docling

Feature

• cli: Add option for html with split-page mode (#1355) (c0ba88e)
• xlsx: Create a page for each worksheet in XLSX backend (#1332) (eef2bde)
• OllamaVlmModel for Granite Vision 3.2 (#1337) (c605edd)

Fix

• deps: Widen typer upper bound (#1375) (7e40ad3)
• Auto-recognize .xlsx, .docx and .pptx files (#1340) (0de70e7)
• docx: Declare image_data variable when handling pictures (#1359) (415b877)
• Implement PictureDescriptionApiOptions.bitmap_area_threshold (#1248) (2503999)
• Properly address page in pipeline _assemble_document when page_range is provided (#1334) (6b696b5)

Feature

• Handle <code> tags as code blocks (#1320) (0499cd1)
• docx: Add text formatting and hyperlink support (#630) (bfcab3d)

Fix

• docx: Adding new latex symbols, simplifying how equations are added to text (#1295) (14e9c0c)
• pptx: Check if picture shape has an image attached (#1316) (dc3bf9c)
• docx: Improve text parsing (#1268) (d2d6874)
• Tesseract OCR CLI can't process images composed with numbers only (#1201) (b3d111a)

Documentation

• Add plugins docs (#1319) (2e99e5a)
• Add visual grounding example (#1270) (71148eb)

Fix

• Fixes tables when using OCR (#1261) (7afad7e)

Fix

• Word-level pdf cells for tables (#1238) (8bd71e8)

Fix

• Improve HTML layer detection, various MD fixes (#1241) (9210812)
• html: Fix HTML parsed heading level (#1244) (85c4df8)

Fix

• converter: Cache same pipeline class with different options (#1152) (825b226)
• debug: Missing translation of bbox to to_bounding_box (#1220) (6df8827)
• docx: Identifying numbered headers (#1231) (f739d0e)

Documentation

• examples: Batch conversion doc raises_on_error (#1147) (0974ba4)

Feature

• SmolDocling: Support MLX acceleration in VLM pipeline (#1199) (1c26769)
• Add PPTX notes slides (#474) (b454aa1)
• Updated vlm pipeline (with latest changes from docling-core) (#1158) (2f72167)

Fix

• Determine correct page size in DoclingParseV4Backend (#1196) (f5adfb9)
• msword: Fixing function return in equations handling (#1194) (0b707d0)

Documentation

• Linux Foundation AI & Data (#1183) (1d680b0)
• Move apify to docs (#1182) (54a78c3)

Feature

• Add factory for ocr engines via plugins (#1010) (6eaae3c)
• Add DoclingParseV4 backend, using high-level docling-parse API (#905) (3960b19)
• actor: Docling Actor on Apify infrastructure (#875) (772487f)
• Equations to latex in MSWord backend (with inline groups) (#1114) (6eb718f)

Fix

• html: Handle nested empty lists (#1154) (f94da44)
• Use first table row as col headers (#1156) (0945973)
• Pass tests, update docling-core to 2.22.0 (#1150) (aa92a57)

Documentation

• Fix spelling of picture in usage (#1165) (7e01798)

Feature

• Use new TableFormer model weights and default to accurate model version (#1100) (eb97357)

Fix

• CLI: Fix help message for abort options (#1130) (4d64c4c)

Documentation

• Add description of DOCLING_ARTIFACTS_PATH env var (#1124) (e1c49ad)

Performance

• New revision code formula model and document picture classifier (#1140) (5e30381)

Fix

• Proper handling of orphan IDs in layout postprocessing (#1118) (c56ab3a)

Documentation

• Enrichment models (#1097) (357d41c)

Fix

• Enable locks for threadsafe pdfium (#1052) (8dc0562)
• html: Use 'start' attribute when parsing ordered lists from HTML docs (#1062) (de7b963)

Documentation

• Improve docs on token limit warning triggered by HybridChunker (#1077) (db3ceef)

Feature

• [Experimental] Introduce VLM pipeline using HF AutoModelForVision2Seq, featuring SmolDocling model (#1054) (3c9fe76)
• cli: Add option for downloading all models, refine help messages (#1061) (ab683e4)

Fix

• Vlm using artifacts path (#1057) (e197225)
• html: Parse text in div elements as TextItem (#1041) (1b0ead6)

Documentation

• Extend chunking docs, add FAQ on token limit (#1053) (c84b973)

Feature

• Implement new reading-order model (#916) (c93e369)

Fix

• Runtime error when Pandas Series is not always of string type (#1024) (6796f0a)

Documentation

• Revamp picture description example (#1015) (27c0400)

Feature

• Support cuda:n GPU device allocation (#694) (77eb77b)
• xml-jats: Parse XML JATS documents (#967) (428b656)

Fix

• Revise DocTags, fix iterate_items to output content_layer in items (#965) (6e75f0b)

Feature

• Add support for CSV input with new backend to transform CSV files to DoclingDocument (#945) (00d9405)
• Introduce the enable_remote_services option to allow remote connections while processing (#941) (2716c7d)
• Allow artifacts_path to be defined as ENV (#940) (5101e25)

Fix

• Update Pillow constraints (#958) (af19c03)
• Fix the initialization of the TesseractOcrModel (#935) (c47ae70)

Documentation

• Update example Dockerfile with download CLI (#929) (7493d5b)
• Examples for picture descriptions (#951) (2d66e99)

Feature

• Add content_layer property to items to address body, furniture and other roles (#735) (cf78d5b)

Feature

• Describe pictures using vision models (#259) (4cc6e3e)

Fix

• Remove unused httpx (#919) (c18f47c)

Feature

• New artifacts path and CLI utility (#876) (ed74fe2)

Fix

• markdown: Handle nested lists (#910) (90b766e)
• Test cases for RTL programmatic PDFs and fixes for the formula model (#903) (9114ada)
• msword_backend: Handle conversion error in label parsing (#896) (722a6eb)
• Enrichment models batch size and expose picture classifier (#878) (5ad6de0)

Documentation

• Introduce example with custom models for RapidOCR (#874) (6d3fea0)

Feature

• Expose equation exports (#869) (6a76b49)
• Add option to define page range (#852) (70d68b6)
• docx: Support of SDTs in docx backend (#853) (d727b04)
• Python 3.13 support (#841) (4df085a)

Fix

• markdown: Fix parsing if doc ending with table (#873) (5ac2887)
• markdown: Add support for HTML content (#855) (94751a7)
• docx: Merged table cells not properly converted (#857) (0cd81a8)
• Processing of placeholder shapes in pptx that have text but no bbox (#868) (eff16b6)
• KeyError in tableformer prediction (#854) (b1cf796)
• Fixed docx import with headers that are also lists (#842) (2c037ae)
• Use new add_code in html backend and add more typing hints (#850) (2a1f8af)
• markdown: Fix empty block handling (#843) (bccb022)
• Fix for the crash when encountering WMF images in pptx and docx (#837) (fea0a99)

Documentation

• Updated the readme with upcoming features (#831) (d7c0828)
• Add example for inspection of picture content (#624) (f9144f2)

Feature

• CLI: Expose code and formula models in the CLI (#820) (6882e6c)
• Add platform info to CLI version printout (#816) (95b293a)
• ocr: Expose rec_keys_path in RapidOcrOptions to support custom dictionaries (#786) (5332755)
• Introduce automatic language detection in TesseractOcrCliModel (#800) (3be2fb5)

Fix

• Fix single newline handling in MD backend (#824) (5aed9f8)
• Use file extension if filetype fails with PDF (#827) (adf6353)
• Parse html with omitted body tag (#818) (a112d7a)

Documentation

• Document Docling JSON parsing (#819) (6875913)
• Add SSL verification error mitigation (#821) (5139b48)
• backend XML: Do not delete temp file in notebook (#817) (4d41db3)
• Typo (#814) (8a4ec77)
• Added markdown headings to enable TOC in github pages (#808) (b885b2f)
• Description of supported formats and backends (#788) (c2ae1cc)

Feature

• New document picture classifier (#805) (16a218d)
• Add Docling JSON ingestion (#783) (88a0e66)
• Code and equation model for PDF and code blocks in markdown (#752) (3213b24)
• Add "auto" language for TesseractOcr (#759) (8543c22)

Fix

• Added extraction of byte-images in excel (#804) (a458e29)
• Update docling-parse-v2 backend version with new parsing fixes (#769) (670a08b)

Documentation

• Fix minor typos (#801) (c58f75d)
• Add Azure RAG example (#675) (9020a93)
• Fix links between docs pages (#697) (c49b352)
• Fix correct Accelerator pipeline options in docs/examples/custom_convert.py (#733) (7686083)
• Example to translate documents (#739) (f7e1cbf)

Fix

• Improve OCR results, stricten criteria before dropping bitmap areas (#719) (5a060f2)
• Allow earlier requests versions (#716) (e64b5a2)

Documentation

• Add pointers to LangChain-side docs (#718) (9a6b5c8)
• Add LangChain docs (#717) (4fa8028)

Feature

• Added http header support for document converter and cli (#642) (0ee849e)

Fix

• Correct scaling of debug visualizations, tune OCR (#700) (5cb4cf6)
• Let BeautifulSoup detect the HTML encoding (#695) (42856fd)
• mspowerpoint: Handle invalid images in PowerPoint slides (#650) (d49650c)

Documentation

• Specify docstring types (#702) (ead396a)
• Add link to rag with granite (#698) (6701f34)
• Add integrations, revamp docs (#693) (2d24fae)
• Add OpenContracts as an integration (#679) (569038d)
• Add Weaviate RAG recipe notebook (#451) (2b591f9)
• Document Haystack & Vectara support (#628) (fc645ea)

Feature

• Create a backend to transform PubMed XML files to DoclingDocument (#557) (fd03480)

Feature

• Updated Layout processing with forms and key-value areas (#530) (60dc852)
• Create a backend to parse USPTO patents into DoclingDocument (#606) (4e08750)
• Add Easyocr parameter recog_network (#613) (3b53bd3)

Documentation

• Add Haystack RAG example (#615) (3e599c7)
• Fix the path to the run_with_accelerator.py example (#608) (3bb3bf5)

Feature

• Introduce support for GPU Accelerators (#593) (19fad92)

Feature

• Add timeout limit to document parsing job. DS4SD#270 (#552) (3da166e)

Fix

• Do not import python modules from deepsearch-glm (#569) (aee9c0b)
• Handle no result from RapidOcr reader (#558) (f45499c)
• Make enum serializable with human-readable value (#555) (a7df337)

Documentation

• Update chunking usage docs, minor reorg (#550) (d0c9e8e)

Feature

• Docling-parse v2 as default PDF backend (#549) (aca57f0)

Fix

• Call into docling-core for legacy document transform (#551) (7972d47)
• Introduce Image format options in CLI. Silence the tqdm downloading messages. (#544) (78f61a8)

Feature

• Expose new hybrid chunker, update docs (#384) (c8ecdd9)
• MS Word backend: Make detection of headers and other styles localization agnostic (#534) (3e073df)

Fix

• Correcting DefaultText ID for MS Word backend (#537) (eb7ffcd)
• Add py.typed marker file (#531) (9102fe1)
• Enable HTML export in CLI and add options for image mode (#513) (0d11e30)
• Missing text in docx (t tag) when embedded in a table (#528) (b730b2d)
• Restore pydantic version pin after fixes (#512) (c830b92)
• Folder input in cli (#511) (8ada0bc)

Documentation

• Document new integrations (#532) (e780333)

Fix

• Improve handling of disallowed formats (#429) (34c7c79)

Fix

• ParserError EOF inside string (#470) (#472) (c90c41c)
• PermissionError when using tesseract_ocr_cli_model (#496) (d3f84b2)

Documentation

• Add styling for faq (#502) (5ba3807)
• Typo in faq (#484) (33cff98)
• Add automatic api reference (#475) (d487210)
• Introduce faq section (#468) (8ccb3c6)

Performance

• Prevent temp file leftovers, reuse core type (#487) (051789d)

Fix

• cli: Expose debug options (#467) (dd8de46)
• Remove unused deps (#466) (af63818)

Documentation

• Extend integration docs & README (#456) (84c46fd)

Feature

• ocr: Added support for RapidOCR engine (#415) (85b2999)

Fix

• Use correct image index in word backend (#442) (767563b)
• Update tests and examples for docling-core 2.5.1 (#449) (29807a2)

Fix

• Fixes for wordx (#432) (d0a1180)
• Force pydantic < 2.10.0 (#407) (d7072b4)

Documentation

• Add DocETL, Kotaemon, spaCy integrations; minor docs improvements (#408) (7a45b92)

Feature

• Add support for ocrmac OCR engine on macOS (#276) (6efa96c)

Fix

• Python3.9 support (#396) (7b013ab)
• Propagate document limits to converter (#388) (32ebf55)

Feature

• Added support for exporting DocItem to an image when page image is available (#379) (3f91e7d)
• Expose ocr-lang in CLI (#375) (ed785ea)
• Added excel backend (#334) (926dfd2)
• Extracting picture data for raster images found in PPTX (#349) (7a97d71)

Fix

• Fixing images in the input Word files (#330) (8533039)
• Reduce logging by keeping option for more verbose (#323) (8b437ad)

Documentation

• Fixed typo in v2 example v2 (#378) (911c3bd)
• Add automatic generation of CLI reference (#325) (ca8524e)
• Add architecture outline (#341) (25fd149)
• Fix parameter in usage.md (#332) (835e077)

Fix

• Skip glm model downloads (#322) (c9341bf)

Fix

• Handling of single-cell tables in DOCX backend (#314) (fb8ba86)

Documentation

• Hybrid RAG with Qdrant (#312) (7f5d35e)
• Add Data Prep Kit integration (#316) (93fc1be)

Feature

• OCR: Introduce the OcrOptions.force_full_page_ocr parameter that forces a full page OCR scanning (#290) (c6b3763)

Fix

• Configure env prefix for docling settings (#315) (5d4a10b)
• Added handling of grouped elements in pptx backend (#307) (81c8243)
• Allow mps usage for easyocr (#286) (97f214e)

Documentation

• Add navigation indices (#305) (1239ade)

Fix

• EasyOcrModel: Support the use_gpu pipeline parameter in EasyOcrModel. Initialize easyocr (#282) (0eb065e)

Fix

• tesserocr: Raise Exception if tesserocr has not loaded any languages (#279) (704d792)
• Dockerfile example copy command (#234) (90836db)

Documentation

• Update badges & credits (#248) (a84ec27)
• Add coming-soon section (#235) (5ce02c5)
• Add artifacts-path param to CLI (#233) (d5e65ae)

Feature

• Pdf backend, table mode as options and artifacts path (#203) (40ad987)

Documentation

• Add explicit artifacts path example (#224) (eeee3b4)
• Update custom convert and dockerfile (#226) (5f5fea9)
• Correct spelling of 'individual' (#219) (41acaa9)
• Update LlamaIndex docs (#196) (244ca69)

Fix

• Simplify torch dependencies and update pinned docling deps (#190) (eb679cc)
• Allow to explicitly initialize the pipeline (#189) (904d24d)

Feature

• Add pipeline timings and toggle visualization, establish debug settings (#183) (2a2c65b)

Fix

• Fix duplicate title and heading + add e2e tests for html and docx (#186) (f542460)

Fix

• Fix header levels for DOCX & HTML (#184) (b9f5c74)
• Handling of long sequence of unescaped underscore chars in markdown (#173) (94d0729)
• HTML backend, fixes for Lists and nested texts (#180) (7d19418)
• MD Backend, fixes to properly handle trailing inline text and emphasis in headers (#178) (88c1673)

Documentation

• Update LlamaIndex docs for Docling v2 (#182) (2cece27)
• Fix batch convert (#177) (189d3c2)
• Add export with embedded images (#175) (8d356aa)

Feature

• Update to docling-parse v2 without history (#170) (4116819)
• Support AsciiDoc and Markdown input format (#168) (3023f18)

Fix

• Set valid=false for invalid backends (#171) (3496b48)

Feature

• Add coverage_threshold to skip OCR for small images (#161) (b346faf)

Fix

• Fix legacy doc ref (#162) (63bef59)

Documentation

• Typo fix (#155) (f799e77)
• Add graphical band in readme (#154) (034a411)
• Add use docling (#150) (61c092f)

Feature

• Docling v2 (#117) (7d3be0e)

Breaking

• Docling v2 (#117) (7d3be0e)

Documentation

• Introduce docs site (#141) (d504432)

Feature

• New experimental docling-parse v2 backend (#131) (5e4944f)

Fix

• Remove stderr from tesseract cli and introduce fuzziness in the text validation of OCR tests (#138) (dae2a3b)

Documentation

• Simplify LlamaIndex example using Docling extension (#135) (5f1bd9e)

Feature

• Add options for choosing OCR engines (#118) (f96ea86)

Feature

• New torch-based docling models (#120) (2422f70)

Feature

• Windows support (#122) (d44c62d)

Fix

• Allow usage of opencv 4.6.x (#110) (34bd887)

Documentation

• Document chunking (#111) (c05b692)

Feature

• Support tableformer model choice (#90) (d6df76f)

Feature

• Add figure in markdown (#98) (6a03c20)

Feature

• Add URL support to CLI (#99) (3c46e42)

Fix

• Fix OCR setting for pypdfium, minor refactor (#102) (d96b96c)

Documentation

• Document CLI, minor README revamp (#100) (f8f2303)

Fix

• Updated the render_as_doctags with the new arguments from docling-core (#93) (4794ce4)