docling Releases

This release introduces new border styles for list rendering, including an 'inline' style for headers and footers, and a new 'dashed' border style. It also improves Vim integration by handling popup window resizing and fixes several keyboard protocol bugs.

Apr 26, 2026

Enhancements include VML image extraction from DOCX, strengthened input validation for METS-GBS processing, and various fixes to VLM chat templates, HTML handling, and PPTX notes assignment.

Apr 23, 2026

This release focuses on significantly improving document embedding performance and user experience by introducing streaming status reporting. It also adds native integrations for Gmail, Outlook, and Google Calendar, along with various UI improvements and bug fixes.

Apr 22, 2026

The release implements GraniteVisionTableStructureModel for VLM-based table extraction and addresses various LaTeX and DOCX parsing edge cases, including nested formatting and list items with formulas.

Apr 17, 2026

Adds explicit TikZ environment handling in the LaTeX backend and fixes RapidOCR asset alignment, list state isolation in table cells, and performance optimizations in Markdown debug logs.

Apr 16, 2026

Introduces a client SDK for the docling serve service and adds support for rapidocr 3.8 mobile model naming, alongside documentation for agent skill bundles.

Apr 13, 2026

Adds Nanonets OCR2 onboarding, explicit MLX support for OCR presets, and VLM tool-calling API support, along with fixes for Transformers v5 compatibility and DOCX SDT references.

Apr 13, 2026

Supports GraniteVision v4 and adds signature/stamp HTML blocks to the DC document, alongside a new PARTIAL_SUCCESS status for VLM pipeline pages.

Apr 10, 2026

Adds support for Falcon-OCR and LightOnOCR-2-1B models, along with improvements to LaTeX custom macro expansion.

Apr 7, 2026

fzf v0.71.0 adds the `--popup` alias for `--tmux` with Zellij support, cross-reload item identity tracking with `--id-nth`, and significant performance improvements.

Apr 4, 2026

This release introduces automatic tool calling, intelligent tool selection, new agents (Filesystem and Document Generation), and Telegram bot integration for AnythingLLM.

Apr 2, 2026

The 'Nightvision' release introduces a diff-based improvement engine, a single-port Docker architecture, and UI enhancements like drag-and-drop reordering.

Apr 2, 2026

Introduces GLM OCR support, upgrades the DocumentFigureClassifier model to v2.5, and removes the deprecation of extraction.

Apr 1, 2026

Upgrades to transformers v5, adds an OCR model for remote KServe v2 API, and fixes PDF hyperlink propagation, Excel last-row bounds, and LaTeX macro parsing.

Mar 31, 2026

Introduces an HTML backend using a headless browser and fixes critical bugs in OMML LaTeX output, PDFium resource management, and VLM content filtering.

Mar 25, 2026

Routes plain-text and Quarto/R Markdown files to the Markdown backend while addressing thread safety issues and list item extraction errors.

Mar 20, 2026

A UI-focused release featuring a new prompt input, improved citations, agent call metrics, and the addition of Perplexity Search as a web search provider.

Mar 18, 2026

Adds CUDA graph mode support for the VLLM engine to optimize inference performance.

Mar 14, 2026

Enhances XBRL support with metadata and linkbase relationships, and improves table structure detection using TableFormer v2.

Mar 12, 2026

Adds TableFormer v2 support and gRPC transport for KServe v2 API, alongside fixes for HTML table complexity and LaTeX parsing.

Mar 10, 2026