1、Democratizing Data AccessEmpower business users withself-service analytics on trusted,governed data in Unity CatalogAs a busy analytics team,we dont have the bandwidth to run ad-hoc data related requests sent to us from members of the business.Is there a way for them to self-serve in a secure way?Th
2、e VisionMulti-agent system for conversational data access.Bridges the gap between users and governed data in Unity Catalog.Users ask questions in natural language.System provides text answers,SQL,and charts.System can be securely shared with users in the OrgConversational AI on DatabrickssupervisorG
3、eniechart generatorDBX AppUCGeniespaceSystem Architecture&Core ComponentsDatabricks AppSystem orchestrated by LangGraph.Agents:The brain manages the workflow.Genie Agent:Translates natural language to SQL via Databricks Genie API.Chart Generator Agent:Creates data visualizations(static charts).Meet
4、the AgentsRole:Central coordinator and router.Determines user intent(data query,chart request,etc.).Routes tasks to Genie or Chart Generator.Manages the state of the conversation and data flow.Supervisor AgentMeet the AgentsRole:Natural language to SQL translator.Implemented as a callable tool for t
5、he Supervisor.Interacts with Databricks Genie Conversation API.Targets a specific,pre-configured Genie Space.Polls the API to get generated SQL and structured data.Key Dependency:Success relies heavily on a well-curated Genie Space(metadata,instructions,examples).Genie AgentMeet the AgentsRole:Creat
6、es static charts from data.Invoked by the Supervisor.Takes structured data(from Genie)and chart instructions.Uses an LLM to generate a Streamlit string for visualizationReturns the final tool output as the final answer.Chart Generator AgentPlatform Integration&ControlLangGraph:Used for building stat