5, 10 or 20 seats+ for your team - learn more
As its public face, a chatbot needs to project the values and character of the organization it represents. In this series of liveProjects, you’ll help out a local pet refuge by building an AI-powered chatbot that shares both the knowledge and the attitudes of that compassionate institution. Starting with prompt engineering and persona design, you’ll progressively layer in curated search, multi-use case handling, and callback scheduling. You’ll take advantage of OpenAI’s API along with LangChain and LangGraph to create a custom assistant that can deliver 24/7 pet care guidance. By the end, you’ll have a sophisticated AI agent that blends empathy, domain-specific knowledge retrieval, and intelligent routing.
In the first liveProject, you’ll create a basic chatbot that can reliably field questions in a way that’s consistent with the mission of the pet refuge. You’ll craft structured, effective prompts and design a persona for your chatbot, implement the basic functionality, and then test, refine, and improve your chatbot with iterative evaluation.
In the third liveProject, you’ll upgrade the capabilities of the RAG-enhanced chatbot, turning it into a multi-use-case agent using advanced reasoning, custom tools, and the ReAct framework. First, you’ll build a robust prompt chain with reasoning to classify queries and route them to the right path. Next, you’ll add practical tools and workflows with LangGraph to handle out-of-scope cases, capture user info, and blend retrieval with classification. Finally, unify everything into an agentic system that can choose tools, search knowledge, collect details, escalate when needed, and keep conversations natural.
This liveProject series is for Python programmers ready to develop basic chatbots into sophisticated AI agents with accurate personas. You should be comfortable with Python scripting and eager to learn advanced prompt engineering, custom tool development, and agent frameworks.
From Prompts to an Agentic System project for free