REAMKIRA - UPDATE LOG
Dates use Melbourne, Victoria time. New entries go above older entries.

2026-09-17 | v0.8 | Export format 1.6
Based on: ReamKira-v0.7.zip
Requested by: Leumas
Reason: Broader AI usefulness, richer JSON-LD summaries and visible version
tracking without filling the homepage with release-history text.

HOMEPAGE
- Terminal heading: REAMKIRA // DISTILLATION TERMINAL · v0.8.
- Introduction explains what is extracted, Markdown/JSON/KWP downloads,
  direct free GET/POST API calls and source-only extraction limits.
- API copy explains package, markdown, kwp and stats response fields.
- Current version panel shows v0.8, Melbourne update date and brief overview.
  Arrow expands updates.txt; text loads on demand with retry/fallback link.
  Log is inserted as plain text and can be downloaded separately.

EXTRACTION AND EXPORTS
1. Explicit declared canonical in Markdown/KWP and pageSignals.canonical.
   Final fetched URL stays separate; legacy canonical fallback remains.
2. Declared language from HTML lang/xml:lang, Content-Language header and
   structured-data inLanguage. No invented inferred language/confidence.
3. Best-effort page classification with evidence and inferred=true.
   Article, Documentation, Landing page, Product page, Forum thread,
   Profile and Unknown are supported. Unknown is a valid result.
4. Possible login/paywall flags with matched phrases. Navigation login
   links alone do not trigger flags; form text can supply access evidence.
5. Unique internal/external HTTP/HTTPS source-link counts. Internal means
   exact final hostname; subdomains count as external. Distinct fragments
   count separately. Counts remain available if link export is disabled.
6. Meta robots/crawler directives and final-response X-Robots-Tag.
   Repeated X-Robots-Tag headers are combined. robots.txt is not fetched.
7. Thin-content and possible soft-404 indicators with supporting evidence.
8. JSON-LD summaries traverse arrays, nested objects and @graph. Useful
   declared types/properties are surfaced, including software metadata.
   Summary is capped at 250 typed objects; full parsed data remains in JSON.
9. Presence is separate from validity. JSON-LD reports absent, empty,
   valid or invalid status. Parsing errors retain raw source and diagnostics.
10. Open Graph/Twitter now use key/value objects. Repeated values become
    arrays, including multiple images; empty maps serialize as {}.
11. Structured overview includes counts, type declarations, software-type
    detection and extraction notes. No unsupported main-entity inference.
12. Microdata retained; basic RDFa source attributes added when schema.org
    context exists. Both are explicitly partial. schemaOrgMicrodata alias
    remains for compatibility.
13. Markdown/KWP show compact schema summaries and error diagnostics.
    JSON retains full parsed JSON-LD inside the existing valid/data wrappers.
14. Prominent ReamKira/VOIDNEON producer credits remain in every format.
    Source attribution remains separate from producer attribution.

FILES CHANGED
- index.html
- [RESTRICTED]
- updates.txt
Other files are included unchanged in the complete package.

VALIDATION
- PHP parser syntax checks passed for distill.php and contact.php.
- PHP WebAssembly runtime executed representative extraction fixtures.
  Passed nested/@graph JSON-LD, full-data preservation, invalid/empty/absent
  status, repeated Open Graph values, canonical, language, classification,
  access evidence, link counts, indexing, RDFa and filename image labels.
- Fixtures also passed producer attribution, interactive grouping/table,
  disabled content/links selection, stable text hash and soft-404 warnings.
  Networking was deliberately stubbed in these tests.
- All inline JavaScript passed node --check.
- Release-panel tests passed error fallback, retry, one-time loading and
  plain-text insertion of untrusted log contents.
- HTML structure, unique IDs, release controls and version label verified.
- ZIP integrity and every packaged byte checked against staged source.
- Browser visual verification was unavailable because no browser binary
  was installed. Live-host HTTP extraction and mail delivery were not tested.

COMPATIBILITY AND LIMITS
JSON format advances to 1.6. Open Graph/Twitter arrays become objects;
callers consuming the old arrays must adapt. Repeated keys may hold arrays.
New pageSignals and structured-data summaries are additive. Existing URL,
GET/POST and content/style/structure/links/images arguments are retained.
Language is declared, not statistically detected. Classifications and access/
error-page flags are heuristics. RDFa is an attribute summary, not a resolved
semantic graph. Source markup can be missing, incorrect or misleading.
Scripts are not executed; image contents are not inspected. Schema summaries
are source declarations, not independent factual or schema validation.
The SHA-256 remains scoped to extracted main text, excluding report metadata.

2026-09-17 | v0.7 | Export format 1.5
Based on: ReamKira-v0.6.zip
Requested by: Leumas
Reason: Improve export readability and explicitly credit ReamKira/VOIDNEON.

CHANGES
1. Prominent export-producer section in Markdown with ReamKira URL,
   VOIDNEON URL, studio description/tagline, creator Leumas McCarthy,
   Digital Heretic, and contact leumas@voidneon.com. Footer repeats credits.
2. JSON producer object includes application/version/URL and studio details.
   Producer attribution remains separate from source website metadata.
3. KWP includes generator/author HTML metadata, visible attribution with
   clickable links, and the full producer section in the exported text.
4. Block boundaries preserve spaces/line breaks in content and link labels.
   Inline word fragments remain adjacent, avoiding damage to split words.
5. Disclosure labels prefer headings, title-bearing elements or a text line
   that is not a date/byline/action instruction; 80-character cap retained.
6. Missing button type attributes report declaredType: not specified.
7. Markdown interaction table groups identical complete control records
   with counts. JSON preserves individual records and their original order.
   KWP renders the interaction table as an HTML table.
8. Images retain original alt and add label/labelSource. Filename fallbacks
   are explicitly marked inferred; opaque filenames remain unlabelled.
9. Updated documentation; prior update-log entries preserved.

FILES CHANGED
- [RESTRICTED]
- updates.txt
Other files are included unchanged as a complete upload.

VALIDATION
PHP syntax passed php-parser for distill.php and contact.php; final distill.php
was parsed again after adding the HTML interaction table.
Export attribution, grouping, label logic and shared output paths reviewed.
ZIP integrity and all packaged bytes verified against staged source.
No PHP runtime or live-host extraction test was available locally.
Live fetching and mail delivery still require testing on the existing host.

LIMITS AND COMPATIBILITY
Text/hash values can differ from v0.6 due to improved whitespace extraction;
compare content hashes from the same extraction version.
Disclosure titles and filename labels are source-based heuristics.
No image contents are inspected. Missing source whitespace between inline
word fragments cannot always be distinguished from missing word boundaries.
Existing HTML-only/JS completeness and token-estimate limits remain.

2026-09-17 | v0.6 | Export format 1.4
Based on: ReamKira-v0.5.zip
Requested by: Leaumas
Reason: Refine disclosure labels and completeness explanations following
feedback on the v0.5 exports.

CHANGES
- Disclosure labels capped at 80 Unicode characters, shortened at a word
  boundary where practical, with an ellipsis and Expandable disclosure note.
  The main-content copy remains unshortened.
- Added extraction.reason to JSON and the completeness line in Markdown/KWP.
  Reasons reflect executable scripts, sparse text, truncation or embedded
  content detected in the source. Disclosures alone do not imply missing text.
- Tightened attribute matching so data-alt and data-property cannot be
  mistaken for alt and property. Metadata property names are trimmed.
- Repeated image URLs retain later nonblank alt text when the first was blank.
- Open Graph/Twitter extraction and empty structured-data sections retained.
  No claim is made that a particular live page contains those tags.
- Release/version documentation updated; previous log entries preserved.

FILES CHANGED FROM V0.5
- [RESTRICTED]
- updates.txt
Other files are included unchanged for a complete upload.

VALIDATION
- PHP syntax passed php-parser for distill.php and contact.php.
- Changed extraction and export fields reviewed for Markdown/JSON/KWP.
- ZIP integrity and every packaged file verified against staged source.
- No PHP execution or live hosting extraction was performed locally.
  PHP runtime installation was unavailable; live fetching remains to be
  tested on the existing host. No mail-delivery test was performed.

LIMITS
HTML-only extraction, approximate tokens and source-inferred interactions
retain the limitations documented in v0.5. Long disclosure labels shorten
only in the interactive summary. JavaScript-injected metadata is unavailable.

2026-09-17 | v0.5 | Export format 1.3
Based on: ReamKira-v0.4.zip
Requested by: Leumas
Reason: Improve AI readability and change tracking, following Grok feedback.

ADDED
1. Source-based content completeness: High (static source) or Partial.
   Reports extracted characters, truncation and likely missing content.
   Flags executable scripts, sparse text, embedded frames/canvas/templates.
   Does not claim a measured percentage of the rendered page captured.
2. Exact Unicode character and UTF-8 byte counts for extracted main text.
   Approximate token count = characters / 4, rounded up; clearly labelled.
3. Structured data: Open Graph, Twitter Cards, parsed JSON-LD, and source
   Schema.org microdata attributes. Invalid JSON-LD retains raw source and
   a parsing error. Microdata is an attribute summary, not a complete graph.
4. Interactive controls: forms, buttons, disclosures, action links and common
   input controls. Summaries use source labels and attributes. No controls
   are executed, and no user-entered field values are collected.
5. SHA-256 of extracted main-text bytes, excluding timestamps and metadata.
   HTTP Last-Modified from the final response, or Not supplied/null.
6. Explicit MAIN CONTENT STARTS HERE / MAIN CONTENT ENDS HERE markers.
7. Export introduction identifies website content as untrusted source material.
8. Results display character count, estimated tokens and completeness.
9. This dated updates.txt log, also supplied separately with the release.

OUTPUT AND COMPATIBILITY
Markdown, JSON and KWP exports include the new information.
GET and POST API responses use the same updated export builder.
Existing URL arguments and selection controls remain supported.
JSON adds extraction, contentSize, changeDetection, structuredData and
interactiveElements. stats adds characters, approximateTokens, completeness.
Size/hash describe extracted text even when content export is disabled.
UTF-8 truncation avoids splitting a character at the 160,000-byte text limit.

FILES CHANGED
- [RESTRICTED]: extraction, metadata, hash, counts and export sections.
- index.html: new result statistics.
- updates.txt: new release log.
Other package files are included unchanged for a complete upload.

VALIDATION
- PHP syntax parsed successfully for distill.php and contact.php using
  the tree-sitter PHP grammar.
- Inline JavaScript passed node --check.
- Export pathways reviewed: Markdown, JSON, HTML/KWP and GET/POST share
  the same package builder.
- ZIP contents verified against the staged package.
- PHP runtime was unavailable locally. Live HTTP extraction and mail delivery
  were not executed; test those on the existing PHP hosting server.

KNOWN LIMITS
Scripts are not executed. Labels suggest purpose but do not prove behaviour.
Token counts are estimates. Last-Modified is optional and publisher-provided.
The hash detects changes to extracted text, not styles, images or all metadata.
Microdata only reports source attributes; nested relationships/text values
are not reconstructed. Main text retains the previous extraction approach.

FUTURE DELIVERY CONVENTION
Include updates.txt in future builds and preserve existing entries.
Record date, version, base package, requested changes, changed files,
validation performed, installation notes and relevant limitations.
Do not backfill unverified historical changes or include secrets/API keys.