Build a DMS in Java with RAG for Enterprise Documents
We’ll add AI to a working document management system. You’ll see how it classifies documents, fills in fields automatically, and uses RAG to find relevant documents based on what each user is allowed to access. We’ll then walk through the architecture and Java code so you can apply the same approach in your own projects.
Register to receive the workshop recording by email.

An invoice arrives. You open the file, select the document type, and enter the amount, due date, and supplier. Looking for documents about farm equipment? You search for “tractor,” “combine harvester,” and “seed drill” one by one and hope you haven’t missed anything.
At the workshop, we’ll show how AI can identify document types, fill in fields automatically, find documents by meaning, and answer questions about their contents.
We’ll start with a working document management system and then open the Java code. You’ll see how we use Jmix AI Tools, build a vector index, prepare context for the model, and check access permissions before passing document content to the model.
The live demo and live coding use the same codebase. You’ll need some Java knowledge to follow the code.
AI-powered document classification
An assistant on the document details screen
Semantic search with RAG
Document access control
You can use the Jmix DocAI source code as a starting point and adapt it to your own document workflows.
Jmix DocAI source code
A RAG architecture that respects access permissions
A way to connect your own AI tools
Practical limitations of AI in document workflows
Jmix and the application architecture
We’ll introduce Jmix and walk through the DMS data model, UI, document storage, and access permissions.
Everyday document workflows
We’ll enter fields for an incoming document, look for an answer in a contract, and run a keyword search. Then we’ll look at where these workflows take extra time and which steps AI can help with.
Live demo of AI features in the DMS
We’ll classify incoming documents, fill in their fields, and use the assistant directly from the document details screen. You’ll see which manual steps these features can save.
Adding AI features in Java
We’ll open the code and use Jmix AI Tools to expose application functions to the model. You’ll see how the model receives data and calls those functions.
From document upload to a RAG answer
We’ll explain how the OpenSearch vector index supports semantic search. Then we’ll upload a document and show how it becomes searchable and available to the assistant when answering questions.
Access permissions and implementation issues
We’ll test the application with users who have different permissions, show how those permissions limit what goes into the model’s context, and discuss the issues we encountered during development.
Running the DMS yourself
We’ll show you where to get the source code and how to get started with the application. Then we’ll take your questions.
Defined application architecture
Structured project context
Jmix Studio
Jmix AI Assistant
Ready-to-use component
Java source code you own
Dmitriy Vaschenko
Community ManagerAlexey Burov
Head of Sales