Creating AI-Enhanced Document Management with the Docker GenAI Stack

Introduction

In this article, we will discuss the Alfresco Gen AI project, which utilizes a Docker stack alongside Alfresco to create an AI-driven content service platform and document management system. The primary aim of this project is to enhance documents with generative AI features, including summarization, classification, prompting, and image description.

Project Overview

The Alfresco Gen AI project leverages a Docker-based setup with the following technologies and tools:

Docker: The environment runs locally with a minimum of 20GB of RAM allocated for the large language models (LLMs).
Java 17: Required for some components of the stack.
Alfresco: Serves as a sample content service platform and document management solution.

Core Features

Document Summarization: It allows users to summarize documents, even those in different languages. The output summary can be configured to be in English, regardless of the original document's language.
Classification: This feature classifies the document based on a predefined list of terms, automatically selecting the most appropriate category.
Prompting: Users can ask questions regarding the contents of a document, and the model returns relevant information based on the content.
Image Description: It can describe images and graphics, providing textual representation for visual contents.

Setup and Requirements

To set up the project, one must clone the repository using git clone and check that:

Docker is running.
Java 17 is installed.
The Docker environment has sufficient memory (minimum 20GB).

Once the necessary checks are completed, the GenAI services can be started using the Docker Compose setup. The services expose their APIs on specified ports.

Sample API Usage

Once the services are up and running, users can test the various features via command-line API calls. Here are some examples of how each functionality can be invoked:

Summarizing a Document: Users can POST a document to the summarization endpoint and will receive a summary in English.
Classifying a Document: Similarly, classification can be tested by specifying a term list for the model to choose from.
Prompting Information: Users can ask specific questions regarding document content and obtain relevant answers.
Describing Images: For images uploaded, users can call the description endpoint to get a textual description of the image.

Integration with Alfresco

To integrate these features with Alfresco, one must build additional components:

Alfresco AI Applier: Connects documents with Gen AI services to apply summarization.
Alfresco AI Listener: Triggers Gen AI features automatically when documents are uploaded or modified in the Alfresco environment.

By applying aspects like "Summarizable with AI" or "Describable with AI," users can set rules within their Alfresco repository. Automations allow for the classification of documents into appropriate folders and generating summaries or descriptions applicable to uploaded files.

Automation and Advanced Use Cases

The project includes automation scripts that allow for more sophisticated actions, such as moving documents into folders based on their classifications automatically. This streamlines document management workflows, making it easier for organizations to manage document lifecycles.

In summary, the Alfresco Gen AI project successfully demonstrates the potential of integrating generative AI into document management systems, providing enhanced capabilities and efficiencies.

Keywords

Alfresco
Gen AI
Docker
Document Management
Summarization
Classification
Prompting
Image Description
Automation
Large Language Models (LLMs)

FAQ

Q1: What technologies are used in the Alfresco Gen AI project?
A1: The project utilizes Docker, Java 17, and Alfresco as the core technologies.

Q2: What features are provided by the Gen AI services?
A2: The main features are document summarization, classification, prompting, and image description.

Q3: How can I test the Gen AI services?
A3: Users can test the services by making API calls to the provided endpoints once the Docker containers are running.

Q4: How does the integration with Alfresco work?
A4: Integration is done through the Alfresco AI Applier and Listener, which connect documents to Gen AI services, enabling automatic summarization and description capabilities.

Q5: Can I automate document management tasks?
A5: Yes, the project supports automation through rules and scripts that apply AI features upon document creation or updates.