established by this simulation.
Thermal region history
Choose 5 × 5 region, then set the upper-left row and column and the width and height. The outlined rectangle stays in image coordinates across all frames; it does not track an object. Image clicks move its upper-left corner. A selection near an edge is clipped to the remaining rows and columns. Single pixel returns to a one-cell trace. The crosshair and upper readout always show the anchor cell; the history readout shows the region statistics.
Every selected cell is read with the existing canonical array_element selector. For each frame the viewer computes the equal-weight arithmetic mean, minimum and maximum in row-major order. It applies no smoothing, interpolation or rounding before calculation, and never omits a missing or non-finite cell. The line is the mean; the shaded band is spatial pixel spread, not measurement uncertainty. Physical calibration remains unknown.
The complete history appears only after all twelve source chunks pass SHA-256 and length checks. On failure the trace and export remain unavailable; retry preserves already verified sources. Changing the rectangle clears the previous result until the new calculation completes. The sample slider seeks the shared recording clock, including the final thermal sample at 238 seconds.
Inspect region & method exposes the exact per-cell canonical selectors for the selected frame, source product, chunk digest, unrounded statistics and method. Download calculated history (CSV) exports all 120 rows with region coordinates, source filenames, digests, local frame indices and method identifier meridian-viewer-thermal-region-v1. This is a view-derived calculation, not a new standard contract or an independent observation. The reproducible implementation is included at viewer/src/components/showcase/thermal-region.ts in the example bundle.
Shared URLs retain pixel (the upper-left cell), thermalWidth, thermalHeight and cursor. The color and history scales stay fixed at 18–24.5°C for this authored source. Two-camera thermal comparison is available under Cameras & UV bands. Native partner-format decoding remains later work.
Source-selection contract
The mixed record packet uses the standard's observation, acquisition-context, acquisition-bindings, measurement-bindings and product-source-selection 0.1.0 contracts. Five instruments connect to sixteen raw logical products and one derived viewing video. All 120 weather values have canonical measurements and capture-time bindings.
The source-selection document describes array encoding, axes, byte order, CSV row partitions and exact rational frame timing. Select layout:thermal-04 with array_element indices [5,20,33] to read the pixel at 90 seconds: byte offset 102532, value 23.93558120727539 Cel. Select layout:spectrum with indices [45,64] for the RF bin at the same time. These are canonical selectors, rather than viewer-specific addressing conventions.
Both environmental meters share one CSV file. Each has a separate product layout with an exact instrument partition. Selecting the wrong meter's row fails the profile. The video remains a derived product with generator, scenario and frame-table inputs; it is not relabeled as raw sensor imagery.
The mixed verification receipt checks 366 source selections: all weather measurements, a thermal pixel and RF bin across every sample, and first/middle/final video and metadata frames. Its source-selection checks pass. Full acquisition and measurement profiles remain incomplete because physical calibration and measured clock uncertainty are absent. No zero uncertainty is invented to obtain a pass. A video check validates bytes and the declared frame interval; media decoding remains not_checked.
The viewer consumes a projection of these canonical layouts and a byte-for-byte copy of the standard's dependency-free reader. index.json only maps product references to local filenames. Panel layout, colors and the held-sample playback policy stay in the application. The download includes the reader, JSON Schema, documents and verifier. These contracts and readers are available in Beta 6. To reproduce this packet from its generator, use the exact canonical commit in the packet README and run:
node /path/to/meridian/recording-bindings-v1/verify.mjs /path/to/disclosureosThis first profile supports contiguous numerical arrays, UTF-8 CSV records and constant-rate video declarations. Compressed scientific containers, variable-frame-rate video, strides and custom missing-value encodings require later profiles.
Spatial context
The Sky & surroundings workspace uses the additive spatial packet. Canonical observation, acquisition bindings and product-source selections connect four source instruments/references and three derived products. The selection document addresses every pointing, track, station and clock row plus the first and final all-sky frame. Validation passes 1,511 selections while physical acquisition profiles remain incomplete.
The 640×640 all-sky clip runs at six frames per second. Its pointing table and clip are derived products of the retained optical frame metadata, not independent observations. Its ideal equidistant projection shares the authored optical light trajectory; the lens model declares synthetic bearing and field-of-view assumptions. Neither image gives a distance.
Aircraft reports retain an independent relative clock and Cartesian frame. The source message IDs survive normalization to CSV; the 100–120-second gap is never interpolated. No geographical transform or alignment to the field clock is supplied, so the viewer has a separate aircraft scrubber and does not overlay the track on the sky. The association review explains this boundary. These are fictional normalized reports, not a native ADS-B decoder.
Station positions retain the sourcebook's local frame: North (0,0,0), South (80,-120,0) metres. No GNSS fix is claimed. The timing review distinguishes the clock-offset model from unknown physical uncertainty; it applies no correction.
Shared links preserve workspace=spatial, spatialPanel, cursor, aircraftSecond, skyAz and skyEl. The same workspace and source packet ship in the downloadable viewer.
Field signals
The Field signals workspace adds pressure waveform and acoustic spectrum, magnetic components and derived magnitude, and passive-radar map inspection. It uses the same source-selection contract and reader as the other workspaces.
- Source guide
- Generation and processing methods
- Observation and derivations
- Acquisition context
- Measurement bindings
- Source selections
- Validation receipt
The acoustic periodogram is calculated from the retained pressure samples using 800-sample periodic Hann windows at 100 Hz with a 400-sample hop. Exact inspected values remain in Pa²/Hz; the heatmap applies a logarithmic display scale. Magnetic magnitude is derived from the three retained components. Every magnetic component has its own canonical measurement and capture-time binding.
Radar maps are authored in Doppler Hz and bistatic excess path length km. No transmitter/receiver geometry, range/speed conversion or geographic track is inferred. Map maxima use a declared selection rule, not a probabilistic detector. The bounded IQ reference has an independent relative clock and is not an input to these maps. This example supplies synthetic float32 products, not a decoder for a provider's NDJSON, compressed or SDR formats.
Shared URLs retain workspace=signals, cursor, signalPanel, acousticBin,
magneticAxis, delayBin, dopplerBin and the independent iqSample. The field
cursor can select the final 100 Hz pressure sample at 239.99 seconds. Source
selection validates exact bytes and values; incomplete physical calibration and
clock characterization remain explicit in the acquisition checks.
Events and counts
Recordings → Events & counts contains additional captures within the Meridian research collection, including a Geiger counter and a muon-detector reference with independent, entirely synthetic event lists. These captures use an authored 15 August clock and retain their own playhead.
Select a ten-second interval in the count timeline, then a pulse-height bin to reveal contributing events. The source list preserves excluded muon-reference pulses when All pulse heights is selected. The histogram shows only events included by the declared rule. Selecting a histogram bin filters the event list; it does not change the interval's count or rate. An event opens its exact canonical table row, ID, timestamp, pulse height and authored coincidence flag.
- 20–30 seconds: a recorded interval with zero events and a zero rate.
- 40–60 seconds: no recording; count and rate are absent, and no histogram is shown.
- 80–90 seconds: five recorded seconds; the rate denominator is five, not ten.
All intervals and histogram bins are half-open. Pulse height is in uncalibrated ADC units. No dose, energy, detector efficiency, dead-time correction or independently verified muon identity is inferred.
The event packet contains observation and processing records, acquisition context, bindings, canonical table selections, methods and a validation receipt. All 106 source rows are checked. Derived count products reference retained event and coverage products; the viewer independently recalculates the counts and rates before displaying them. The histogram is a transient calculation over those same event rows.
The acquisition profile remains incomplete because physical calibration and timing characterization are absent. This example uses the existing CSV source-selection contract; it does not add a new event schema or implement native NDJSON decoding. The source packet and same renderer ship in the standalone ZIP. Shareable selections use workspace=bench, benchDetector, benchWindow and pulseBin; they preserve the field cursor independently.
Imaging bench
Open Cameras & UV bands for independent numerical UV/night-vision sequences, two thermal resolutions, a timestamped event camera, and three UV-band channels. These are compact synthetic reference products, separate from the Meridian field recording and detector-count bench.
The sticky playback control advances through 24 authored seconds. Select a source pixel with the image or keyboard row/column controls, follow its full-sequence trace, and inspect its canonical array_element locator. Capture settings resolve to their exact CSV row. Thermal A and B retain separate pixel coordinates; their common authored clock does not establish spatial registration or physical synchronization. Directly generated Celsius values are not a radiometric MJPEG conversion.
The event-camera view recomputes the retained accumulation from all x/y/polarity rows before rendering. Select a pixel to see its contributing row selectors; open individual events to inspect their exact source. Both polarity signs count positively. Second 12 deliberately demonstrates recorded zero events. Choose a 1, 2 or 4 second accumulation, with half-open bounds and the one-second event-rate history alongside it. Windows clip at the end of the retained recording; the rate uses actual exposure. Summed pixels retain each contributing canonical cell and event-row selector.
UV-band bars preserve UVA, UVB and UVC as separate uncalibrated ADC channels. There is no interpolation into a continuous spectrum or irradiance conversion. Fixed color and chart scales are presentation choices, not calibration.
The imaging source guide documents the bounded representations and remaining native-format limits. Observation, acquisition context, source selections, and the validation receipt are included in the same download as the renderer. Calibration and clock context remain explicitly incomplete.
Reference signals
The Recordings catalog includes Waveforms, Environment and Biometric references. Twelve compact synthetic examples extend the camera and detector benches.
Each device has its own local timeline and channel units. The sticky sample scan runs at ten samples per second for inspection, independently of capture cadence. Exact source selection, not screen pixels or interpolated chart values, drives the inspector. Shared links retain benchFamily, referenceDevice, referenceSample, referenceBin and referenceWindow; the Meridian field cursor remains separate.
- Ultrasonic microphone: 4,096 float32 samples at 384 kHz, displayed in 256-sample windows without decimation. Its retained 4,096-point DFT covers the entire 10.67 ms capture. A frequency-bin inspector exposes amplitude, transform method and every contributing canonical array selector. No audible derivative or pressure calibration is implied.
- Microbarometer: ten minutes of relative pressure, with one sample per second. This is a numerical pressure reference, not a MiniSEED decoder.
- Environment: magnetic flux density and RF power density, TVOC and equivalent CO₂, PM2.5, lightning times and distance estimates, and body-frame acceleration/angular rate. The gas example retains five unavailable readings. Traces stop at the gap. Equivalent CO₂ is not directly measured CO₂; lightning distance has no bearing; body-frame acceleration is not a trajectory.
- Biometric references: four fictional EEG channels, skin resistance, independent SpO₂/heart-rate scales, and left/right pupil diameters with a numerical eye schematic. No subject or clinical interpretation is present. No EDF/BDF or eye-video decoding is claimed.
- Random-bit reference: a deterministic 256-bit sequence, exact bit selection, 32/64/128/256-bit windows, sample count and fraction of ones. The count inspector exposes each contributing standard selector. Counts demonstrate reproducibility, not true randomness or a significance test.
The source guide, methods and quantity definitions, observation, acquisition context, canonical source selections and validation receipt ship alongside the same renderer in the standalone ZIP. The spectrum's processing record links its retained waveform input. All 8,141 source selections validate against the existing contracts; physical calibration and clock characterization remain incomplete.
Instrument variants
Meridian is one fictional research collection. Its field study, imaging captures, detector sessions and other instrument activities share this presentation, while retaining their original source files, clocks and relationships. Collection membership does not assert a common object or physical cause. The original case.json continues to describe the original field-study record; this expanded collection is an application grouping, not a silent expansion of that case document.
The Survey instruments, Camera contexts and Other measurements families use the additive variants packet. Its original twenty instrument/artifact examples connect observation products to acquisition context and canonical source selections. Validation checks 1,517 retained selections. Structural, semantic and source-byte checks pass; physical calibration and clock characterization remain incomplete. All values are generated, and all models are declared in methods.json.
| Capture | Interaction and boundary |
|---|---|
| Original radar references | Select a detection, switch plan/profile, and inspect bearing, slant range, elevation and platform pose. Local east/north/up uses the declared transform. Each context has its own frame. Secondary replies and phased-array steering are authored metadata. |
| Original sonar references | Inspect authored return delays or a separate passive waveform. These initial references are distinct from the expanded sonar packets. |
| Lidar | Select a point in plan/profile; retain x/y/z and return intensity. No segmentation or terrestrial registration. |
| Satellite raster | Select a numerical pixel and inspect its geographic center and footprint edges. The declared WGS84 grid is not a GeoTIFF decoder or a real satellite acquisition. |
| Civilian, security, vehicle and phone cameras | Eight numerical frames per capture, with exact DN cells and separate per-frame exposure/motion settings. Phone orientation metadata remains visible; raw stored pixels are not silently rotated. |
| Gravity, seismic and optical spectrum | Inspect authored values and their source rows. Gravity differences, ground velocity and wavelength/intensity retain different units and meanings. No anomaly or chemical identification is inferred. |
| GNSS and clock | Select fix-quality metadata or clock offsets. Four absent GNSS fixes stay absent; no interpolation or cross-capture clock correction is applied. |
The expanded survey views use separate radar, active/passive sonar, lidar and satellite packets. Use their notes and profile guides for the current default visualizers; the table above describes the retained original variants.
The reference data reader is the same canonical implementation used throughout the showcase. Every table row is retained as a selector. Larger arrays retain a bounded set of example selectors, while the declared layout and reader support any in-bounds cell. Display geometry and pixel-center calculations are application methods over those exact source values.
Shared links preserve benchFamily, variantDevice, variantSample, variantPixel, variantProjection, variantRange and formatExample. They do not change the field recording cursor.
Expanded radar and sonar
The survey workspace includes seven radar scenarios, an active-sonar echo field and a passive-sonar analysis. These use their own source packets, rather than adding detail to the original variants-v1 files.
| Viewer | Source packets | Implementation guide |
|---|---|---|
| Scanning, ground, civilian, military, airborne, shipborne and secondary radar | The corresponding radar-…-v1/ directories | Radar |
| Active sonar | sonar-active-v1/ | Active sonar |
| Passive sonar | sonar-passive-v1/ | Passive sonar |
Each packet's README identifies its source revision, processing assumptions and reproduction command. The guides explain what each profile checks and include runnable examples. Use these declared models when adapting geometry, returns, detections or tracks; a display label alone does not supply platform motion, sound speed or an identification.
The passive sonar example shares its interactive frequency map, selected spectrum and source waveform with the downloadable viewer. Its optional passive sonar profile declares the analysis and its source windows.
Lidar survey
The lidar example shares its linked spatial views, pulse register and source inspector with the showcase and downloadable viewer. The optional lidar profile declares the point coordinates, return groups, intensity convention and authored surface labels.
The lidar-v1 packet retains both the pulse register and point table. A pulse with no detected return remains distinct from a pulse whose result is unavailable. Each selected point keeps its exact source row across view changes and onward navigation. A separate point-only example, lidar-legacy, contains no pulse register.
Multiband satellite raster
The satellite example shares its multiband map, selected-cell inspector and quality controls with the showcase and downloadable viewer. The optional raster profile declares six bands, scaling, missing data and an authored UTM grid.
The satellite-v1 packet retains both the band cube and quality grid. Natural color, infrared, individual-band and NDVI views inspect the same cell. Cloud and shadow values stay available for inspection but remain excluded from the derived index. A separate brightness-only example, satellite-legacy, contains no band or quality grid.
Format acceptance
Format acceptance exercises five fixed source states using actual byte verification:
- Declared CSV: the fingerprint and table layout pass, then an exact row is readable.
- Declared numerical array: the fingerprint and array layout pass, then an exact cell is readable.
- Unrecognized format: the artifact fingerprint passes, but no sample layout or renderer is supplied. Download remains available.
- Changed source bytes: an intentional test copy fails the original fingerprint. No chart or accepted-sample action appears.
- Source not supplied: no fetch or numerical result is invented.
This is a reproducible acceptance demonstration, not an arbitrary-file importer. ASTERIX, GeoTIFF, LAS/LAZ, MiniSEED, EDF/BDF and provider-specific compressed formats require their own adapters. A vocabulary token describes a sensor or method; it does not certify a native decoder. Existing CSV, float32 and media paths retain their own documented scope.
Accounts, documents and materials
Trace the other relationships
| Showcase interaction | Standard documents | Technical reference |
|---|---|---|
| Compare the two reported shapes | context.json, entities.json, observation.json | Context, Testimony |
| Switch editions and inspect a cited page | entities.json, history.json, report-a.txt, report-b.txt | Documents |
| Select a sample and inspect its results | specimens-r2/entities.json, specimens-r2/observation.json, collection/handling/preparation/laboratory source files | Physical samples |
| Compare current and earlier reviews | history.json | Claims |
| Inspect the case grouping | case.json, observation.json | Records |
The research claim history and instrument evaluator use distinct history contracts. instrument-history.json is an explicit, empty history for the instrument checks; both histories reference the same observation snapshot. It does not discard or convert the research assessments.
The field report text uses form-feed characters to delimit pages. Review citations bind to an exact edition, artifact digest and page. In the earlier short packet, laboratory results bind to portion A. The expanded sourcebook binds its analysis to A1, with independent field-control and water-blank results. Select sourceRevision=original to inspect the earlier packet; those historical records have not been rewritten.
Read expanded reports and accounts
The document reader defaults to the additive
reader-r2 sourcebook. Report A contains
three substantial logical pages; B contains four. The text reader supports
contents, per-edition search, independent A/B page selection, figures, tables and
linked source files. The account reader presents two
first accounts and Observer A's dated follow-up, with two explicitly authored
illustrations. Three accounts do not become three witnesses.
This packet uses existing observation, archival research-entities 0.2.0 and
archival claim-history 0.4.0 contracts at canonical commit
88bcbfcc8b17af30837313aff596cec8e1bde43a. Its
entities identify witnesses,
accounts, source editions, creation/release events and digital artifacts. Its
source assertions use canonical
edition citations: entity-document snapshot, edition ID, artifact fingerprint
and logical page. An inspected paragraph additionally reports application line
positions and exact excerpt text. Those positions do not introduce a new
normative selector.
The source files are Markdown with explicit page boundaries, not PDF facsimiles.
The browser checks hashes and byte lengths before rendering text or illustrations.
The receipt separates reference
validation from source authenticity and scientific validity. From the downloaded
reader-r2 folder, with Node 22 and the built canonical checkout:
node build.mjs /path/to/disclosureosThe original short packet and its historical assessment citations remain intact.
sourceRevision=original opens that earlier reader; it must not be interpreted as
a citation to the expanded sourcebook. The richer specimen sources can be read
through report links, but their later analytical portion A1 does not replace the
original material workspace's portion A. PDF derivatives, synthetic account audio and the richer specimen workspace ship as separate packets; their source notes appear below.
Report facsimiles and file lineage
In Documents, Page images shows a generated PDF page for each authored report page. Edition A keeps three pages and edition B keeps four. Use thumbnails, width fitting or 150%/200% display zoom; compare independent pages in both editions; download the selectable-text PDF; or return to the original Source text view to search and inspect passages. The presentation, page selections and zoom survive shared URLs.
These are generated layout derivatives, not scans of acquired documents. The PDF has its own source edition, fingerprint and page citations. Each displayed page also retains an explicit mapping to the original Markdown citation. Equal page numbers do not make those two artifact identities interchangeable. Earlier assessment citations are unchanged.
Open How this page was made to follow authored Markdown → PDF → page image → thumbnail, with extracted PDF text as a separate branch from the PDF. Every file has its own hash and declared processing inputs. The conversion keeps source text unchanged while formatting emphasis, retaining local-link labels and normalizing typographic dashes/quotes. Figure geometry is retained with an explicitly mapped embedded font. PDF container metadata is newly authored and uses a fixed date for deterministic reproduction; that date is neither the report issue date nor a physical acquisition time. Declared creation/storage actions are illustrative, not independently authenticated custody.
The source guide, conversion method, artifact inventory, archival entities, PDF page citations and validation receipt ship with both the portal and standalone viewer. Existing observation processing and archival contracts represent the relationships. No PDF renderer dependency is required in the browser.
From the downloaded facsimiles-v1/ folder, retaining its sibling reader-r2/:
node build.mjs /path/to/canonical-checkout
python -m pip install reportlab==4.4.9 svglib==1.6.0 pypdf==6.10.0
python render.py ../reader-r2 . --checkThe Node build verifies pinned assets and canonical relationships. The Python check reproduces PDF bytes and checks every prose block, table cell and figure caption on its declared page. Page PNGs are retained Poppler derivatives with independent hashes; byte-identical rasterization across platforms is not promised. Source failures withhold the affected page image or verified PDF download and preserve a text fallback. Timed audio, sketch annotations, source history and the source library retain separate source packets, described in the notes below.
Account media and sketch annotations
The Accounts section offers Written text, Listen, and Explore sketches. Observer A’s first account has a complete synthetic reading, with eight paragraph intervals and a downloadable WAV and WebVTT transcript. Selecting a passage seeks to it; playing that passage stops at its end. Continue reading resumes uninterrupted playback. The selected paragraph is shareable, and opening a link never starts playback automatically.
This is a generated reading of retained fictional text, not an interview or an original voice recording. Audio time starts with the reading and is separate from the field-session timeline. Observer B and the follow-up remain written sources. The packet retains the same two witnesses and three account records.
Both first-account illustrations have three selectable features and display zoom. A mark opens its exact written paragraph and source inspection. The 720 × 440 illustration coordinates are application annotations, not measurement selectors. These authored drawings do not establish physical shape, distance or trajectory.
The account media packet binds the synthetic audio to a separate source edition and canonical time_range citations. The original Markdown keeps its original digest and page citations. Observation processing records the text-to-audio and transcript derivatives. The method explains how cue boundaries come from PCM frame counts; the retained verifier checks byte identity, exact paragraph wording and timing bounds. Regenerating a platform voice is not promised to reproduce identical audio bytes.
node account-media-v1/build.mjs /path/to/canonicalThe portal and standalone viewer share these components. Damaged audio is withheld while the full text remains readable. Native audio controls, labeled feature buttons and the text view provide keyboard access. This is a bounded PCM/WAV example, not a general interview-import or transcription service. Qualification reviews, procedure examples, retelling highlights and witness-group variants are described under Account context and retellings.
Source context and release history
In Documents, Source history opens Source details, Addresses & checks, and Release history for either report or a separate administrative release example. Source details retain each field's attribution; inherited report assertions still point to the original text. Added catalog assertions point to the authored catalog. Display choices persist through shared URLs and browser Back.
The existing source_edition, external_identifier, identifier_check and document_event contracts represent this information. Each identifier names an exact source edition; each check names one identifier value assertion. Access and identity outcomes are displayed separately, with the latest dated result of each kind and the full earlier log. A reachable address can return the wrong edition. No check is different from an explicit unknown outcome. All addresses use the reserved .invalid domain and are never requested by the viewer. These are fictional recorded scenarios, not live network tests.
The public extract demonstrates a visible withheld-field placeholder. The earlier working original is unavailable; no hidden contact value is shipped. Creation, marking, review and release events are tied to this separate administrative edition. They do not change the reports' markings or establish authenticity, government action, complete custody or release permission.
The archival entities, exact extract citation, method, source guide and validation receipt ship in the download. From its root, retaining the sibling reader-r2/ snapshot:
node source-context-v1/build.mjs /path/to/canonicalBuild the pinned canonical checkout first, as described in the source guide. Validation checks byte identity, structure and references; it does not verify the truth of the authored check outcomes. The portal and standalone viewer share the same implementation.
Account context and retellings
Accounts → People & context opens dated source declarations, separate qualification reviews, oath/procedure examples, and witness groups. Observer roles refer to 14 August 2026; the reading reviews are dated 22 August. Unknown professional status, aviation experience or years of experience remain unknown. No review authenticates a person or assigns a credibility score.
Compare accounts now adds four topic selections for A/B and A/follow-up comparisons. Each highlighted span comes from one exact retained paragraph. Both sides retain their source date, file fingerprint, line and canonical edition/page citation. The annotations are authored reading aids rather than an automatic semantic diff or exhaustive list of changes. The complete accounts remain available below. Observer A's later exposure to the first report and plots is explicit.
The four procedure examples cover a procedure not administered, a reported pass, a reported fail and unknown administration/result. They use the existing procedure entity's polygraph declaration, with explicit links to their exercise account and witness. The examples also demonstrate public-name, pseudonym and withheld identities; exact, approximate, estimated and unknown date declarations; and supplied/unknown oath states. They belong to separate fictional collection exercises, not the North/South event. A procedure outcome is not a determination that an account is true.
The field group has two named pseudonyms and three linked accounts. Separate aggregate examples show approximate attendance without named membership and an unknown count. Those examples do not add witnesses to the field event, establish independence or turn a reported total into individual records.
The catalog, canonical entities, reviews and citation history, method, source guide and validation receipt ship in the standalone download. From its root, with the sibling reader-r2/ present:
node account-context-v1/build.mjs /path/to/canonicalUse the built canonical checkout pinned in the source guide. This reproduces the supplemental archival research-entity snapshot and claim history using existing contracts. Original reader assertions, dates and source bytes remain intact. The viewer verifies catalog and account source hashes before displaying context or highlights. URL parameters accountMode, accountContext, accountTopic, procedureExample and groupExample preserve selections in both the portal and downloaded app.
Source library and report comparison
Open Source library from any showcase view. The library indexes retained files across the collection, including the original reference, expanded reports, instruments, accounts, materials and archival exercises. Written Markdown/text sources are searchable in full. Other files are searchable by title, filename, collection, source description and fingerprint; numerical arrays, audio, video, images and PDFs are metadata-only entries. Search does not imply OCR or transcription. Report PDF text derivatives remain separately identified files.
Filter by source type, open a result, and select a page where an existing citation supplies one. The index and selected source must pass their SHA-256 and byte-length checks before source text or citation export is available. Copy citation includes readable attribution plus structured JSON; if clipboard access fails, a selectable text area and JSON download remain available. Read in case opens a report/account, while Explore collection takes other sources to their owning section. Return to preserves the originating view and selections. Search, source type, selected file and citation page also survive sharing or reload through library, libraryQuery, libraryType, librarySource and libraryPage.
library-v1/index.json is an application projection, not a new standard contract. Exports preserve available source/product declarations, the enclosing observation snapshot, existing canonical edition citations and the locations of their unchanged document snapshots. A file without an edition citation is exported as a whole-file reference, with no invented page or source identifier. Display excerpts retain their exact source text and line span as application annotations. Equal bytes in separate packets keep separate inventory entries and record context. The index includes declared source inventories and standard documents, not every build helper in the download.
In Documents → Compare editions, three selected comparisons follow the interpretation (A2/B3), account clarification (A2/B3) and specimen results (A3/B4). Each highlight is checked against its exact source paragraph during generation and opens its complete page. reportTopic preserves the selected comparison. Highlights are editorial reading aids rather than exhaustive diffs or new research claims. The retained short-example assessments continue to cite the original reports; expanded report annotations never rewrite those earlier inputs. The expanded assessment workspace below declares its own source snapshots.
The standalone viewer includes the same library, context packets and controls. The library builder checks all declared file hashes, resolves exported edition references against retained entity snapshots and reproduces the shipped index.
Follow a specimen
The default Materials view uses specimens-r2: seven materials, an uncollected trace, two preparation steps, six declared custody actions and five laboratory results. The parent M-014 is split into A and B; A is then divided into analytical A1 and a retained remainder. The field control and water blank are independent specimens.
The entity document uses the existing research-entities 0.4.0 contract. Results belong to exact analysis inputs. A1 results never appear as results of its parent or siblings. Copper values retain their standard uncertainties; lead and water-blank bounds remain below-limit reports, not zero. The water blank uses mg/L and is not plotted with the soil results in mg/kg. No enrichment or statistical significance is inferred from the comparison.
Nominal masses and volume are sourced observation measurements. Attributed source statements identify each specimen. index.json maps the material inventory row to its measurement for the application; this is not a new material-mass field in the standard. The amounts describe initial authored allocations, not current remaining balances. They are not summed across parent and descendants.
The collection gallery retains two AI-generated illustrations with exact prompts and hashes, a deterministic label drawing, and the authored local site plan. Artwork is not a photograph, calibration, weighing or compositional measurement. Gallery context always describes the parent at collection, not a selected child specimen. Source images and replicate rows are verified before display. The projection is generated only after canonical structural and relationship checks; the receipt does not authenticate custody or assess scientific adequacy.
From the downloaded specimens-r2/ directory, run node build.mjs /path/to/canonical-checkout after building its pinned checkout. This reproduces the documents and projection from retained sources. It retains image bytes rather than attempting deterministic AI image generation. The portal and standalone viewer use the same specimen workspace. The supplemental material-studies packet below adds microscopy-style numerical imagery, a modeled spectrum, mixtures, material classes and custody edge cases. Material PDF facsimiles, parent-gallery zoom and trace context are described below.
Material documents and collection views
The Materials collection workspace offers two-page Collection sheet and Laboratory report PDFs. Each reproduces the complete retained source text, with the laboratory result table intact. Switch pages, enlarge the page, download the verified PDF or open accessible source text. Document/page/zoom selection is shareable in the URL. Viewing a report while a parent or retained portion is selected does not assign its laboratory results to that portion.
The material document guide, conversion method, file inventory, archival entities, page citations and validation receipt retain the exact artifacts and relationships. The original source files remain unchanged. Each generated PDF has its own edition, separate from the original Markdown; PDF page numbers are not invented for that Markdown. Source issue dates remain separate from fixed PDF container dates.
The parent gallery supports fit, 2× and 4× enlargement with keyboard-scrollable image exploration. The T-014 trace includes a dedicated authored schematic and direct navigation to the collected soil and independent control. The trace detail is not to scale; the site inset uses the retained fictional local positions. Neither image supplies a physical measurement or establishes an age or cause.
From the downloaded material-views-v1/ directory, retaining sibling specimens-r2/, run node build.mjs /path/to/canonical-checkout. With ReportLab 4.4.9 and pypdf 6.10.0 installed, python render.py ../specimens-r2 . --check reproduces PDF bytes and checks every paragraph/table cell on its assigned page. PNG previews retain their own digests; no cross-platform pixel reproduction is claimed. The portal, source library and standalone viewer include the same files and controls.
Microscopy, spectra and material reference
The Materials view offers the original collection, Microscopy & spectra, and a Material reference. Both laboratory products explicitly describe prepared portion A1. The original specimen snapshot and historical results remain unchanged; belonging to the same specimen does not establish agreement between independently authored studies.
The material-studies source guide retains the generator, model and source sheets. Its research entities, attributed band labels, source selections and validation receipt use the existing canonical contracts.
- Microscopy-style field: a generated 96 × 96 float32 relative-intensity array. Select a feature or any pixel, move with arrow keys, zoom the display and inspect the exact array selector and byte offset. There is no acquired microscope image, physical scale, grain-size measurement or mineral identification.
- Spectrum: 201 retained CSV rows of relative response versus wavenumber. Select a row or an attributed band label and inspect its values, source and exact A1 analysis. The dashed reference is the generator's baseline model, not an acquired blank. Values are not calibrated absorbance or concentration; band labels do not assign chemicals. The image and spectrum are independent models.
- Material reference: all eight material types and four trace types have named examples. New specimens include collection, packaging, declared handling and source-authored qualitative review. The unused training insert has no bodily association; the biological example is fictional plant tissue. Uncollected traces have no specimen mass, custody chain or laboratory result.
- Mixture and custody: select either soil input and then the separately identified mixed output. The preparation names both inputs, while its nominal 2 g + 3 g allocations do not certify recovery or current balance. A separate debris record explicitly preserves unknown custody before receipt.
The product-to-analysis association is an application projection from the retained laboratory source report, not a new normative field. Existing research-entity contracts bind analyses and results to specimens; existing product-source-selection contracts bind displayed cells and rows to bytes. The receipt checks all 201 spectral rows and 100 retained image positions. Arbitrary browser pixel selections use the same canonical reader.
With the built canonical checkout pinned in the source guide and the sibling specimens-r2/ directory retained, run this inside the downloaded material-studies-v1/ directory:
node build.mjs /path/to/canonical-checkoutThis reproduces the numerical sources, standard documents and projection. The standalone viewer includes the same interactions.
Reviews and reproducibility
Expanded assessment reasoning
Assessments → Read a review presents three substantive, attributed reviews: the initial shared-source question, its revision, and an independent separate-sources explanation. Each exposes its exact rationale, comparison sources, limits and proposed follow-up. What changed compares the first two source sets and preserves the earlier wording. How it was reviewed explains the qualitative method and its limits.
The revised claim supersedes only the initial claim from the same review team on the same topic. The alternative has a separate author and topic. Current means current in this history, not correct or established. All three outcomes remain inconclusive; no probability, confidence score or authenticity finding is inferred.
The assessment packet retains the exact authored review sections from the expanded sourcebook, the exact report/account editions and the original scalar CSV files and generator. These dated reviews do not silently incorporate later instrument or laboratory examples. The original short-example history remains unchanged and is still available from the workspace.
The visible source-set comparison covers the records being compared. Each canonical claim additionally retains its method, the authored catalog of reading aids and a source assertion containing the verbatim review note. Inspect assessment exposes the complete claim. Contribution notes and text line ranges are application reading aids, not computed weights, support/challenge relationships or new normative source locators. Written sources retain existing edition/page citations; scalar files retain whole-file source references.
The history, method, source guide and validation receipt ship in the standalone download. From the download root, retaining reader-r2/ and the original sources/, fixture.mjs and history.json:
node assessments-v1/build.mjs /path/to/canonicalUse the built canonical checkout pinned in the source guide. Existing observation, archival entity and claim-history contracts validate the packet, including exact editions and same-evaluator supersession. The receipt checks structure, references and retained bytes; it does not establish scientific validity or source authenticity. The portal and standalone viewer lazily load the same hash-verified projection and verify a selected source before displaying it. Review, mode and source choices survive URL sharing and browser Back.
This section covers observation-level qualitative reasoning using contracts available in Beta 6. The following sections cover catalog exploration, precise review subjects and a reproducible calculation.
Observable and origin exploration
Assessments → Observables exposes the complete 12-entry technology and biologics catalog. Origin hypotheses exposes all 96 nodes of the proposed origin framework, with hierarchy navigation, search and category filters. Catalogs are retained from the pinned canonical source and checked against the same definitions used by the technical documentation. The twelve observable definitions and shared vocabulary APIs are available in Beta 6. Inspect or download the complete definition, including scientific explanations, signals, methods, references, aliases and any caveat.
The origin framework retains its attribution to Col. Karl Nell, Proposed Taxonomy of UAP Origin Hypotheses, Sol Foundation Symposium, Stanford University, November 2023. Its testability flags are framework declarations. Inclusion, including speculative entries, does not establish a test, scientific consensus or an explanation for Meridian.
Two observable terms and three origin nodes have fictional reviews dated 13 October. Each uses the existing canonical assessment contract with an inconclusive outcome, an attributed evaluator, a declared method and exact source inputs. The observable reviews distinguish angular motion from physical acceleration and examine whether differing accounts establish a comparable sensor-detection difference. The three origin questions concern optical/viewing effects, celestial or orbiting sources, and conventional human-operated sources. Compare three example questions displays the rationale, missing checks and proposed discriminating work side by side without a rank, probability or default winner.
Other entries show Not assessed, meaning no assessment is included for that entry in this packet. This is not an absence finding, rejection or failed documentation profile. A selected origin leaf does not confer an assessment on its parent, children or siblings. The material examples do not establish a biological specimen or a biologics finding.
The catalog, authored reviews, canonical history, method, source guide and validation receipt are included in the source library and standalone ZIP. These dated reviews use exact copies of the sourcebook and original scalar inputs retained in assessments-v1; later instrument and laboratory examples are separate captures. The vocabulary packet is updated directly to the revised V2 baseline.
The catalog-entry-to-claim association uses the standard’s shared vocabulary_bindings contract. Each binding identifies the exact catalog revision, catalog bytes, term and claim-history snapshot, with attribution for the association. The subject and supporting source inputs remain separate. The canonical evaluator resolves supplied documents and terms without fetching a catalog from the network. It does not authenticate sources, execute suggested tests, validate the catalog’s scientific statements or compute probabilities. See shared vocabulary references for descriptive associations and declared assessment criteria. Display contribution notes, limits and proposed follow-up remain authored reading aids.
From the downloaded root, retaining the sibling assessments-v1/ folder and using the built canonical checkout pinned in the source guide:
node vocabulary-v1/build.mjs /path/to/canonicalThe portal and standalone viewer verify the catalog projection and each opened source. Search, category, selected entry and origin-comparison state are retained through catalogQuery, catalogFilter, catalogEntry and catalogCompare; assessmentInput selects a review source. Corrupt source bytes withhold the source reader. The following sections cover precise review subjects and a reproducible calculation. The standard APIs used here are available in Beta 6; the viewer is supplied as source.
Precise subject reviews
Assessments → Specimens & relationships presents five fictional reviews: a lead-result field, the analytical portion A1 before and after a revision, the parent specimen's handling, and a directed material-to-field corroboration question. Each review shows its exact subject, attributed reasoning, limitations, next checks and declared inputs. Inspect exact subject opens the existing canonical reference and resolved field or record.
The field review targets only laboratory_result/lead-a1/result. It retains the below-limit declaration and threshold without inventing a numerical concentration. Specimen reviews use whole-entity subjects; results stay attached to the tested portion. Material-review contamination and representativeness declarations remain unknown. The initial and revised A1 claims share one subject, topic and evaluator; the revision adds preparation, result-table and repeated-value inputs while retaining the earlier report-only review. Both use the same specimen snapshot, so the comparison does not claim the other sources did not yet exist.
The case claim uses the standard's directed case_relation subject. It asks whether the material record corroborates the pinned field observation and remains inconclusive. Case membership is organizational: no relationship fact, reverse corroboration, observation merge or publication change is performed. The new case retains the original field observation and a separate specimen observation with the qualitative review method. Original case, specimen and history bytes remain under retained/.
The laboratory history uses laboratory claim-history 0.5.0 and entities 0.4.0. The case history uses case claim-history 0.1.0. subject, entityInputRefs, materialReview, caseInputRefs, observationInputRefs and supersedes are existing standard contracts. Display labels and resolved reading aids are application projections. Later microscopy and spectral examples are outside these declared input sets.
The authored reviews, method, source guide and validation receipt ship in the source library and standalone download. The builder verifies manifest-listed source bytes and invokes both canonical evaluators for reference and revision checks. Other artifacts mentioned by a retained observation are outside this packet's byte checks. Scientific validity, physical custody and authenticity are not checked.
From the download root, retaining specimens-r2/ and the original field observation, case and station log, use the built canonical checkout pinned in the source guide:
node subject-reviews-v1/build.mjs /path/to/canonicalThe portal and standalone viewer verify the projection and each selected source before exposing its text or download. assessmentMode=subjects, subjectReview, subjectCompare and assessmentInput preserve the selected review, revision comparison and input through shared URLs and Back.
Gaps and disagreements
The Assessments workspace includes nine fictional companion observations for missing, approximate, conflicting and restricted values. They use existing observation envelopes, source assertions, selections and access declarations. They do not revise the original Meridian field session.
Unknown values retain the reasons not_recorded, not_collected and unavailable. An approximate date keeps its calendar precision. The conflicting-time example preserves both attributed declarations without selecting or averaging them. A separate indexing example shows an explicit selection with its method, evaluator, rationale and considered assertions. Selection does not establish clock accuracy.
The restricted, withheld and unknown-access examples export a redacted position and only a source identifier, kind and access state. The bounded exporter does not copy internal coordinates, source URLs, titles, hashes or assertions. No underlying location register is included in the page, library or ZIP. The downloadable public JSON contains only the declaration. This example is not a general sanitizer or an authorization service; a production integration needs a field-specific public export policy.
The view also links to existing exposure-aware counts, below-limit laboratory results, unsupported files and integrity failures. These remain distinct from unknown values. Integrity failures display an error and no invented data; source-access labels alone never grant access.
Use assessmentMode=states&recordState=conflict on the showcase route to share a selected example. Unknown selections fall back to the first example. The public states-v1 packet includes exact companion records, timing notes and a canonical validation receipt. Its inventory observation groups the files without combining their event times. Each indexed companion retains its own observation identity.
node states-v1/build.mjs /path/to/pinned-disclosureos states-v1The pinned standard checkout and build requirements are the same as the verification packet below. Public source notes describe the examples in states-v1/README.md.
Investigation and replies
Open Reviews & questions → Investigation & replies to follow three authored fictional questions: recording alignment, specimen scope and a request for additional capture metadata. Two have custodian replies and inconclusive reviewer assessments; the third retains no reply and stays unassessed. Proposed next actions are not recorded as performed work.
investigation-v1/ uses the existing case-record and case-claim-history 0.1.0
contracts at pinned canonical commit 88bcbfcc8b17af30837313aff596cec8e1bde43a.
Each request is a source_assertion targeting an investigation; each reply is a
response_event; each review is a separate assessment with exact reply assertion,
source and method references. There is no canonical request-response edge in these
contracts. The question grouping is explicitly editorial navigation.
Pinned questions.json and method.md generate the fictional correspondence.txt
and review-log.md. The packet retains the earlier field, specimen and case bytes,
adds a separate observation containing workflow sources, and emits a new case
snapshot. Earlier membership, reviews and measurements remain unchanged. The
source observation is not a new physical capture or an additional field member.
Message dates have day precision; the assigned packet recordedAt instant is not a
send or receipt time. No real person or organization was contacted.
The lazy-loaded investigation-history.tsx verifies the projection, case, history,
correspondence, review log, method and selected retained review before showing the
thread or its actions. Sources can be inspected or opened in the searchable
library. overviewMode=investigation and investigationQuestion=timing|specimen|capture
survive reload, Back and library return. The case timeline links the dated requests,
replies and assessed reviews into the same question view.
The packet guide describes reproduction. Canonical evaluation checks contract and reference consistency. It does not verify authenticity, scientific interpretation, completion of proposed work or the existence of additional data outside this fictional packet.
Verification and reproducible calculation
Assessments → Checks & calculation separates saved standard-validation receipts from a calculation the reader can run. Recorded checks presents three runs: the precise specimen reviews, the material-to-field relationship history, and a new calculation observation. Each shows the evaluator's structural, semantic, external and profile statuses, its exact checked snapshots and the pinned canonical implementation. External means resolution of supplied cross-record snapshots, not outside scientific review. No documentation profile or score policy was applied; no overall scientific grade is displayed.
The specimen and case evaluators rerun against byte-identical snapshots from subject-reviews-v1. The calculation record uses existing observation methods, processing and derived products contracts. The artifact's specimen label and arithmetic result are application content, not a new standard measurement or normative cross-observation binding. The builder compares its mean and units to the existing copper-a1 measurement without rewriting that measurement or its separately declared uncertainty.
Reproduce a result uses exactly three authored values for analytical portion A1: 1.21, 1.25 and 1.29 mg/kg. The fixed method sums decimal hundredths as integers before dividing by 100 and three, producing 1.25 mg/kg. It rejects missing values, other units, other specimens and unsupported table shapes. The original ±0.08 mg/kg standard uncertainty remains a source declaration; it is not calculated from these repeated values.
The browser bundles the same fixed method file included in the verification packet. It verifies the input CSV, retained method artifact and expected-output bytes before enabling reproduction, then compares the complete result with the saved expectation. It executes no uploaded code, remote script or user-supplied expression. The material page's mean uses the same function. A reload or shared link preserves the selected view and inputs, but never carries forward a browser success state. The button does not rerun the Node-based standard validators.
Download the complete example and run this from its root with Node 22; no dependency installation or canonical checkout is required for the bounded arithmetic check:
node verification-v1/reproduce.mjsThe source CSV, fixed implementation, expected output, method, calculation observation and receipt are retained in the source library and ZIP. To rerun canonical checks, use the pinned, built checkout and the sibling source packets described in the source guide:
node verification-v1/build.mjs /path/to/canonicalassessmentMode=verification, verificationMode, verificationRun and verificationInput retain the chosen view, check and input. The checks apply only to their listed records and manifest files. Artifacts merely referenced inside a record are not fetched. Successful reproduction is not source authentication, physical custody verification, laboratory validation or support for an unusual origin.
What this demonstrates
The example connects recordings, accounts, documents, material lineage and attributed reviews. It demonstrates selected standard capabilities, not every optional field. Its application grouping does not establish that every capture shares a subject, clock or physical cause.
Data and fictional text use CC0-1.0; example code uses the MIT license. The viewer is not an intake, redaction or publication service.
Continue with Renderers & support to choose a component and understand its format limits, or Prepare a dataset contribution for the real-data workflow.