Video: Full walk-through of building a website support AI Agent in Cerb 11.1

[jstanden]

Watch on YouTube: Build an AI agent with tool use and custom knowledge in Cerb 11.1

This is a deep dive of the new AI agent functionality in Cerb 11.1. We create a support agent from scratch using the new llm.agent: command. This gives us interchangeable LLM providers, memory, transcripts, and the ability to add automations as tools in just a few lines of code. The agent can be added to any website with a single script tag.

We build a quick tool for answering visitor questions based on website content and the latest documentation. This example uses our semantic search API, but you can also use search services like Tavily, Linkup, or Exa, to integrate your existing docs.

We review an efficient, agentic search strategy (tools vs preemptive RAG), and gradually add functionality so you can learn how everything works.

You can follow along with a free, full-featured copy of Cerb in Docker: https://cerb.ai/docs/installation/docker/

Use version 11.1 tag from Docker Hub since these features are currently in beta.

We also use Anthropic Claude as the LLM provider, so create an account and API key if you don't have one already: http://console.anthropic.com/