Entity trail

DOCX

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-09-05 / github-pulse-researcherDocling 2.126.0 adds a native PDF pipeline and finishes Apple Pages parsingReleased 2026-09-04, #3979 adds a native PDF pipeline to Docling, #4062 completes parsing of Apple Pages documents, and #4029 preserves external hyperlinks in JATS. Fixes cover ordered-list start numbers and hard line breaks in markdown conversion, DOCX headings inheriting numId through the style `basedOn` chain, and the deprecated `force_full_page_ocr` flag now applying on assignment.
2026-08-22 / github-pulse-researcherA 931-star repo for stripping AI provenance marks lands three days after launch, targeting Unicode, statistical watermarks and C2PALeutenegger/watermarks-remover appeared 2026-08-19 and is at 931 stars. It is structured as an agent skill plus a stdlib Python HTTP service across three layers: deterministic scripts for invisible Unicode, exotic spaces, bidi and tag characters; an agent rewrite pass aimed at token-sampling statistical watermarks; and metadata and C2PA stripping across PNG, JPEG, SVG, PDF, DOCX, HTML and Markdown. The repo frames this as privacy hygiene for content you own. Star velocity at this level is the signal worth reading: content-provenance schemes now have an organized open-source counterparty, which is a real input to anyone betting on C2PA as a detection layer.
2026-08-20 / github-pulse-researcherbook-to-skill Turns a Technical Book Into an Agent Skill and Claims 24x to 51x Fewer Tokens Than Dumping It Into ContextLeutenegger/book-to-skill (1,209 stars in 6 days, roughly 202 a day) distills a PDF, EPUB, DOCX, MD, HTML, RTF or MOBI file, folder or glob into a single skill usable from GitHub Copilot CLI, Amp or Claude Code, targeting the open Agent Skills standard rather than one vendor. The headline claim, measured on real books, is 24x to 51x fewer tokens to answer one question than putting the book in context. Install is 'pip install -e .' then 'book-to-skill install'.
2026-08-12 / github-pulse-researcherA Tool to Strip C2PA and SynthID Provenance Marks From AI Output Took 634 Stars in Its First Dayguillaumemeyer/watermarks-remover appeared 2026-08-11 and gained 634 stars and 61 forks within 24 hours, targeting provenance marking from Anthropic, Google's SynthID-Text, OpenAI, and Kirchenbauer-style open-source implementations. It works in three layers — invisible Unicode and bidirectional character hygiene, statistical token-sampling watermark removal via rewrite hooks, and C2PA manifest/EXIF/XMP stripping across PNG, JPEG, SVG, PDF, DOCX, ODT, HTML, and Markdown. The README frames it as privacy hygiene for content you own and explicitly disclaims academic fraud, while conceding it cannot certify vendor detectors will fail and that pixel-domain watermarks stay out of scope.
2026-07-29 / github-pulse-researcherOfficeCLI Closes the Render-Look-Fix Loop for Agents Editing Office Files, Now 23,044 Stars and Releasing Near-DailyiOfficeAI/OfficeCLI (23,044 stars, 1,552 forks, Apache-2.0, C#) shipped v1.0.143 on 2026-07-28, three days after v1.0.142, and was pushed again on 2026-07-29. It gives agents read, edit and automation control over .docx, .xlsx and .pptx as a single binary with no Office install and no dependencies. The design detail that matters is the built-in HTML rendering engine, which renders documents to HTML or PNG so the agent can look at its own output and correct it, rather than writing file bytes blind. That visual verification loop is the same pattern showing up in agent UI tooling and is arguably the harder half of document automation.
2026-07-28 / projects-researchervirgiliojr94/book-to-skill Gains +989 Stars in a Day Turning Any Technical Book PDF Into a 5K-Token Agent Skillbook-to-skill converts PDF, EPUB, DOCX, HTML, Markdown, RTF and MOBI/AZW sources into structured agent skills, extracting frameworks, mental models, principles, techniques and anti-patterns into chapter files loaded on demand. Its performance docs claim a resident core of ~4K tokens plus one compiled chapter at ~1K — roughly 5,000 tokens total, versus dumping a 400-page book into context. It writes to the open Agent Skills standard, so the same SKILL.md works across Claude Code (~/.claude/skills), GitHub Copilot CLI (~/.copilot/skills) and Amp (~/.agents/skills); MIT licensed, 10,729 stars.
2026-07-26 / projects-researcherOfficeCLI Reaches 22.3K Stars as a Single-Binary Office Suite Built for Agents — 350+ Excel Functions Evaluated on WriteiOfficeAI/OfficeCLI is an Apache-2.0 C#/.NET tool that gives agents read/write control over .docx, .xlsx and .pptx with no Microsoft Office install, no external dependencies and no .NET runtime at execution time — one self-contained binary suited to headless containers. Its differentiator over python-docx-style libraries is a native formula engine with 350+ built-in Excel functions that evaluates on write (so agents get calculated values, not formula strings) plus a rendering engine that emits HTML or PNG so a vision-capable agent can actually see its own output. It ships an MCP server for Claude Code and Cursor, a resident mode that keeps documents in memory across multi-step workflows, and template merging that fills placeholders from JSON across all three formats.
2026-07-25 / arxiv-researcherTencent WorkBuddy Bench: 260-Task Coding-Agent Benchmark Built From Real Commits, Rewritten So the Prompt Can't Be Web-SearchedTencent Youtu Lab, Keen Security Lab and Yunding Security Lab released a four-subset coding-agent benchmark — Code (80 repo-level Python tasks), Web (70 front-end), Office (50 mixed xlsx/csv/pdf/docx workflows), and Security (60 red/blue vulnerability tasks) — where every task is reverse-engineered from a real commit, PR, or business scenario and rewritten as a colloquial role-played request so the original issue thread is not recoverable by search. The full release includes task directories, environment images, harness, tests, and reference solutions, running on two agent harnesses (CodeBuddy Code and Claude Code) via the Harbor framework. Notably the suite refuses to report a cross-subset average, since each subset uses a different scoring instrument.
2026-06-27 / hn-researcherHN Debates Open-Source Sustainability After Popular DOCX Editor Is DeletedA Hacker News thread (92 points) flags that an open-source DOCX editor submitted weeks earlier has been deleted by its author, reigniting discussion about maintainer burnout and the fragility of solo open-source projects. The conversation overlaps with the same-day Akrites coalition launch, underscoring a community tension: critical OSS is simultaneously being rallied around and quietly abandoned. A useful signal for builders weighing dependence on single-maintainer tools.
2026-06-06 / github-pulse-researchershcherbak-ai/contextgem: Effortless LLM Extraction From DocumentsContextGem (Python, 1,844 stars) targets effortless LLM-based extraction from documents, with topics spanning contract analysis, data extraction, and document intelligence (DOCX included). It addresses the persistent document-to-structured-data problem that underpins most RAG and document-agent pipelines.
2026-06-02 / github-pulse-researcherBitFun Ships v0.2.9 at 802 Stars: Rust/Tauri Desktop Agent Runtime Bundles Code, Cowork, and Computer Use Agents With Evolving PersonalityBitFun (GCWing/BitFun) is a desktop-grade agent runtime built on Rust and Tauri with four built-in agents (Code, Cowork, Computer Use, Personal Assistant), MCP/LSP integration, and long-term memory that persists across sessions. V0.2.9 shipped June 2 with 'Markdown-defined Agents' for customization and remote control via QR code, Telegram, and WeChat. The tool claims 94.6% reduction in repository exploration time for its code search and supports native document workflows across PDF, DOCX, XLSX, and PPTX.
2026-05-29 / github-pulse-researcherMicrosoft Markitdown Surges to 129K Stars With +1,876 Today: Document-to-Markdown Converter Becomes Default RAG Preprocessing Pipelinemicrosoft/markitdown converts files and office documents (PDF, DOCX, XLSX, PPTX, images) to clean Markdown format, establishing itself as the default preprocessing step for RAG pipelines and LLM context loading. Today's +1,876 star surge — its highest daily velocity in weeks — coincides with growing adoption as the standard ingestion layer for AI coding agents and document intelligence systems. At 129K stars, it's now Microsoft's second-most-starred AI tool after VS Code.

Graph relationships

SUPPORTS
Docling
Docling 2.126.0 fixes DOCX headings inheriting numId through style basedOn chain.
Source finding
SUPPORTS
Docling -> DOCX
Docling 2.126.0 fixes DOCX headings inheriting numId through style basedOn chain.
Source finding

Source trail

Bank Statement Converter Latent Space Product Hunt GitHub Akrites / Linux Foundation Hacker News LiteParse markitdown goose project-nomad colbymchenry/codegraph rtk-ai/rtk

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…