Docling Studio

Project website: https://pjmalandrino.github.io/docling-studio-landing/ · Live demo: https://pier-jean-docling-studio.hf.space/ · Repository: https://github.com/scub-france/Docling-Studio Launched publicly 31 of march 2026, after an 8-month POC phase. Current traction: 85+ GitHub stars, 410+ package downloads, live demo on Hugging Face Spaces. Third-party editorial coverage: technical article "Where did your RAG go wrong, and can you fix it?" on PunchTape (https://punchtape.substack.com/p/where-did-your-rag-go-wrong-and-can). DZone article: launch feature and architecture deep-dive "Designing Docling Studio" (https://dzone.com/articles/designing-docling-studio). Launch announcement on Medium (https://medium.com/scub-lab/introducing-docling-studio-see-what-docling-sees-bb43fbcefa8b). Community relay by independent practitioners on LinkedIn: Academic use case (RAG/document analysis): https://www.linkedin.com/posts/sonia-tabti-phd-b6993835_vous-utilisez-docling-pour-analyser-vos-documents-ugcPost-7451255140642254848-WIGu Editorial / industry relay (#docling #rag #aiagents): https://www.linkedin.com/posts/rachelroumeliotis_docling-rag-aiagents-activity-7450563057455452161-d6SQ Community relay: https://www.linkedin.com/posts/aairom_httpslnkdinehw-kppr-introducing-docling-share-7445373116366839808-_Lrp Growing organically through the Docling ecosystem with adoption signals from academia, industry, and editorial communities.

Teams using Docling as their document extraction engine for RAG pipelines who need to debug what's actually happening inside their ingestion: visualize bounding boxes from OCR, inspect chunks before embedding, modify them directly when something is wrong, and push the corrected pipeline to their vector store. Built for researchers, engineering teams, and educators who don't want a black box between their PDFs and their vector store — and who value a single self-hosted Docker image over another SaaS dependency.

Vue 3 + TypeScript (Frontend), FastAPI + Python With Hexagonal Architecture / Ports & Adapters, SQLite, OpenSearch As First VectorStore Adapter (Protocol Based, Swappable), Sentence Transformers For Embeddings, Multi Arch Docker (Amd64/Arm64). Neo4j Graph Adapter In Active Development As Second VectorStore Implementation, Demonstrating The Hexagonal Architecture'S Extensibility., Neo4j, Vue 3 + TypeScript (frontend), FastAPI + Python with hexagonal architecture (backend), SQLite + OpenSearch (storage), Docling + sentence-transformers (RAG pipeline), packaged as a single multi-arch Docker image. CI/CD via GitHub Actions with Trivy security scans, 541+ tests across Pytest/Vitest/Karate. Deployable on Hugging Face Spaces or self-hosted. Neo4j is an available vector Store since last release 0.5.0.

Full project presentation: https://pjmalandrino.github.io/docling-studio-landing/ Ecosystem engagement (IBM Research / LF AI & Data): Launch announcement publicly engaged by Peter Staar (TSC chair, Docling / LF AI & Data Foundation) and members of the IBM Research Docling team — https://www.linkedin.com/posts/pier-jean-malandrino_docling-opensource-architecture-activity-7448023570561064960-zZlu. Roadmap alignment with ecosystem direction under discussion. Academic adoption: PhD researcher publicly documented Docling Studio's usefulness for analyzing documents in RAG contexts, reaching her professional and student network — https://www.linkedin.com/posts/sonia-tabti-phd-b6993835_vous-utilisez-docling-pour-analyser-vos-documents-ugcPost-7451255140642254848-WIGu Editorial / industry relay: Public relay on LinkedIn tagged #docling #rag #aiagents, signaling editorial attention from tech-publishing circles — https://www.linkedin.com/posts/rachelroumeliotis_docling-rag-aiagents-activity-7450563057455452161-d6SQ Third-party technical article: "Where did your RAG go wrong, and can you fix it?" — PunchTape, covering Docling Studio's approach to RAG pipeline debugging: https://punchtape.substack.com/p/where-did-your-rag-go-wrong-and-can Industry evaluation: Evaluated internally by ThinkDeep AI / Isiadoc /APL Logistics / La Banque Postale for debugging their Docling pipelines. Some of them expressed intent to contribute upstream. Community endorsement: Independent LinkedIn relay — https://www.linkedin.com/posts/aairom_httpslnkdinehw-kppr-introducing-docling-share-7445373116366839808-_Lrp Unsolicited positive feedback on Reddit describing the tool as solving "mystery RAG failures" and praising the single-Docker-image approach. Product signals: 65+ GitHub stars, 410+ Docker package downloads, live public demo on HF Spaces, comprehensive CI/CD (multi-arch builds, Trivy security scans, Karate E2E tests).