What if you could talk to your app’s architecture the way you talk to your favorite LLM? “Show me the top 5 classes by CPU usage in domain A,” “Find all classes from package B that sneak into other domains.” That’s exactly what we’ve built: a query engine that lets you ask your monolith questions—no custom scripts, no guesswork.
vFunction’s new GenAI-powered query engine lets architects and developers run natural language prompts against the structure of their monolithic application. Just ask a question, and we’ll handle the rest: translating it to safe, validated internal queries, running it against our database, and returning results in a readable table. All you need to do is type.
Why build a query engine?
Monoliths are famously opaque, and do you really want to spend the precious hours in your day trying to decode them? Understanding how the system behaves, what calls what, where coupling occurs, and how methods evolve is often buried under layers of code.
Customers asked us, “Can we export what we see in the call tree?” They wanted to include it in architecture reviews, technical documentation and diagrams. Screenshots weren’t cutting it. That’s when we realized the architectural graph powering vFunction should be queryable with natural language. That got us thinking—what else could we do?
Here are some examples of queries users can run that would previously require exporting and manually filtering a full call graph:
- Show me all classes used in more than four domains that aren’t common.
→ Reveals architectural coupling or candidates for shared libraries. - Find all methods in the call tree under the domain ProductController that use beans.
→ Useful for mapping data access patterns, often buried in complex trees. - Which domain shares the most static classes with the domain InventoryService?
→ Helps determine which domains could be merged with the current domain.
How does the query engine work?
The query engine is not just a search box. It’s a full-blown architectural Q&A powered by GenAI, tied into your application’s live architectural state.
Here’s how it works:
- You write a prompt like “Show me the classes using the SalesOrderRepository across domains.”
- Our GenAI turns the prompt into a query.
- We send only the query to the GenAI provider—no data, no context, just the natural language prompt.
- The GenAI returns a query.
- We run the query locally against your vFunction server’s architecture data and display the results in a table or CSV format.
Security first
LLMs can hallucinate. We don’t let them.
vFunction never sends your application data to the GenAI provider. Only the user’s natural language prompt is shared. Nothing else. The GenAI is used strictly to translate the prompt into a query tailored for vFunction’s internal schema. At no point is your measurement data exposed outside your environment.
After generating the query, vFunction validates and sanitizes it, then runs it locally on your server. You get the benefits of natural language interfaces with complete data privacy and protection.
The result: conversational architecture analysis
With the new GenAI-powered query engine, you don’t need to dig through call trees or guess how classes relate. Just ask.
Want to explore stack traces, track class reuse across domains or filter down a call path for documentation? Open vFunction’s query engine, describe what you’re looking for, and get the answer. Even the most complex monolith is now an open book—saving you hours of effort digging through code, tracing dependencies, and assembling documentation.
Curious how vFunction helps teams tackle technical debt and turn monoliths into modular, cloud-ready apps? Explore the platform and see what architectural modernization looks like in action.
