Microsoft Adds Knowledge Graphs to Foundry IQ to Improve Multi-Hop Retrieval
Microsoft has added Knowledge Graphs and a Microsoft-Retriever-1 component to its Foundry IQ product, aiming to reduce reliance on large frontier models for planning multi-hop retrieval over flat vector chunks. Starburst separately launched Context Studio, which exposes governed semantic context to agents over MCP to limit conflicting business definitions. Both moves point to a growing view that vector search alone is insufficient and that prompting frontier models with everything is slow and costly. The author, who is building an open-source self-hosted alternative called PipesHub, reports practical lessons such as combining deterministic structural links with selective entity linking rather than relying solely on costly LLM-based triplet extraction. The author also describes implementing separate identity handling for service-account tokens and encountering OAuth Dynamic Client Registration failures with Dropbox's MCP server, arguing that pre-registered clients should be supported alongside existing methods.
This is an AI-generated summary. ShortSingh links to the original source for the complete article.
Discussion (0)
Log in to join the discussion and vote.
Log in