Embed analytics experiences in your applications using the Conversational Analytics API

Key Takeaways

Ever wish you could just ask your data questions and get accurate answers right away? No SQL, no dashboard clicks. That's exactly what the conversational analytics API is built for. It's part of the Gemini stack and it lets you [music] embed conversational intelligence right into your applications powered by your own data. So what is the conversational analytics API? The API is a developer interface to the same engine that powers conversational analytics inside Looker. But instead of using it on a web page, you can call the API from anywhere. You create a data agent, an AI agent that understands your schema, business logic, and data connections. Then through [music] the API, you set natural language queries, and the agent translates the queries into SQL or look [music] and returns the structured results. But it's not just a static model that you can't control. The API supports system instructions and authored context so you can define your own instructions and examples to keep your model grounded. For example, you can tell the agent to always filter on specific fields, provide sample natural language questions and their equivalent queries, define a glossery of key terms, and [music] explain how to join tables using schema relations. The API also supports various options for conversation management with stateful sessions and stateless chats. These options let you decide whether to use manage state to an existing agent or conversation object or build an entirely stateless application with dynamically generated inline context. The API also includes AM access controls builtin with predefined rules to grant permissions to create, share, [music] and chat with data agents and to use stateless chats. And since it's just an API, you can combine it with the Mini Live API or other LLM tools for multi-turn multi-source reasoning. For example, you can use the Looker MCP server or agent development kit known as ADK to connect to other agents. In fact, if you've used the ADK's ask data insights tool, you've already seen the conversational analytics API in action. This is the API to use when you want to go beyond static dashboards. You can stop guessing which filters to click and instead ask questions dynamically with the agent inferring how to structure the query. This provides analytics that feels like a conversation but still runs on your covered secure data. Let's see this now in action. You can clone the quickserts repo which has a readyto run examples in Python. Here's a simple Python call to send our chat message and handle the response. The agent interprets the question, runs the right BigQuery query, and gives you the response with both natural language and structured results, so your app can handle the response however [music] you want. So, how do you get started? Check out the links below and try the quick start app and explore the demos and tools repos to see code examples. Then you can start building your data agent using an SDK in a language of your choice. The conversational analytics API turns your data into something more than numbers. It turns it into a dialogue. Thank you for watching and happy building. [music] >> [music]