🌐 Open the interactive Scriptorium Reader (GitHub Pages) →
Russian-first public Reader (merged as PR #333; live HTTP not independently verified): GitHub Pages uses master / (root). The new home page groups 10 Russian-language works: seven with actual source-pinned observations and three clearly marked as work in progress. The 7 measured books provide 182 fields, 153 numeric observations and 29 explicitly unavailable values. Search, book-to-book rate comparisons, chart bars, JSON exports and source metadata are reader-facing; cross-language original/translation analysis is not yet present. The multi-route Python-generated Reader remains a separate local/CI build in build/site, rather than the root branch Pages experience; its retained FantLab matrix currently covers four measured books and one unmeasured source. The Hyperboloid generated route described below is separately measured and does not silently alter that matrix.
Seventh measured Russian work (merged to master in #342; HTTP unverified): Tolstoy’s Resurrection has 29 source-frozen fields (24 numeric, five NOT_RUN) from 129 pinned revisions, body SHA-256, two public FantLab count references and original CI-artifact hashes. ?book=resurrection#measurements is implemented in the root Reader; independent live-route verification remains outstanding. FantLab input identity is UNKNOWN, comparison DIAGNOSTIC_ONLY, M2 0/5. No source prose is published.
Shareable book views (root Reader source): Every measured Russian edition can now be opened directly in the same-page metric explorer, for example Beketova's Russian translation or Silver Dove. These addresses select only allow-listed, verified measurement records; pending/invalid IDs fall back to Anna Karenina. Back/forward, keyboard focus and source-JSON fallbacks are retained. The URL is a target for the existing branch Pages deployment after the relevant commit is published—the README link alone is not proof of live deployment. This public root navigation is separate from the generated Reader's multi-page routes.
Shareable two-book comparisons (root Reader source): ?left=anna&right=beketova#pairs identifies exactly two measured Russian editions in the side-by-side rates table. Both selectors update browser history; Back/Forward and reload restore a valid pair. Invalid, pending, partial and duplicate URL pairs fall back to Anna Karenina × Brothers Karamazov. Selecting the same book in both in-page controls instead displays an error and hides the link. The URL cannot specify arbitrary data sources; M2 remains 0/5. The pair-link feature merged in #337 and its master Pages deployment workflow succeeded on 2026-10-27; browser/HTTP LIVE_VERIFIED remains outstanding.
Source-free pair JSON and CSV exports (root Reader): The comparison table can be downloaded with exactly the jointly measured rows, left/right values, numeric right-minus-left differences with explicit delta units (percentage points for percentages), original measurement units, versioned shared profiles, each edition and translator identity, immutable text hashes and links to the evidence records. The deterministic export labels DIAGNOSTIC_ONLY, unknown FantLab input identity and M2 0/5 explicitly; it includes no literary text and does not send data anywhere. The export is unavailable for duplicate, unmeasured or algorithm-incompatible pairs. The JSON comparison export merged in #338, but its live HTTP result is not independently verified. The additional CSV table uses the same validated measurements, percentage-point delta units, translators, edition/source SHA-256, complete shared algorithm profiles and evidence URLs. CSV text is quoted and spreadsheet-formula-neutralized; numbers remain numeric. CSV export was merged in #339; live browser/HTTP confirmation remains outstanding.
Generated Reader stage — local/CI only: The separately built build/site/index.html contains a measured-stage panel derived at build time from the retained FantLab evidence matrix: four works with numeric observations, 94 measured values, 19 unavailable actuals, and M2 0/5 source-matched works. It links to the generated build/site/evidence/fantlab/ table. This specific stage panel and multi-route evidence view are not published by the branch-root Pages URL. The public homepage has its own seven-book interactive charts, ten-title catalogue and source-free metric explorer. The full generated Reader additionally registers Hyperboloid of Engineer Garin at build/site/works/hyperboloid-garin-measured/: 29 pinned source-free fields (24 measured, five NOT_RUN), exact-value CSV/JSON and CI/source receipts. The same local/CI build now also contains a separately pinned Beketova Russian translation report at build/site/works/captain-grant-beketova-measured/: all 11 verified derived observations, edition/translator attribution and source-free JSON/CSV exports. Both extra generated-only reports lack confirmed FantLab input identity or parity and do not change the branch Pages catalogue.
The separately built full Reader can show real measured differences between two books, independently search and filter Anna Karenina's 28-row and Running on Waves' 27-row diagnostics, inspect Silver Dove and Brothers Karamazov's 29 fields each, filter their twenty shared measured rates independently by metric name or family, examine a full-work diagnostic, browse the source-free FantLab evidence matrix, or compare two of your own Reader JSON reports privately in the browser. Charts and a visible exact-value table show retained, validated derived measurements, units and links to methodology and scientific limitations. No manuscript is hosted or uploaded.
GitHub Pages branch deployment (verified 2026-10-27): GitHub reports has_pages=true. The GitHub-managed pages build and deployment workflow completed successfully on master, with the Jekyll checkout configured as source: .. The previous blocker was not lack of Pages administration: it was a missing root index.html and public assets. Merged PR #332 supplies them. A successful deployment is not a claim that the public site is already live with the expected Reader until an independent HTTP/browser check confirms it. The richer Python-built Reader remains available offline:
python -m scriptorium.site_renderer --repo-root . --output build/site
# Open build/site/index.htmlThe separate, historical Actions artifact deployment job remains guarded by SCRIPTORIUM_PAGES_DEPLOY_ENABLED=true, but is not required for GitHub's already configured branch Pages publication. Do not change the deployment gate or repository settings to make the root homepage work. See #285 for the distinction and #284 for unresolved software/data reuse licensing.
Retained Resurrection field diagnostic (CI/replay and root Reader, source-free): The 129 exact pinned Russian Wikisource chapters of Tolstoy's Resurrection can be replayed into 29 versioned Scriptorium fields in one verified pass. The output retains numerical values, units, two publicly documented FantLab count references and signed deltas, with dictionary-dependent values as NOT_RUN. It includes no chapter prose, and explicitly records unknown FantLab input edition, diagnostic-only comparison and M2 0/5. CI creates separate create-only replay/field receipts; the verified source-free 29-row receipt is retained in master (24 numeric and five NOT_RUN). GitHub Pages live-route verification remains outstanding. Reproduce when the pinned source endpoint is reachable:
python -m scriptorium.resurrection_diagnostic \
--manifest corpus/candidates/source-edition-traces/tolstoy-resurrection-ru.revisions.json \
--receipt resurrection-replay-receipt.json --output resurrection-fields.jsonWhat the Russian-first demo shows: seven source-frozen, measured Russian-language texts (including a separately identified translation by Alexandra A. Beketova), with 182 recorded fields / 153 numeric / 29 unavailable. Three further Russian-text candidates are listed as preparation only: Road to Nowhere, Petersburg and Darwin's Origin of Species in Sergey Rachinsky's 1864 Russian translation. Ten catalogue entries do not mean ten analyzed books. Translated texts keep named translators, original language, family, exact edition and source identity; a French original must not be silently treated as an English text. The English analyzer is a separate future module with zero measured original-language works today. The root Reading Room offers descriptive comparisons of normalized candidate metrics on the same pinned Russian algorithm profile; original-vs-translation and translation-vs-translation comparisons require independently verified per-version data. M2 FantLab parity remains 0/5. Browser LIVE_VERIFIED requires independent HTTP/screenshot evidence; publication must not be inferred from CI builds alone. Root Pages deployment and the Python-built multi-route Reader are distinct.
Understand the language of a text without uploading the manuscript. Scriptorium measures vocabulary, sentence length, dialogue and punctuation, then produces readable offline reports and reproducible data. This 0.4.0a1 preview adds private comparison of two analyses to the canonical Reader and its real-work gallery.
The Russian module (ru) is the initial product priority. Every analyzed literary body has an analysis language and an independent edition identity. A translated work additionally records the author's original language, a named translator and a specific edition/source family; another translator or translation revision becomes a separate identity, not another spelling of the title. The measured Russian translation of Дети капитана Гранта by А. А. Бекетова has 11 real derived fields, but no measured French original. The researched translation of О происхождении видов by С. А. Рачинский has no retained numeric fields, and its English source is not analyzed here. Never place invented English/French numbers into Russian graphs. See versioned language/translation contract.
Coverage target: 10 analyzed works, each individually rights-cleared, byte/source-frozen and measured. Current committed root-Reader data (through merged PR #342): 7 measured / 10 catalogued; the remaining three are labelled honestly and require separate reproducible measurements. This is unrelated to the stricter FantLab M2 source-matched parity requirement (0/5).
With Python 3.13 or later, from this checkout:
python -m pip install .
scriptorium analyze manuscript.txt --format html --output report.html
scriptorium analyze manuscript.txt --output analysis.json
scriptorium analyze manuscript.txt --format csv --output metrics.csvOpen report.html locally. All 29 metric rows retain their units and algorithm
profiles; unavailable measurements stay unknown. The report has readable metric names,
a four-value overview, exact numeric values and keyboard-accessible tables. JSON and
CSV include source digests, not manuscript content. Choose a new output filename for
each run: existing files are never overwritten.
No runtime dependencies, account, server or telemetry are required. Installation may
fetch build tools; see offline wheel installation.
python -m scriptorium exposes the same commands from a checkout.
Import an EPUB 2/3 with a standard XHTML reading-order spine without uploading the book or writing its extracted chapters to disk:
scriptorium epub book.epub --output analysis.json --provenance-output epub-identity.json
scriptorium epub book.epub --format html --output analysis.html
scriptorium epub book.epub --format csv --output metrics.csvThe regular Reader 29-metric reports work with scriptorium compare.
The separate source-free receipt records the archive, extracted UTF-8
body and engine-normalized text SHA-256 identities plus the versioned
extraction profile. EPUB reading order follows OPF spine order; unsupported
encrypted, malformed, unsafe or image-only books fail closed. No DRM bypass,
external network resources or embedded scripts are processed. This local
EPUB text projection is not FantLab edition/input parity or corpus admission.
EPUB formats, safety limits and reproducibility.
Select up to 16 UTF-8 manuscripts explicitly in a JSON manifest, then generate source-free per-work reports and an HTML/JSON/CSV index locally:
printf '%s\n' '{"schema_version":"scriptorium-private-batch-manifest-v1","files":["first.txt","second.txt"]}' > manifest.json
scriptorium batch manifest.json --output-dir private-resultsThe manifest paths are relative to its directory. No implicit crawling, uploads, overwrites or source-text publication. Failed items are marked unavailable, never assigned zero metrics. Re-running an unchanged batch reuses identical outputs; conflicting files fail closed. Use a private POSIX filesystem: directory-relative no-follow operations are required. This workflow does not admit works to the corpus or establish FantLab parity. Batch use, limits and exit codes.
scriptorium analyze second-manuscript.txt --output second-analysis.json
scriptorium compare analysis.json second-analysis.json --output comparison.html
scriptorium compare analysis.json second-analysis.json --format json --output comparison.json
scriptorium compare analysis.json second-analysis.json --format csv --output comparison.csvOpen comparison.html for the same 29 measurements, explicit units, source identities
and right minus left differences. Percentage differences use percentage points.
Unknown measurements do not become zero; different methods or dictionaries are labelled
incomparable where applicable. JSON decimal observations and their differences retain
precision. Contradictory counts, means, punctuation rates and same-text results are rejected.
Imported JSON and hashes are declarations, not authenticated proof: rerun the analyzer
on the original bytes to verify them. Comparison contract and examples.
Build the canonical site with scriptorium site --repo-root . --output build/site
and open build/site/compare/index.html, or use the Compare private reports
action in the generated homepage. Select two local scriptorium analyze JSON
reports; no manuscripts or report bytes are uploaded. Inspect all 29 measurements,
incomparable/missing reasons, exact decimal right-minus-left differences, and
download source-free JSON/CSV. The browser validates the same versioned report
contract as the installed Python Reader and rejects duplicate keys, integer
lexeme spoofs, impossible counts, nonfinite values and changed units. Browser
support requires JavaScript; the offline scriptorium compare CLI always works.
Browser validation and limitations.
This is a local build capability, not a live GitHub Pages deployment. No private manuscripts are included; imported report hashes are unauthenticated declarations. Scientific/FantLab parity remains unverified (M2 0/5).
The repository contains a source-free, historical full-work diagnostic for Anna Karenina: 28 expected/actual/delta rows, with 23 numeric results and five unavailable values. No novel download is needed:
scriptorium diagnostic benchmarks/fantlab/work270306-wikisource-diagnostic.json \
--title "Anna Karenina — full-work diagnostic" --output anna-report.html
scriptorium site --repo-root . --output build/siteOpen anna-report.html or build/site/index.html. The canonical site retains the five
earlier evidence/report routes and the Reader 0.3 two-book comparison route.
A successful local build is not a live GitHub Pages deployment. The Pages activation
interlock is unchanged; this preview does not enable publishing by itself. Atomic site
replacement currently requires Linux renameat2; unsupported platforms fail without
deleting the previous build. Private text analysis and comparison are separate capabilities.
No manuscript download or morphology setup is needed for the frozen Running on Waves and Children of Captain Grant — Beketova translation comparison:
mkdir -p build
python -m scriptorium.full_work_gallery --repo-root . --format html > build/full-work-gallery.html
python -m scriptorium.full_work_gallery --repo-root . --format csv > build/full-work-evidence.csvOpen build/full-work-gallery.html for ten exact measurements and seven row-wise charts.
These are Scriptorium-side results, not FantLab parity or author-style scores. The pinned
canonical site route is build/site/works/two-full-work-differences/index.html, with
measurements.json and measurements.csv downloads. Navigation and export links have
44-pixel minimum touch targets; visible focus and reduced-motion styles are retained.
Newly approved replacement evidence must declare compatible source units before display;
legacy Grin diagnostics are not assigned invented source-unit fields.
Building this route does not prove GitHub Pages is deployed. Shell redirection replaces
its destination; use new filenames to keep older exports. Method and limits.
Inspect five frozen public literary sources and the distance between their character counts and the displayed FantLab counts, without downloading books:
python -m scriptorium.benchmark_readiness --root . --format markdown
python -m scriptorium.benchmark_readiness --root . --format json --output readiness.jsonEach source-free report includes source rights/edition evidence, immutable body hashes, dated FantLab references and explicit source-matching status. Differences are diagnostic only: FantLab's analyzer-input identities remain unknown and M2 stays 0/5. This character-count readiness command labels other metrics NOT_RUN in this view only; it does not erase existing, independently validated field-level observations in the metric matrix below. Readiness methodology, CSV and limitations.
To inspect actual retained per-metric differences rather than character counts alone,
run python -m scriptorium.benchmark_metric_matrix --root . --format markdown.
The source-free matrix validates 142 field-level diagnostic/extension rows: 113 across the core five-work source-readiness set (four works with retained measurements, one missing), and an additional 29 for Resurrection as a separately pinned, source-unmatched diagnostic. The original five-work set contains: Anna Karenina has 28 (23 numeric,
five without actuals), Running on Waves has 27 (23 numeric, four NOT_RUN),
and The Silver Dove and The Brothers Karamazov have 29 each (24 numeric,
five dictionary-dependent NOT_RUN each). There are 94 measured and 19 unavailable
actuals among these four. The remaining one frozen work, The Hyperboloid of Engineer Garin,
is part of the five-source readiness matrix with NOT_RUN fields, but has a separate 24-measured / 5-NOT_RUN generated Reader report. Both Silver Dove and Karamazov have
original SHA-pinned source-free CI receipts; only two FantLab displayed counts
are retained for each. The supplemental Resurrection adds 29 fields (24 numeric,
five NOT_RUN) from 129 pinned chapters; the only retained FantLab references are
characters 881,244 and words 126,457, with unmatched-source differences of
+9,591 and +9,672 respectively. In total the six retained diagnostics have
118 measured fields and 24 unavailable actuals, without increasing M2 progress. Other reference values and exact input identity remain
UNKNOWN; none establishes FantLab parity. Anna's historical rows include 4/4 general, 4/4 dialogue,
14/14 punctuation and 1/6 vocabulary measurements; its other five vocabulary
actuals are two NOT_RUN and three unresolved.
The generated evidence Reader displays all 29 Silver Dove, 29 Brothers Karamazov and 29 supplemental Resurrection fields in accessible static tables,
including separate unknown FantLab references and an explicit extension marker.
It also labels metric units and percentage-point differences, and offers
independent search/filter controls for both books by metric name/ID,
family and availability. Tables and source-free JSON/CSV remain available
without JavaScript and in print. A separate 20-row side-by-side comparison
shows only shared measured word/sentence lengths, punctuation rates and dialogue
percentages (Running on Waves minus Anna Karenina, percentage differences in
percentage points). Unlike total characters or vocabulary counts, these carry
explicit per-unit denominators, but remain descriptive, source-unmatched
Scriptorium candidates, not author-style or FantLab-parity evidence.
Neither retained work is source-matched to FantLab's analyzer input, displayed
rounding is unproven, and M2 stays 0/5.
Build the site with scriptorium site --repo-root . --output build/site and open
build/site/evidence/fantlab/index.html to browse the same verified five-work readiness evidence plus the separately labelled Resurrection diagnostic
and download its source-free JSON/CSV. The same page now also shows each
frozen work's character count, signed diagnostic difference from FantLab's
displayed count and source/edition/rights link. It also shows signed
relative drift as (frozen minus FantLab displayed) divided by FantLab
displayed, expressed in percent and rounded to two decimal places.
This display-only diagnostic is not a parity score and does not alter JSON/CSV.
These counts do not establish FantLab analyzer-input identity. The evidence page
also offers source provenance JSON/CSV downloads generated from the same
five validated readiness witnesses as the visible work table. The JSON contains
raw and normalized SHA-256 digests, reference dates and legal/source metadata;
the CSV retains raw digests and reference dates. No manuscript text is included.
The homepage links to this route.
This is a
locally built page, not a confirmed live GitHub Pages deployment.
Field-level matrix and evidence checks.
To join a separately generated source-free POS artifact from the same normalized
text into a FantLab benchmark, add --pos-metrics pos.json to the
python -m scriptorium.benchmark --reference ... --text ... --scriptorium-revision ...
command. This adds 40 explicitly unresolved/diagnostic POS rows, including
17 visible category counts and percentages, but never claims POS parity.
Unmatched text digests, profiles or inconsistent 17×17/20×17 structures are
rejected. The work488 FantLab category counts are not a simple partition of
defined words, and service-word actuals remain unknown.
POS benchmark usage and evidence boundaries.
FantLab-shaped metrics are inferred compatibility candidates, not demonstrated parity. Scriptorium-only features are extensions. The historical Anna diagnostic has no established identity match to FantLab's input text. M2 remains 0/5; none of M1–M5 is declared complete by this preview. The AOT-lineage morphology target and field-by-field parity criteria remain unchanged.
Texts shorter than 300,000 characters including spaces are valid inputs with explicit representativeness limits. That length is a corpus-admission floor, not a guarantee of quality; legal provenance and source identity require separate evidence. Source books, private manuscripts and dictionary contents are not included in reports or uploaded. Hashes can identify known works; share derived reports deliberately.
The CLI bounds manuscripts to 32 MiB and dictionaries and comparison reports to 8 MiB per file, rejects special files and symlinks, and uses a create-only atomic writer. Operational errors do not echo private paths or source text. CSV neutralizes formula-like text while preserving numeric cells; spreadsheet import/save behavior is not universally guaranteed.
python -m unittest discover -s tests -v
python -m pip wheel --no-deps . --wheel-dir distThe Reader workflow runs the complete canonical suite, builds a full wheel and tests it outside the checkout: actual manuscript analysis/comparison, HTML/JSON/CSV, historical report, safe failures and deterministic source-free gallery/site. No analyzer stubs are used in the private-comparison acceptance. Package, project and CLI versions must agree. The full Python public API, morphology modules, benchmark data and corpus records remain in place. No external package-registry release is performed by CI.
Site locks are persistent regular files whose kernel locks are released on process death. Do not delete a lock file while a publisher could be using it. Legacy directory locks from an interrupted older preview fail closed and are not automatically deleted.
Reader installation and commands · Local analysis and dictionaries · Private comparison · Private batch · Frozen diagnostic methodology · Five-work benchmark readiness · Canonical site and publication boundary · Roadmap and repository assessment
The detailed Research and evidence index retains project history, corpus investigations and original reference links. Private EPUB and batch Reader capabilities have been integrated into the repository; their inputs remain local. The browser comparison is merged into master as a local-build capability; public deployment remains unverified and separately gated. Code/data licensing and Pages activation retain separate gates (#284/#285).
Reader FantLab evidence now includes an offline-only metric explorer (label/ID, family and measurement availability) with a complete no-JavaScript/print table. Scientific source-match and parity status remain unchanged; see benchmark evidence.