Master OpenAI Agent Builder - Deploy Chatbot to Your Website

Explore the OpenAI workflow canvas to design agents with templates and 12 nodes, then publish, rename, and evaluate workflows using the OpenAI agent SDK, hosted on OpenAI servers.

Learn to evaluate and grade agent output using rules for politeness, conciseness, helpfulness, accuracy, and creativity, and review evaluation reports across traces.

Learn how to configure an AI agent, rename it, and set a system instruction for high school students, while managing chat history and message continuity.

Manage agent chat history with system, user, and assistant messages, including pre-written history. Tune model parameters and reasoning effort for testing vs production, and access the agent code and SDKs.

Create a classifier agent to route weather queries using a logic node and a JSON two-category schema (weather or other) to the weather agent.

Demonstrate using set state to initialize and update a count, and apply a transform to modify the state value before sending it to the agent.

Explore how to use a while loop with set state and transform node to iterate tasks, update a count number, and log progress in an agent workflow.

Explore how to implement guardrails in an OpenAI agent builder workflow, including PII data protection, moderation, and jailbreak detection, using guardrail and user approval nodes with GPT-5 mini.

Learn to use a user approval node to gate complex queries, classify questions as simple or complex, and route to simple or high-performance models with JSON output.

Design health and finance rag using OpenAI agent builder, with two vector stores, a file search tool, classifiers, and web search for general questions, producing answers from relevant data.

Design a classifier-driven rag architecture that routes health, finance, and general queries to dedicated agents, using a json output and if-else routing.

Demonstrate health and finance rag by validating classifier and general agents and using a web search tool for time-sensitive queries like latest global news and earnings.

Design a simple Q&A chatbot with guardrails for moderation and jailbreak, configure a file-driven agent, craft a strict system prompt that searches files and delivers polite final answers with references.

Upload your website data into a named vector store to power a Q&A agent, then tailor the system prompt with emojis, bullet points, and markdown formatting for a website chatbot.

Set up your qna chatbot locally by installing node.js, configuring a dot env with your OpenAI API key and workflow ID, and running the OpenAI chat kit starter in VSCode.

Freeze a working OpenAI chat starter app version in a GitHub repository and push changes. Deploy via Vercel after configuring the OpenAI API key and workflow ID as environment variables.

Deploy your Next.js chatbot to production, whitelist the domain in OpenAI settings, and test live interactions through the deployed URL and iframe integration for your website.

Integrate a QnA chatbot with your website by embedding an iframe and adding a header script. Configure the chat bubble and API URL via OpenAI Agent Builder.

Refine the prompt template to generate clickable references for the website qna chatbot by reconstructing full URLs from file references and publishing changes to production for visible, clickable sources.

Set up an AWS RDS MySQL database (Aurora) for the Creepy Vibes Cafe, configure an admin user with an auto-generated password, and capture endpoint, port, and MySQL Workbench for debugging.

Download and install MySQL Workbench locally to connect to your AWS RDS MySQL instance. Create a new connection using the RDS host, port 3306, and credentials, then test the connection.

Connect mysql workbench to the rds database, create the creepy vibes database with customers, products, orders, and order items, and upload sample data for testing the website chatbot.

Design the OpenAI agent workflow by connecting a start node, guardrail, and intent classifier to four intent agents—order status, product search, recommender, and order placement—with a general fallback.

Classify customer queries into five intents: order status tracking, product search, discovery, product recommendations, and order replacement, then route them via an intent classifier in a JSON workflow.

Create an AWS Lambda function to access an RDS database, expose it via API gateway, and connect to Zapier with a webhook tool and a MySQL dependency layer.

Configure lambda environment variables and deploy to production, connect the lambda to an rds database, test queries, and set up an api gateway for chatbot deployment.

Learn to set up an AWS HTTP API gateway, connect it to a Lambda function to access an RDS database, configure routes and stages, and test the integration with curl.

Connect a MCP server to Zapier using a GET webhook, paste the API key, and configure tools for automated, user-unapproved calls to access the database.

Rename the workflow and test the AI assistant in preview mode to validate the product search and behind the scenes data access and approval flows.

Publish and update your OpenAI powered chatbot, configure production environment, swap API keys and workflow IDs, and redeploy on Vercel to launch your order status and product recommender AI assistant.