{"type":"video","version":"1.0","html":"<iframe src=\"https://www.loom.com/embed/9617730818074e918b08414ab7322e70\" frameborder=\"0\" width=\"1920\" height=\"1440\" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>","height":1440,"width":1920,"provider_name":"Loom","provider_url":"https://www.loom.com","thumbnail_height":1440,"thumbnail_width":1920,"thumbnail_url":"https://cdn.loom.com/sessions/thumbnails/9617730818074e918b08414ab7322e70-176a900013ba5c4d.gif","duration":292.458333,"title":"Hydroscope Beats Vector Search with Graphs","description":"This Loom explains how Hydroscope improves coding agents by using a self hosted HydraDB dependency graph instead of vector search. The author describes cases where vector search misses downstream dependencies and returns outdated rules, while HydraDB traverses supersedes edges to block retired policies and uses graphontology to prevent false matches, stopping pre flight at zero tokens when a subsystem like Kafka does not exist. Hydroscope unifies skills, 64 MCP tool schemas, AST code symbols, and resolved entity aliases in a single graph model, achieving 100% code based task pass rate versus 12.5% for vector search, raising tool accuracy from 85.7% to 100%, reducing agent turns from 5 to 1, and shrinking prompt tokens by 94% while writing resolved entity IDs back after each run."}