Ozgar AI Release Notes

🧠 GPT-5 Support

Ozgar AI now supports GPT-5 for both the AI assistant and automated documentation generation, delivering significantly improved reasoning, contextual understanding, and higher-quality insights across complex enterprise systems. The new recommended Ozgar model settings are as follows:

We will provide updated guidance on recommended model usage without requiring a new release, allowing you to benefit from newly available models (e.g., GPT-5.4).

For "Knowledge Model" we still recommend also the gpt-4o-mini model, offering the best balance of speed, output quality, and overall efficiency. For customers who prefer to prioritize maximum quality over efficiency, gpt-5.4-mini is also supported, though current testing showed no significant improvement in generated knowledge for most scenarios.

Supported Assistant models in v1.4.5:

• GPT-4.1 - fast, cost-efficient
• GPT-5.2 - balanced
• GPT-5.4 - highest quality, higher cost

This helps optimize performance, cost, and quality as new models become available.

🤖 New AI Assistant Pipeline

A completely redesigned assistant pipeline improves accuracy, completeness, and reasoning reliability.

• Inference & reasoning validation loop to verify responses before generating outputs
• Improved Source Code Agent specialized in analyzing large enterprise codebases
• Migration to the LangChain4J framework for improved orchestration and scalability

These changes substantially improve how Ozgar processes complex questions about code, systems, and dependencies.

🔎 Source File Intelligence

Ozgar now provides deep intelligence across source files, enabling accurate answers for detailed questions about your codebases.

• Query and list programs, source files, tables, display files, and other objects across libraries and packages
• Answering questions that need pattern and text search across program sources (RPG, CL, etc.)
• Answering questions related to modifications in source files

📊 Programs Flow Visualizations

The new programs flow visualization provide graphical insights into program relationships, helping developers quickly understand program dependencies.

🔄 Data Refresh Updates

Ozgar now supports more flexible and efficient data refresh workflows:

Incremental refresh with hourly update intervals

Incremental refresh with hourly update intervals

This allows to keep Ozgar's knowledge graph and system metadata continuously up to date while minimizing processing overhead.

🔌 New Integrations

Git Connector

A new Git connector and configuration UI allows Ozgar to directly integrate with Git repositories, enabling seamless ingestion of modern development environments.

JDBC Connector for PostgreSQL

Ozgar now supports PostgreSQL via JDBC, enabling direct analysis of database structures and objects.

MCP Support

Ozgar now supports the Model Context Protocol (MCP), enabling integration with external AI tools and clients for advanced AI workflows.

The MCP Client Setup Guide for Claude Code, Claude Cowork, Codex, and Github Copilot is available here.

🧩 Expanded Components Coverage

New component pages are now available for:

• User-defined database functions
• Stored procedures
• Database triggers

These components now appear in the Components section of the documentation outline, providing a more complete view of application logic.

🌍 Localization

Ozgar AI now supports Dutch language, expanding accessibility for teams working in Dutch-speaking environments.

🛠 Platform Improvements

Chatbot Personalization

Assistant conversations can now be personalized by user type, improving relevance and usability across different roles such as developers, analysts, and architects.

Improved Source File View

Source file pages now include summary information, making it easier to quickly understand file purpose and context.

🐞 Bug Fixes

Fixed an issue that could cause chatbot conversations to disappear.