On This Page

elDoc is a Hong Kong software company with two document products. The first is elDoc IDP OSS, an open-source application under AGPL-3.0 that takes PDFs and scans through OCR, classification and LLM-based extraction to a human review screen and CSV export. The second is the proprietary elDoc platform, which adds document management, search, AI chat over company documents, workflows and eSignature. Only the first is open source.

How this profile came about: Maria of elDoc requested the listing through the contribute form on 20 November 2026. IDP Software wrote the profile from elDoc's website, its documentation, the GitHub repositories and the LinkedIn page. Figures marked as vendor claims have not been verified independently. About the author: Maria.

Company and ownership

The licence agreement names elDoc Limited as licensor under Hong Kong law. A Hong Kong government listing gives an office in Kwai Fong, New Territories. elDoc's own website has no About, Contact or legal page with an address. The LinkedIn page gives 2022 as founding year, 11 to 50 employees and Dublin as location, without a country. Collabora's partner page mentions offices in Hong Kong and the United States. The database Tracxn says elDoc was founded in 2016 by Lilia Kanevska and lists 12 employees; elDoc's pages name no founder or manager. Funding and ownership are not published.

Older blog posts deal with eSignature and document management, including an elDoc vs DocuSign comparison and a market entry in Mexico. The GenAI and IDP features came later. In Hong Kong, DMS Solutions sells and implements elDoc as "official redistributor and certified implementation partner".

Where elDoc fits / limits

The open-source module fits developers and small teams who want a complete extraction application with a review screen instead of assembling one from libraries such as Docling, Unstructured or PaddleOCR. The full platform fits organisations that want document management, extraction, search and eSignature from one vendor, on their own servers or air-gapped.

It is a weaker fit in these cases:

  • Production use of the open-source module as it ships. The README says the embedded MongoDB is "intended for local development and evaluation only". There is no API, no user management and no Docker image for the module; results leave the system as CSV through the web interface.
  • Mature open-source project. The repository was created on 6 November 2026 and had 3 commits from 1 contributor, 4 stars and no release on 20 November 2026. Older open-source document processing projects such as Unstract and DocRouter AI have a longer public history.
  • Fully local by default. The default configuration sends documents to OpenRouter, a cloud LLM service, and needs an API token. Local models and PaddleOCR are possible but must be configured.
  • Closed products built on the code. AGPL-3.0 requires that modified versions offered over a network publish their source. elDoc sells a commercial licence for other cases. Have your legal team read both.
  • The free Community Version for larger organisations. It is a separate, closed licence. Commercial use is allowed only for businesses with under 100,000 USD annual revenue; government agencies are excluded.
  • Published prices or accuracy. Neither is on the website.

How elDoc processes documents

The README describes the flow as "upload → understand → extract → review → export CSV". The architecture article and the README name these parts:

Step Open-source module Full platform adds
Input Machine-generated and scanned PDFs, common image formats, multipage TIFF Online preview and editing via Collabora Online
OCR Embedded Tesseract (default), PaddleOCR, vision-language models Google Cloud Vision
Classification and extraction LLM with tool calling and vision; optional field definitions per document type IDP plugins, Anchor and Regex processors
Human review Web interface for validating results Workflows and form designer
Output CSV download in the interface REST API, search (Solr full text, Qdrant vectors), AI chat
Administration One admin account Users, groups, Active Directory, multitenancy

The code is Java 21 on Jakarta EE (Payara Micro) with a PrimeFaces interface, MongoDB for storage and LangChain4j for the model calls. Supported model providers are OpenRouter, AWS Bedrock and OpenAI-compatible endpoints, which covers local servers such as Ollama or vLLM. The README itself recommends PaddleOCR over Tesseract for complex layouts and poor scans in production. The compose file of the full platform also starts Docling and Qdrant next to the elDoc container.

Open source and editions

elDoc has three ways to get the software, and the names are easy to confuse:

Edition Licence Who may use it Price
elDoc IDP OSS AGPL-3.0, plus commercial licence; contributors sign a CLA Anyone Free
Community Version (full platform) Closed elDoc Community License, access after a request is approved Businesses under 100,000 USD revenue in production; others for evaluation only Free
Enterprise, White Label Commercial Anyone On request

Code. The open-source module is at github.com/elDocAI/elDoc-idp-oss. Its version number 7.2026.4 matches the full product, but the README says this "does not imply feature parity". Builds exist for Windows and Linux; macOS is untested. The second repository, eldoc-genai-container, holds MIT-licensed compose files only. The images themselves come from elDoc's own registry with credentials from support.

Timeline. The code went public on 6 November 2026, the blog announcement followed on 8 November and the LinkedIn announcement on 19 November. That post names Andrew Grabovetskyi and Kostiantyn Kozlov as key contributors. The Community Version page still describes the open-source release as a plan. The homepage and the licence agreement do not mention it.

Technical specifications

Feature Specification
Document Types Invoices, bank statements, utility bills, payment slips, company registry forms, service reports, transcripts, handwriting (solution pages)
Input Formats PDF (digital and scanned), common image formats, multipage TIFF
Output Formats CSV (open-source module); API and integrations in the full platform
OCR Tesseract, PaddleOCR, vision-language models such as Qwen VL; Google Cloud Vision in the full platform
LLM Providers OpenRouter, AWS Bedrock, OpenAI-compatible endpoints, local models
API/Integration REST API with JWT (full platform only); Collabora Online, Active Directory
Deployment Options Open-source module: Java 21 on Windows or Linux; full platform: Rocky Linux 9 or RHEL 9, containers, on-premises, air-gapped, private or public cloud, SaaS
Hardware GPU only for self-hosted LLMs; smallest tier 96 GB total VRAM and 128 GB RAM
Certifications None published; ISO/IEC 27001, GDPR and NIST logos without certificate
Claimed Accuracy None published

Deployment, security and pricing

The full platform runs on Rocky Linux 9 or RHEL 9 with MongoDB 8.2 and Java 21; Solr, Collabora Online and LibreOffice are optional (software requirements). The hardware page says no GPU is needed when cloud LLM services are used. Hosting the language model in-house starts at 96 GB of total GPU memory and 128 GB of RAM and reaches 512 GB of GPU memory for the largest tier. Read this before you plan an air-gapped installation: the cost of the model servers can exceed the cost of the software.

There is no price list. The homepage offers proofs of concept and pilots. The homepage shows ISO/IEC 27001, GDPR and NIST logos; it does not say whether elDoc holds a certificate, who issued it or what it covers. Ask for the certificate. The privacy policy still refers to the EU-U.S. Privacy Shield, which the Court of Justice of the EU invalidated in 2020, and names no address. See the guide to self-hosted document processing for what to check in an on-premises setup.

Evidence

The open-source module can be checked by anyone: the code, licence and README are public, and the README states its own limits. Everything else is vendor statement. The homepage says elDoc is "Proven across enterprise, government, defence and highly regulated environments" and shows customer logos, but there is no case study, named project or quote. The "Document Retrieval Accuracy" counter on the homepage has no method. The partner DMS Solutions calls elDoc "The #1 GenAI Document Management Platform in Hong Kong" without naming a ranking. The listing on the Hong Kong government's Smart Government Innovation LAB is a catalogue entry, not a reference. Founding year (2016 or 2022) and team size come from LinkedIn and Tracxn and differ. Test the open-source module with a sample of your own documents before you talk about the full platform.

Resources

Company Information

elDoc Limited is a Hong Kong company; a government listing gives its office as Unit 2305, Tower 2, Metroplaza, Kwai Fong. LinkedIn shows 11 to 50 employees and founding in 2022; Tracxn says 2016. Funding, ownership and revenue are not published.