Most AI applications are built by calling a model through an API rather than training one. This module covers working with LLM APIs from OpenAI and Google: getting an API key, sending messages, choosing a model, streaming responses, and building chat interfaces on top.
It also looks at getting usable results out of model responses so they can feed the rest of your software. The LangChain projects connect models to external data and tools, and include a multi-language invoice extractor that pulls key fields out of documents.
These courses are long, so treat them as a reference: watch the parts you need, then come back for the rest. When you're done, continue on to RAG & Vector Databases.
A first look at building with LLMs: the ChatGPT playground, OpenAI API basics, a ChatGPT-style interface with streaming and a stop button using Chainlit, then vector databases, document Q&A, and agent tools.
Gemini from the ground up: getting an API key, choosing models, initializing the generative model, text-to-text, image-to-text, chat and embeddings, ending with an AI code buddy chatbot.
Six projects with LangChain and GPT-4, Gemini Pro and Llama 2: chat with a PDF, blog generation, end-to-end apps with a Pinecone vector database, a multi-language invoice extractor, and a Gemini Q&A chatbot.