Lurkee
A monitoring system that turns website changes into prioritized business intelligence through scheduled collection, deterministic diffing, structured LLM analysis, and reports.
The system behind the interface.
A changed page is not automatically useful intelligence. The product needs to separate noise from meaningful signals without hiding the source evidence from the operator.
Architecture choices that shaped the product.
Hashing and diffs establish what changed before an LLM is asked to explain why the change may matter.
Severity and findings use typed schemas so the result can drive alerts, reports, and filters instead of becoming an unbounded paragraph.
Routine analysis can use a lower-cost model while stronger reasoning is reserved for findings that cross a severity threshold.
What exists beyond the concept.
- Scheduled collection with content hashing and unified change detection
- Structured severity classification before reporting and alerts
- Operator-facing history, report, billing, and system-health workflows
A production-minded result.
A working end-to-end monitoring pipeline and dashboard foundation. The current implementation is described as multi-stage, not as a LangGraph system.