Feedback on AI Integration in DBeaver Lite 25.3 — UX Gap and Architecture Limitations

[torin2175]

Dear DBeaver Development Team,

As a SQL Server DBA, I have recently tested DBeaver Lite 25.3.0 specifically to evaluate the new AI features and the integration with Google Gemini. Unfortunately, I must report that the current implementation significantly lags behind modern standards for AI-assisted development (such as those found in Cursor, VS Code, Augment, Claude or JetBrains AI).

Based on my professional experience, here are the critical issues that make the Lite version's AI features impractical for a high-performance workflow:

Lack of Contextual Code Mutation: The @ai command in the editor functions strictly as a "text-to-SQL" generation tool. It is completely unaware of the existing code surrounding the cursor and lacks the capability to refactor or modify selected code blocks.

AI Chat Isolation: The current workflow is inefficient, requiring manual copy-pasting of code into the chat window for any modifications. By 2026 standards, users expect a seamless "Edit with AI" experience where the assistant has direct, bi-directional access to the active editor.

High UX Friction:

AI-related functions are scattered across different menus, and default hotkeys often conflict with system-level shortcuts.

The "Smart Completion" feature is not intuitive and requires deep diving into settings to activate, which is unacceptable for a product marketed as "AI-powered".

Absence of an Agentic Approach: While modern IDEs utilize agents capable of analyzing multiple files and providing differential updates (diffs), DBeaver remains limited to basic API requests and manual insertions.

At this stage, using external AI agents (via browsers or specialized IDEs) is substantially faster and more effective than DBeaver’s built-in tools due to their superior context management and iterative editing capabilities.

I hope this feedback helps you reconsider the AI architecture in future releases to better serve professional database administrators and developers.

Best regards,
Alex

[mergisi]

This is incredibly valuable feedback, Alex! 👏

You've articulated the exact pain points many database professionals face with current AI integrations. The points about contextual code mutation and bi-directional editor access are spot-on — these are table stakes for a truly AI-native workflow in 2026.

Some observations:

1. Agentic vs. Static AI: You're right that tools like Cursor, VS Code Augment, and Claude have set a high bar. The ability for AI to understand schema context, previous queries, and refactor existing code is crucial.
2. UX Integration: AI features shouldn't feel bolted-on. They need to be woven into the natural workflow (like inline completions, not just chat windows).

I hope the DBeaver team takes this feedback seriously. The demand for seamless AI-assisted database development is only growing! 🚀