Finds non-obvious connections between episodes in the knowledge graph. Use when linking a new episode to the Memory MCP knowledge graph.
Find connections that aren't obvious from keywords alone. The goal is to surface surprising links — two episodes that discuss the same underlying principle from completely different angles.
Memory:search_nodes to find concepts related to the new episodeMemory:open_nodespython scripts/similarity_scorer.py [episode_id] for TF-IDF similarityMemory:create_relations to link nodes with descriptive edge labelsUse descriptive relation types:
supports — evidence reinforces an existing claimcontradicts — directly conflicts with an existing claimextends — builds on an existing conceptapplies_in — same concept used in a different domainreferences — mentions the same person/book/ideaIn episode cards, phrase connections as insights: