Today, we’re introducing Guided Coverage Improvement—a breakthrough capability in the next generation of Diffblue Cover that transforms how development teams systematically achieve their coverage goals. Working in tandem with our new Test Asset Insights and LLM-Augmented Intelligence features, it creates the connection between your existing test infrastructure and the coverage levels your enterprise demands.
Guided Coverage Improvement: Your Strategic Coverage Advisor
Manually identifying areas for coverage improvement and crafting the necessary tests can be time-consuming and complex. It requires a deep understanding of the codebase, meticulous analysis of coverage reports, and the skill to write effective, targeted unit tests.
Traditional methods often involve:
- Prioritization Headaches: Deciding which uncovered lines or branches are most critical to address first.
- Manual Analysis: Sifting through code and coverage reports to understand why certain paths aren’t being tested.
- Input Construction Complexity: Finding the right test inputs, which may involve the complex construction of specific object instances, to cover certain uncovered code branches.
Introducing Guided Coverage Improvement
Diffblue Cover’s Guided Coverage Improvement is designed to address these challenges. Its purpose is clear: to significantly increase coverage beyond out-of-the-box generation, prioritize issues effectively, and provide step-by-step guidance and automation to get you there faster.
How it Works: A Seamless Collaboration
One of the most innovative aspects of Guided Coverage Improvement is its unique approach to leveraging AI coding assistants without adding additional overhead or approval processes.
- Automated Improvement Plan: Diffblue Cover automatically generates a detailed improvement plan for your codebase. This isn’t just a list of uncovered lines; it’s an intelligent, prioritized roadmap identifying specific areas where additional test cases would yield the greatest coverage benefit.
- Auto-Generated Prompts: For each step in this plan, Cover doesn’t just tell you what to do; it gives you how. It generates a full, ready-to-use prompt specifically designed for your favorite AI coding assistant (like GitHub Copilot Chat, for example).
- User-Driven Execution: This is how it works:
- Cover generates the precise prompt.
- The user simply pastes this prompt into their AI chat/assistant.
- The AI assistant then takes over, leveraging its understanding of the prompt and the surrounding code to generate the necessary test code or modifications. The user remains in control, guiding the AI and integrating its suggestions.
Crucially, there is no direct LLM interaction from Diffblue Cover itself, and no AI approvals are required within your enterprise environment. This design ensures that companies can deploy this powerful feature without navigating complex new AI governance or security protocols. It augments the abilities of your existing AI coding assistant, making it even more effective for test generation.
An Example of the Workflow
Let’s walk through a typical workflow to see how Guided Coverage Improvement tackles a difficult coverage issue.
- Run dcover create: You start by running the standard Diffblue Cover command. It generates tests for your codebase, but also outputs a list of “output codes” that explain why certain lines or branches couldn’t be covered automatically.
- Run dcover issues: This command takes the list of output codes, aggregates and prioritises them into a concise list of “issues.” For example, a common issue might be a high-priority R013 code, indicating that a certain object instance is too complex to create automatically and requires a dedicated factory method.
- Resolve the Issue: Diffblue Cover provides an auto-generated prompt for this specific R013 issue. You simply copy this prompt and paste it into your AI coding assistant. The assistant, guided by the prompt, helps you write the required factory method to simplify the object creation.
- Run dcover create again: Now that the necessary factory method is in place, you run dcover create once more. This time, Diffblue Cover can leverage the new factory method to successfully create the required object instance, leading to new test cases being generated and the R013 issue being resolved.
Watch the Demo
Designed for Enterprise Reality
We understand that enterprise environments come with constraints. That’s why Guided Coverage Improvement is built with pragmatic deployment in mind:
No Additional AI Approvals Required
The initial release focuses on generating prompts that you control. There’s no direct LLM invocation by Diffblue Cover itself, meaning you can deploy this feature immediately without navigating complex AI governance processes. You maintain full control over which AI tools you use and how.
Complements Your Existing Workflow
Whether you’re using GitHub Copilot, Claude Code or any other AI coding assistant, Guided Coverage Improvement provides optimized prompts that work with your chosen tools. It’s not about replacing your workflow—it’s about making it dramatically more effective.
Progressive Enhancement Path
Start with manual prompt execution today. As your organization’s AI policies evolve, the feature is designed to scale toward more automated workflows, including direct integration with approved LLMs for fully autonomous coverage improvement.
The Technical Edge: Why This Works
Collaborative Foundation Guided Coverage Improvement achieves superior results because it builds on the Test Asset Insights capability:
- Tests live in unified classes with clear ownership annotations (@DiffblueTest, etc.)
- Full visibility into existing @Before/@After methods and test fixtures
- Access to user-defined factory methods and test data builders
- Understanding of existing mock configurations and Spring contexts
Sophisticated Prioritization Engine Our analysis shows that not all coverage gaps are created equal. Some issues block entire test suites, while others affect single methods. Guided Coverage Improvement uses advanced heuristics to identify high-impact issues like:
- Environment configuration problems (Jacoco issues)
- Missing constructors preventing object initialization (R013)
- Static initializer failures (R006)
- Spring context configuration gaps (R026/R027)
Context-Rich Prompt Generation The prompts are not generic instruction, they includes:
- Specific details about your codebase structure
- References to your existing test patterns
- Contextual information about related tests
- Targeted suggestions that respect your testing conventions
Part of a Complete Testing Transformation
Guided Coverage Improvement builds upon Diffblue Cover’s proven foundation and works synergistically with our next-generation capabilities:
- Reinforcement Learning-Powered Autonomous Generation is the core engine of our signature agentic approach, which has always set Diffblue apart. With a single command, it autonomously generates deterministic unit tests at scale, guaranteed to compile and pass, running unattended for hours across entire codebases without human intervention.
- Test Asset Insights enhances this foundation by analyzing and learning from your existing test patterns, enabling the agent to generate tests that follow your established conventions and reuse your test infrastructure.
- LLM-Augmented Intelligence fuses our reinforcement learning engine with your approved language models, combining the reliability of autonomous generation with the contextual understanding of LLMs for complex business logic.
- Guided Coverage Improvement orchestrates everything above, providing the intelligence layer that identifies, prioritizes, and guides resolution of coverage gaps—transforming our autonomous engine from a powerful tool into a complete solution.
Start Your Coverage Transformation Today
Don’t let coverage gaps compromise your code quality any longer. With Guided Coverage Improvement, you have a clear, actionable path from wherever you are to the coverage levels your project demands.
Ready to break through your coverage ceiling?
Download the latest Diffblue Cover | View Documentation | Schedule a Demo
