0.6.0

• A citation can carry an abstract, your reading of the record in your own words; the site shows it with the citation. See Summarise a record.
• Projects can define their own confidence grades under [grades] in genealogy.toml, each counted as a standard grade. A relationship, claim, evidence assessment or hypothesis with a grade shows the project's wording, and Reading the evidence explains each grade. See Use your own grades.
• Large sites stay quick: past 300 records, the catalogue, timeline, places and sources pages write one page of records and keep the rest on paged pages (catalogue/page-<n>.html, timeline/<decade>s.html, places/<region>.html, sources/page-<n>.html), while the script searches everyone from the shared data. Person menus, relationships, claims and findings no longer repeat on the tool pages. See Large sites.
• data/site-data.js carries only the fields the pages read: under a third of its former size on a tree of 4,800 people. data/site.json still has every public field.
• A rebuild removes findings archive pages that a shorter log no longer needs.

Genealogy Research Workbench 0.5.0

Changelog

0.5.0 — 20 October 2026

• Findings can state their limitations (Finding.limitations), cite records directly (Finding.citation_ids) and carry a project's own type (Finding.kind). The site shows all three; finding add takes --limitations, --cite and --kind. A finding citing a non-public record is withheld.
• Research prose supports pipe tables, including links and record references in cells.
• A findings log of more than 30 dated findings opens with the newest 30. Older ones move to archive pages numbered from the earliest finding, so their links stay stable. The home page lists the newest 10 findings.

0.4.1 — 20 October 2026

• The README and Your first site install the toolkit from PyPI, so the PyPI project page shows the current instructions.

0.4.0 — 20 October 2026

• GEDCOM 7 .gdz archives import their gedcom.ged without unpacking. Media files are counted and kept only in the private archive copy.
• Importing a path that does not exist names the missing file instead of reporting a size limit.
• The local preview server now serves the CSV, GEDCOM and XML downloads it builds (previously 404).
• genealogy.feeds.atom_feed writes deterministic Atom feeds; the latest-findings feed gains a self link and full-text summaries.
• Releases publish to PyPI through a trusted-publishing workflow.
• A step-by-step guide, Your first site, from a tree export to a hosted site.

0.3.0 — 20 October 2026

Import

• GEDCOM follows its HEAD.CHAR declaration: UTF-8/UTF-16 with or without BOM, ANSI (Windows-1252), ANSEL, ASCII, MACINTOSH and ISO-8859; undecodable bytes are reported by line.
• Raw line breaks inside long notes (as written by WikiTree's GEDCOM export) are kept as continued text and reported instead of rejecting the file.
• Parent links keep their supplied nature (PEDI, _FREL/_MREL, ADOP, WikiTree non-biological) in Relationship.pedigree; the tree's own marks (WikiTree uncertain/confident/DNA confirmed, GEDCOM 7 STAT) in Relationship.tree_status, never raising confidence. DISPROVEN links are disproved and not published.
• Names keep given/surname parts, nicknames, married and other names; GEDCOM identifiers (_UID, REFN, EXID, _FSFTID, AFN, WWW), source AUTH/PUBL/call numbers, citation QUAY, GEDCOM 7 SNOTE and about 40 event types are imported. Unsupported structures are counted in one warning; SSN/IDNO are never imported.
• Burial or cremation records mark a person deceased. Invalid coordinates are reported and dropped rather than failing the import.
• OneTree: WikiTree YYYY-00-00 dates and DataStatus guess/before/after qualifiers, MiddleName, LastNameOther, Nicknames, IsLiving, one relationship per couple, marriage events from spouse rows, hidden spouses kept private, one summary warning for relatives outside the export.
• import --update refreshes a previous import of the same tree and keeps research added to the case, refusing without changes if that research would be orphaned. --update and --replace first back up the previous case privately.

Publication

• An approximate birth (ABT/EST/CAL) is treated as up to ten years later for the 120-year rule; open-ended dates still need review. genealogy.publication.is_historical_birth exposes the rule.
• The recorded sex of eligible people is published as an unassessed tree assertion.

Website

• Dossiers have a family section (parents, partners and marriages, children, siblings) with roles, labels and a competing-parent-records notice.
• One labelled evidence vocabulary (genealogy.labels) and a reader's guide, evidence.html.
• Interactive pages share one data/site-data.js instead of embedding the dataset in every page (works from file://).
• Surname filter and column in the catalogue; outstanding research listed first.
• Public GEDCOM 5.5.1 download (data/public-tree.ged, export gedcom), and sitemap.xml/canonical links when site.base_url is set.

Research and website (continued)

• Study records: a written study with summary, rationale, a small escaped Markdown body with [[record]] references, a conclusion and linked subjects. Each study gets a page and an index entry, and dossiers link to it. A study is published only when every record it lists or references is public.

• Finding.recorded_on and finding add --recorded-on date findings. A Latest findings log lists them newest first, with an Atom feed when site.base_url is set.

• Families & branches page: surname groups with years, places and earliest people, plus connected groups of recorded relatives. The catalogue accepts ?surname=/?q= and the tree accepts ?person=.

• Place.precision (exact … country, or approximate). Broad locations are drawn as dashed rings and labelled on the map and place list.

• genealogy check-links and genealogy.sitecheck: a static-site link and anchor checker with a ratcheting baseline, extracted from the Glasgow project's release gate.

• Places are grouped by their broadest recorded jurisdiction ("…, Scotland"), with a country or region summary of places, coordinates, events and people. Nothing is looked up.

• Relationship.estimate: an optional assessed percentage (1–99), shown as "Probable (70%)". Imports never set it.

• Muted and eyebrow text colours now meet WCAG AA contrast. An opt-in axe-core browser check (GENEALOGY_AXE_PATH, run in CI) covers every page type at desktop and mobile widths.

Fixes

• The fictional demo stored parent links from parent to child, contrary to the documented child → parent direction, so its tree drew descendants as ancestors. It is regenerated.
• Browser tests no longer assume a macOS Chrome path (GENEALOGY_BROWSER_EXECUTABLE).
• genealogy.dates adds year_of and wikitree_date; sorting understands Julian/dual years, circa/c., interpreted dates and month names.

Compatibility: cases gain optional fields (Name.given/surname, Relationship.pedigree/tree_status, Source.author/publication, Finding.recorded_on, Place.precision, Relationship.estimate) and a Study record type. Older 0.2 readers cannot load cases written with them. GEDCOM-imported people now record living as "deceased" rather than false. OneTree spouse relationship IDs are now one per couple, so re-import with --replace (not --update) if research refers to the old per-profile spouse IDs.

0.2.0 — 19 October 2026

• Standalone MIT-licensed package, Python 3.11+, wheel/source distribution metadata and CI.
• Local browser upload wizard and explicit file-import workflow for GEDCOM and OneTree JSON.
• Expanded GEDCOM events, coordinates, names, notes, sources and citations; OneTree wrappers, dictionaries, lists, spouses and parent links.
• Explicit publication review for basic deceased-person facts and optional source bibliography; protected material and raw input stay private.
• Connected offline website: people dossiers/search, whole tree, family focus, branches, timeline, coordinate map, sources, research, comparisons, connections and downloads.
• Linear-time bulk case insertion and bounded per-person pages for larger trees.
• Synthetic 42-person example, import fixtures, browser end-to-end tests and installed-wheel verification.
• Real-export compatibility checks; opaque external record IDs containing VCF no longer trigger the raw-genetic-field guard, while nested genetic payloads remain rejected.

0.1.0

Initial extracted research model, cases, privacy projection, platform adapters, basic views and Glasgow compatibility bridges. This version was used as a source snapshot by the Glasgow project.

Genealogy Research Workbench 0.4.0

Changelog

0.4.0 — 20 October 2026

• GEDCOM 7 .gdz archives import their gedcom.ged without unpacking. Media files are counted and kept only in the private archive copy.
• Importing a path that does not exist names the missing file instead of reporting a size limit.
• The local preview server now serves the CSV, GEDCOM and XML downloads it builds (previously 404).
• genealogy.feeds.atom_feed writes deterministic Atom feeds; the latest-findings feed gains a self link and full-text summaries.
• Releases publish to PyPI through a trusted-publishing workflow.
• A step-by-step guide, Your first site, from a tree export to a hosted site.

0.3.0 — 20 October 2026

Import

• GEDCOM follows its HEAD.CHAR declaration: UTF-8/UTF-16 with or without BOM, ANSI (Windows-1252), ANSEL, ASCII, MACINTOSH and ISO-8859; undecodable bytes are reported by line.
• Raw line breaks inside long notes (as written by WikiTree's GEDCOM export) are kept as continued text and reported instead of rejecting the file.
• Parent links keep their supplied nature (PEDI, _FREL/_MREL, ADOP, WikiTree non-biological) in Relationship.pedigree; the tree's own marks (WikiTree uncertain/confident/DNA confirmed, GEDCOM 7 STAT) in Relationship.tree_status, never raising confidence. DISPROVEN links are disproved and not published.
• Names keep given/surname parts, nicknames, married and other names; GEDCOM identifiers (_UID, REFN, EXID, _FSFTID, AFN, WWW), source AUTH/PUBL/call numbers, citation QUAY, GEDCOM 7 SNOTE and about 40 event types are imported. Unsupported structures are counted in one warning; SSN/IDNO are never imported.
• Burial or cremation records mark a person deceased. Invalid coordinates are reported and dropped rather than failing the import.
• OneTree: WikiTree YYYY-00-00 dates and DataStatus guess/before/after qualifiers, MiddleName, LastNameOther, Nicknames, IsLiving, one relationship per couple, marriage events from spouse rows, hidden spouses kept private, one summary warning for relatives outside the export.
• import --update refreshes a previous import of the same tree and keeps research added to the case, refusing without changes if that research would be orphaned. --update and --replace first back up the previous case privately.

Publication

• An approximate birth (ABT/EST/CAL) is treated as up to ten years later for the 120-year rule; open-ended dates still need review. genealogy.publication.is_historical_birth exposes the rule.
• The recorded sex of eligible people is published as an unassessed tree assertion.

Website

• Dossiers have a family section (parents, partners and marriages, children, siblings) with roles, labels and a competing-parent-records notice.
• One labelled evidence vocabulary (genealogy.labels) and a reader's guide, evidence.html.
• Interactive pages share one data/site-data.js instead of embedding the dataset in every page (works from file://).
• Surname filter and column in the catalogue; outstanding research listed first.
• Public GEDCOM 5.5.1 download (data/public-tree.ged, export gedcom), and sitemap.xml/canonical links when site.base_url is set.

Research and website (continued)

• Study records: a written study with summary, rationale, a small escaped Markdown body with [[record]] references, a conclusion and linked subjects. Each study gets a page and an index entry, and dossiers link to it. A study is published only when every record it lists or references is public.

• Finding.recorded_on and finding add --recorded-on date findings. A Latest findings log lists them newest first, with an Atom feed when site.base_url is set.

• Families & branches page: surname groups with years, places and earliest people, plus connected groups of recorded relatives. The catalogue accepts ?surname=/?q= and the tree accepts ?person=.

• Place.precision (exact … country, or approximate). Broad locations are drawn as dashed rings and labelled on the map and place list.

• genealogy check-links and genealogy.sitecheck: a static-site link and anchor checker with a ratcheting baseline, extracted from the Glasgow project's release gate.

• Places are grouped by their broadest recorded jurisdiction ("…, Scotland"), with a country or region summary of places, coordinates, events and people. Nothing is looked up.

• Relationship.estimate: an optional assessed percentage (1–99), shown as "Probable (70%)". Imports never set it.

• Muted and eyebrow text colours now meet WCAG AA contrast. An opt-in axe-core browser check (GENEALOGY_AXE_PATH, run in CI) covers every page type at desktop and mobile widths.

Fixes

• The fictional demo stored parent links from parent to child, contrary to the documented child → parent direction, so its tree drew descendants as ancestors. It is regenerated.
• Browser tests no longer assume a macOS Chrome path (GENEALOGY_BROWSER_EXECUTABLE).
• genealogy.dates adds year_of and wikitree_date; sorting understands Julian/dual years, circa/c., interpreted dates and month names.

Compatibility: cases gain optional fields (Name.given/surname, Relationship.pedigree/tree_status, Source.author/publication, Finding.recorded_on, Place.precision, Relationship.estimate) and a Study record type. Older 0.2 readers cannot load cases written with them. GEDCOM-imported people now record living as "deceased" rather than false. OneTree spouse relationship IDs are now one per couple, so re-import with --replace (not --update) if research refers to the old per-profile spouse IDs.

0.2.0 — 19 October 2026

• Standalone MIT-licensed package, Python 3.11+, wheel/source distribution metadata and CI.
• Local browser upload wizard and explicit file-import workflow for GEDCOM and OneTree JSON.
• Expanded GEDCOM events, coordinates, names, notes, sources and citations; OneTree wrappers, dictionaries, lists, spouses and parent links.
• Explicit publication review for basic deceased-person facts and optional source bibliography; protected material and raw input stay private.
• Connected offline website: people dossiers/search, whole tree, family focus, branches, timeline, coordinate map, sources, research, comparisons, connections and downloads.
• Linear-time bulk case insertion and bounded per-person pages for larger trees.
• Synthetic 42-person example, import fixtures, browser end-to-end tests and installed-wheel verification.
• Real-export compatibility checks; opaque external record IDs containing VCF no longer trigger the raw-genetic-field guard, while nested genetic payloads remain rejected.

0.1.0

Initial extracted research model, cases, privacy projection, platform adapters, basic views and Glasgow compatibility bridges. This version was used as a source snapshot by the Glasgow project.

Genealogy Research Workbench 0.3.0

Changelog

0.3.0 — 20 October 2026

Import

• GEDCOM follows its HEAD.CHAR declaration: UTF-8/UTF-16 with or without BOM, ANSI (Windows-1252), ANSEL, ASCII, MACINTOSH and ISO-8859; undecodable bytes are reported by line.
• Raw line breaks inside long notes (as written by WikiTree's GEDCOM export) are kept as continued text and reported instead of rejecting the file.
• Parent links keep their supplied nature (PEDI, _FREL/_MREL, ADOP, WikiTree non-biological) in Relationship.pedigree; the tree's own marks (WikiTree uncertain/confident/DNA confirmed, GEDCOM 7 STAT) in Relationship.tree_status, never raising confidence. DISPROVEN links are disproved and not published.
• Names keep given/surname parts, nicknames, married and other names; GEDCOM identifiers (_UID, REFN, EXID, _FSFTID, AFN, WWW), source AUTH/PUBL/call numbers, citation QUAY, GEDCOM 7 SNOTE and about 40 event types are imported. Unsupported structures are counted in one warning; SSN/IDNO are never imported.
• Burial or cremation records mark a person deceased. Invalid coordinates are reported and dropped rather than failing the import.
• OneTree: WikiTree YYYY-00-00 dates and DataStatus guess/before/after qualifiers, MiddleName, LastNameOther, Nicknames, IsLiving, one relationship per couple, marriage events from spouse rows, hidden spouses kept private, one summary warning for relatives outside the export.
• import --update refreshes a previous import of the same tree and keeps research added to the case, refusing without changes if that research would be orphaned. --update and --replace first back up the previous case privately.

Publication

• An approximate birth (ABT/EST/CAL) is treated as up to ten years later for the 120-year rule; open-ended dates still need review. genealogy.publication.is_historical_birth exposes the rule.
• The recorded sex of eligible people is published as an unassessed tree assertion.

Website

• Dossiers have a family section (parents, partners and marriages, children, siblings) with roles, labels and a competing-parent-records notice.
• One labelled evidence vocabulary (genealogy.labels) and a reader's guide, evidence.html.
• Interactive pages share one data/site-data.js instead of embedding the dataset in every page (works from file://).
• Surname filter and column in the catalogue; outstanding research listed first.
• Public GEDCOM 5.5.1 download (data/public-tree.ged, export gedcom), and sitemap.xml/canonical links when site.base_url is set.

Research and website (continued)

• Study records: a written study with summary, rationale, a small escaped Markdown body with [[record]] references, a conclusion and linked subjects. Each study gets a page and an index entry, and dossiers link to it. A study is published only when every record it lists or references is public.

• Finding.recorded_on and finding add --recorded-on date findings. A Latest findings log lists them newest first, with an Atom feed when site.base_url is set.

• Families & branches page: surname groups with years, places and earliest people, plus connected groups of recorded relatives. The catalogue accepts ?surname=/?q= and the tree accepts ?person=.

• Place.precision (exact … country, or approximate). Broad locations are drawn as dashed rings and labelled on the map and place list.

• genealogy check-links and genealogy.sitecheck: a static-site link and anchor checker with a ratcheting baseline, extracted from the Glasgow project's release gate.

• Places are grouped by their broadest recorded jurisdiction ("…, Scotland"), with a country or region summary of places, coordinates, events and people. Nothing is looked up.

• Relationship.estimate: an optional assessed percentage (1–99), shown as "Probable (70%)". Imports never set it.

• Muted and eyebrow text colours now meet WCAG AA contrast. An opt-in axe-core browser check (GENEALOGY_AXE_PATH, run in CI) covers every page type at desktop and mobile widths.

Fixes

• The fictional demo stored parent links from parent to child, contrary to the documented child → parent direction, so its tree drew descendants as ancestors. It is regenerated.
• Browser tests no longer assume a macOS Chrome path (GENEALOGY_BROWSER_EXECUTABLE).
• genealogy.dates adds year_of and wikitree_date; sorting understands Julian/dual years, circa/c., interpreted dates and month names.

Compatibility: cases gain optional fields (Name.given/surname, Relationship.pedigree/tree_status, Source.author/publication, Finding.recorded_on, Place.precision, Relationship.estimate) and a Study record type. Older 0.2 readers cannot load cases written with them. GEDCOM-imported people now record living as "deceased" rather than false. OneTree spouse relationship IDs are now one per couple, so re-import with --replace (not --update) if research refers to the old per-profile spouse IDs.

0.2.0 — 19 October 2026

• Standalone MIT-licensed package, Python 3.11+, wheel/source distribution metadata and CI.
• Local browser upload wizard and explicit file-import workflow for GEDCOM and OneTree JSON.
• Expanded GEDCOM events, coordinates, names, notes, sources and citations; OneTree wrappers, dictionaries, lists, spouses and parent links.
• Explicit publication review for basic deceased-person facts and optional source bibliography; protected material and raw input stay private.
• Connected offline website: people dossiers/search, whole tree, family focus, branches, timeline, coordinate map, sources, research, comparisons, connections and downloads.
• Linear-time bulk case insertion and bounded per-person pages for larger trees.
• Synthetic 42-person example, import fixtures, browser end-to-end tests and installed-wheel verification.
• Real-export compatibility checks; opaque external record IDs containing VCF no longer trigger the raw-genetic-field guard, while nested genetic payloads remain rejected.

0.1.0

Initial extracted research model, cases, privacy projection, platform adapters, basic views and Glasgow compatibility bridges. This version was used as a source snapshot by the Glasgow project.

Genealogy Research Workbench 0.2.0

Turn GEDCOM and OneTree exports into a local research workbench and an offline family-history website.

• Browser file chooser and CLI import; explicit publication review, private raw-input storage, and living/restricted/licensed-data protection.
• Searchable people dossiers, whole tree and family branches, timeline, offline coordinate map, sources, comparisons, connection paths, and portable exports.
• Python 3.11+, no runtime dependencies, MIT license, fictional demo and screenshot tour.
• Built wheel and source archive attached. Download a wheel and run python -m pip install /path/to/genealogy_research_workbench-0.2.0-py3-none-any.whl, then genealogy --project my-workspace serve.

Local verification: 87 unit tests passed; 4 separate browser tests passed; installed-wheel imports/builds passed outside the checkout. Two real GEDCOM exports (169 and 2,079 people) and two real OneTree exports (238 and 4,394 people) passed import, privacy checks and complete site generation. Real data is not included in the repository or release.

Imported relationships remain unverified assertions. Missing relatives are omitted with warnings, maps require supplied coordinates, UTF-8 input is required, and arbitrary OneTree biography markup does not become verified source citations. Read the README and import/privacy guides before publishing a generated site. This release is not on PyPI.