Pinecone

A short walkthrough demonstrating the Pinecone Nexus integration with Microsoft OneLake. In this demo, Pinecone Staff Data Engineer Simon Lu showcases how easy it is to use Nexus to build a reliable, production-grade knowledge layer directly over structured data stored in Microsoft OneLake or a Fabric Data Warehouse. By creating an intuitive semantic layer, Nexus allows AI agents to accurately interpret complex enterprise schemas and answer business questions using natural language. What You'll Learn: The Setup (0:00): A look at the sample enterprise datasets in the Fabric console (Census, NYC Taxi, Holidays, and NYC 311 data). Connecting to Fabric (1:05): How to configure the connection using an Azure Service Principal and your SQL connection string. Adding Context Documents (1:55): How to upload markdown documentation, data dictionaries, or semantic models to bootstrap your knowledge base. Background Initialization (2:50): How the background agent explores tables, maps entity relationships, and automatically writes analytical data reports. Nexus in Action (3:42): A demo of the agentic chat playground handling complex, multi-table SQL queries from natural language prompts. Programmatic Access & Curation (4:54): How to integrate Nexus into your own agent workflows as a tool, and how to dynamically update the knowledge layer with feedback. Under the Hood (5:51): An inside look at the source page and the atomic knowledge artifacts Nexus generates to ensure consistent, highly accurate RAG performance. Learn more about the Pinecone Nexus integration with Microsoft Fabric: https://www.pinecone.io/newsroom/microsoft-onelake-nexus/

A quick walkthrough demonstrating the Pinecone Nexus integration with Microsoft OneLake. Video Timeline: 0:00 - 0:08 | Setup & Connection: Connecting Nexus to Microsoft Fabric via Azure Service Principal and SQL endpoints using a sample enterprise dataset (Census, NYC Taxi, and 311 data). 0:08 - 0:24 | Bootstrapping the Knowledge Layer: Ingesting existing onboarding docs and launching an background agentic exploration loop that probes schemas, maps entity relationships, and analyzes data distributions. 0:26 - 1:04 | Agent Playground in Action: Watch a thin agentic layer leverage Nexus to accurately write SQL, cite its sources, and answer complex natural language questions over structured data. 1:06 - End | Knowledge Artifact Curation: A look inside the Nexus backend to see how raw data is pre-compiled into structured, atomic knowledge artifacts, domain documentations, and data quality caveats for deterministic retrieval. Learn more about the Pinecone Nexus integration with Microsoft Fabric: https://www.pinecone.io/newsroom/microsoft-onelake-nexus