AI For JavaScript Developers (OpenAI, Embeddings, Vercel AI)

Sign up for OpenAI, obtain an API key, and claim $5 in free credits for three months while using the GPT-3.5-turbo model.

Assume your prior JavaScript experience and install Node.js using nvm or download the LTS from nodejs.org, then verify the installation by running node -v.

Learn how the project is structured with per-lecture git repos and an intro folder, scripts run by tsx and npm, and setup for OpenAI and environment variables.

Learn to set up the OpenAI API with the official SDK, configure a dot env with your API key, and build a chat completion client using GPT-3.5 turbo with messages.

Respond to feedback prompts by rating this course and sharing your thoughts. Your input drives continuous improvement, as the instructor cites a 4.7 rating from over 200,000 students.

Learn to manage conversation state with a history array in OpenAI chat completions, using system, user, and assistant roles, embeddings, vector stores, and function calls.

Explore how tokens drive tokenization, context window, and pricing in OpenAI models, and learn to manage input and output tokens effectively.

Discover how system prompts and system messages steer large language models by injecting constraints before user input, with examples like a best shoes bot.

Provide real-time stock data to a chat bot via a system prompt. Use a JavaScript object and json stringify to feed Nike stock in New York into the LLM.

Explore how hallucinations occur in chatbots by misreporting stock data across locations, compare GPT-3.5 and GPT-4 capabilities, and discuss prompt engineering strategies to improve accuracy.

Convert user goals into a structured json of tasks using prompts and a strict schema, producing a front-end friendly array of task objects with title, description, and difficulty.

Learn OpenAI function calling to produce structured Json and trigger real functions, using Json schema, calculator examples, and practical integrations with email and CRM tools.

Learn to manage huge data with embeddings and a vector store to power a bot that references internal pdfs via chunking.

Learn to ingest a PDF with llama index, chunk into 1000-token overlaps, and build a vector store of embeddings. Retrieve top matching chunks and query OpenAI with a system prompt.

Explore real-time data querying in a chat-based product recommender using vector stores, embeddings, and OpenAI function calling to return clickable product URLs.

Build real-time product recommendations using an OpenAI function, fetch a product id, and render a user interface link. Extend with more functions and a vector store.

Build a Next.js and React front end to upload a PDF and turn it into embeddings stored in a local llama index vector store for chat using the Vercel SDK.

Clone the repository, checkout the course start here branch, install dependencies, configure your OpenAI API key, and run the Next.js app to explore the back end.

Explore the structure of a Next.js 13 app router project with tailwind styling, loading components, drag-and-drop uploads, and a chat and upload pages workflow using Vercel AI.

Learn to upload a PDF file, parse its contents with a PDF parser, and store the resulting chunks in a llama index vector store via a Next.js API route.

Demonstrates creating or reading a vector store index from a PDF by chunking text with overlap, persisting to disk, and using a storage and service context for Next.js routing.

Demonstrates how the Vercel AI SDK uses the useChat hook to manage messages and input, wire a /chat API endpoint, and render OpenAI-style chat messages.

Create a chat API endpoint in a Next.js app by exporting an async post function that processes messages and streams responses to the client using OpenAI and Vercel SDK.

Learn to enhance a chat AI with document knowledge by querying a vector store with embeddings, retrieving a relevant node, and conditionally updating the system message for enriched responses.

Learn to manage chat context limits by counting tokens and measuring token length with a tokens utility, a token library, and GPT 3.5 turbo encoding.

Learn how to enforce a token limit in a chat by preserving the system message, trimming the latest non-system messages, and capping the response at a safe 4097-token limit.