Description

While reviewing the Knowledge Graph extraction logic, I noticed that edges could be duplicated during sequential processing if the same relationship is mined multiple times.

This causes redundant data in the SQLite store and can impact traversal efficiency.

Proposed Fix

I would like to propose a tiny, focused PR that deduplicates edges directly before the graph traversal/insertion phase in topology.py or knowledge_graph.py.

Let me know if you agree with this approach and I will open a PR with just this fix.