Normalize exported Kindle, Kobo, and CSV highlights with provenance, filters, dedup review, and portable reports.
ProductivityGitHub @loganpendragonmultiverseOnlineOwner unconfirmed
Have you tried it? One tap is enough. No sign-in needed.
Repository information has not been collected yet.
No updates detected yet. GitHub releases will appear here.
GitHub star growth between the latest completed 8-hour observations
DeepSeek Harness Remote can securely connect a phone or another computer to the computer where you run DeepSeek Harness, so you can continue your work anytime and anywhere, check progress anytime, and continue working.
AI translation
Ayakaleaf Pro is a free, open-source, self-hosted Overleaf alternative: Overleaf Community Edition with almost all Server Pro features — track changes, project history, admin panel, SSO, Git-bridge, Z
Open-source desktop app to view and edit HWP/HWPX documents on macOS, Windows, and Linux
AI summary · website content
A modern, high-performance local file search and productivity tool for Windows. A sleek, customizable alternative to Everything and Listary built with C# WPF and NT Services, featuring instant NTFS MF
The maker's introduction consists of the line above. Read the repository README for more details. Repository README ↗
Loading the full README.
Read the full README ↗Normalize exported highlights and notes from multiple ebook platforms into portable records. The command uses explicit UTF-8 JSON input and produces reviewable JSON or Markdown output.
python -m pip install .
annotation-extract examples/sample.json
annotation-extract examples/sample.json --format json --output report.json
annotation-extract kobo.csv --adapter kobo-csv --format csv --output annotations.csv
The example documents the v1 input shape. Existing report files are never overwritten. Source inputs are read-only except where the documented purpose explicitly creates a new output artifact.
Version 1.1 reads the original JSON format plus exported Kindle text/HTML, Kobo CSV, and generic CSV files. Every normalized item includes source provenance. Repeated --work and --platform filters, date bounds, and --kind highlight|note narrow a report. Markdown groups annotations by work, CSV provides a portable table, and duplicate candidates remain visible for review.
The tool runs locally and does not upload input or include telemetry. Python 3.10 or newer is supported on Windows, macOS, and Linux.
The tool only normalizes files the user has already exported. It does not bypass DRM, log into platforms, or recover unavailable annotations.
python -m pip install -e ".[dev]"
ruff format --check .
ruff check .
mypy src
pytest
python -m build
Release metadata must stay aligned across the package, changelog, GitHub release, and Logan Pendragon Forge catalog.
Part of the Logan Pendragon Forge open-source collection. Licensed under the MIT License.
Use annotation-extract "My Clippings.txt" --adapter kindle-clippings --format json for the supported English Kindle My Clippings text dialect. Each block has a title line (preserved verbatim, including an author in parentheses), - Your Highlight on Location 12-13 | Added on Thursday, October 15, 2026 10:30:00 AM, a blank line, multiline body, and a line containing exactly ==========. Note and empty-body Bookmark blocks are supported, as is on page 3 | Location 12-13 metadata. Dates become local ISO timestamps without an invented timezone. Note bodies populate note; highlights populate text; bookmarks retain their location and type. Malformed blocks fail the import with a numbered diagnostic instead of becoming annotations. The tests/fixtures/My Clippings.txt file is wholly synthetic. Other languages, date layouts and HTML exports are outside this dialect.
kindle-text and kindle-html retain their legacy line-based behavior for compatibility; they do not parse My Clippings metadata or general HTML. JSON, Kobo/generic CSV and JSON/CSV report output remain supported. Work/platform/date/kind filtering now precedes deduplication, and duplicate review includes only eligible records. --kind bookmark selects bookmarks. This tool uses exported local files only, with no DRM access, login or network calls.
Annotation text, notes, titles, platform, source and location fields must be strings.
Malformed values fail clearly instead of becoming invented annotation text.
An empty source list produces an empty report; missing sources remain an error.
Dates may be null when no date filter is requested. Bounds use valid YYYY-MM-DD
calendar dates and include the entire exported local day at both ends:
annotation-extract "My Clippings.txt" --adapter kindle-clippings --date-from 2026-10-15 --date-to 2026-10-15
Undated annotations are excluded when a bound is present. Reversed or malformed bounds fail clearly. Markdown displays supplied markup literally and preserves multiline annotations. JSON and CSV preserve validated, normalized text. Existing report files are protected by exclusive creation.
The installed wheel includes all four project documents and examples under
share/doc/book-annotation-extractor/ inside the Python environment. Use
python -c "import sys; print(sys.prefix)" to locate that environment. Run the
installed annotation-extract command against the packaged examples/sample.json
or the synthetic examples/My Clippings.txt without needing a source checkout.
CSV imports require nonempty, unique headers and the same number of cells in every nonblank record. Quoted commas and multiline fields remain supported. Ambiguous or malformed CSV fails before report generation instead of discarding an earlier column or inventing a missing cell.
Existing output paths, including dangling symbolic links, are refused before exclusive file creation. This protects ordinary output collisions; it does not promise protection against hostile concurrent file-system changes.
Use annotation-extract input.json --preserve-occurrences --format json when repeated highlights or notes have locations or source details you want to keep. The same option works with all existing input adapters and with --format csv or --format markdown.
The opt-in JSON schema is version 3: each unique annotation contains occurrences, including the kept occurrence and every duplicate in input order, with platform, work, location, text, note, kind, exported date and source/source-index/item-index provenance. count remains the number of unique annotations; occurrence_count counts all included occurrences. CSV emits one row per occurrence, adding annotation/occurrence/source/item indices; Markdown displays all occurrence details as literal text. These indices describe this input/report and are not persistent cross-run IDs.
Existing filters apply to each occurrence before grouping. Repeated identical locations remain separate occurrences; bookmark locations and annotation kinds retain their existing identity rules. Text fields retain existing trimming/normalization, and timestamps keep their supplied interpretation: no timezone is inferred. Only documented normalized fields are retained, not arbitrary export metadata or the original raw file bytes.
Without the option, JSON stays version 2 and CSV/Markdown retain their existing shape and deduplication behavior. Source files remain read-only and existing output files are refused. Occurrence details can reveal source filenames and annotation content; review exported reports before sharing.