OpenAI Python API Bootcamp (2023): Learn AI, GPT, and more!

Translate natural language queries into SQL with OpenAI, enabling non-technical users to retrieve tabular data from an in-memory database.

Explore text completion API parameters, compare text and code models, and learn how temperature, max tokens, top-p, and penalties shape SQL query generation.

Learn to implement an OpenAI API call, craft prompts from a DataFrame with prompt engineering, and generate SQL queries using code-davinci-002 with careful response handling.

Create an automatic exam creator that generates topic-based, multiple-choice quizzes with answer sheets, an interactive testing experience, and grading via the OpenAI text completion API and prompt engineering in Python.

Explore prompt design and prompt engineering for OpenAI API completion models, addressing prompt leakage and safety, plus jailbreak risks. Learn principles: model choice, instructions, details, examples, and direct, code-oriented prompts.

Create a random exam by crafting a prompt that specifies number of questions and four options, then extract a separate answer key while addressing model hallucination with certainty checks.

Refactor the quiz tool into python scripts structured as classes, with main.py, teacher.py, an exam simulator, and a test creator to create, take, and grade full tests for production purposes.

Explore building an automatic recipe creator that turns a list of ingredients into a full recipe and a generated dish image using the completion API and DALL-E 2.

Explains how to use the dalle-2 image API to generate images from prompts, request multiple images, vary size up to 1024 by 1024, and save results with Python.

Follow a Python script walkthrough that generates a recipe from user inputs, creates a detailed prompt, renders a matching image with DALL·E, and previews future GitHub Pages blog automation.

Learn to build an automatic blog post creator by combining OpenAI API outputs with GitHub Pages, generating text and a DALL-E image, and publishing via Python functions.

Set up a GitHub Pages repository to host a static HTML and CSS site on your username.github.io, then clone, create index.html, push changes, and view the live page.

Automate updating a GitHub Pages repository with Python and GitPython, performing add, commit, and push to index.html, while preparing for AI-generated blog content in a future lesson.

Automate blog post creation in Python by generating HTML templates with OpenAI text and DALL-E images, organizing posts in a content directory, and updating the index with links.

Execute OpenAI API calls to generate blog text and images through well-crafted prompts with context and tags. Save assets and update a GitHub blog workflow to publish the post.

Walk through a python script to automate blog posts: set up paths, insert titles, generate content with open ai utilities and images, and update the blog index using blog utilities.

Explore Reddit sentiment analysis with the OpenAI completion API, prompting for company extraction and sentiment rating (positive, negative, neutral) from investing subreddit posts.

Create an official Reddit app to obtain a client ID and secret key, then connect with Python using PRAW and a user agent to access subreddit data.

Learn to fetch top posts and comments from stocks subreddit using a Python reddit API wrapper, then craft prompts for the completion API with prompt engineering in a Jupyter notebook.

Generate prompts from Reddit top post titles and comments, skip pinned posts, and format them into a dictionary for GPT. Use OpenAI to classify sentiment and extract stock tickers.

Learn to use Codex models to generate Python function docstrings and automate documentation with prompt engineering, including a script that adds docstrings to every function in a .py file.

Design prompts for the OpenAI API by converting a Python function to a string with inspect, then generate a docstring with a code-davinci model, managing quotes and escapes.

Merge a Python function with its docstring by splitting the function source, inserting the docstring with proper quotes, and reassembling the complete function for further testing.

Learn to automatically generate docstrings for Python functions with a main.py script. The walkthrough shows inspecting a module, generating docstrings via OpenAI, and writing a new file with formatted functions.

Explore a translation project with OpenAI, translating top headlines to English and summarizing them. Compare OpenAI's translation capabilities with cloud provider models for higher quality results.

Learn how to scrape international newspaper sites by country to retrieve top headlines using BeautifulSoup and requests, then translate and summarize with OpenAI.

Build an OpenAI prompt to translate and summarize country headlines, scrape Spain and France news via Beautiful Soup in a Python Jupyter workflow, and refine results with the OpenAI API.

Fine-tune GPT models with proprietary data to expand context, reduce token limits, and lower latency, while learning data formatting, costs, and practical uses like customer support.

Create a prompt and completion JSON for fine-tuning OpenAI, using the trimmed Kaggle 10% Stack Overflow dataset converted from python_qa.csv to qa_openai_format.

Learn how to fine-tune a non‑Davinci model, estimate training costs using tiktoken token counts, choose encodings like gpt2 for Babbage, and balance dataset size with four epochs.

Learn to fine-tune OpenAI models via the command line, preparing training data JSON and running fine_tunes.create with a Babbage model. Set up the API key and monitor the upload.

Explore text embedding with the OpenAI API to convert text to vectors, enable cosine similarity, clustering, and fuzzy search, and build context-aware completion by inserting vectors into prompts.

Learn how model hallucination arises when large language models sound confident yet give false facts. Explore prompt engineering, certainty prompts, and strategies to reduce hallucinations in practice.

Learn to convert document data into context for prompts by building a matching index of vector embeddings and a single summary string with pandas and literal_eval.

Estimate embedding costs by counting tokens with tiktoken, use CL100K base, and generate OpenAI embeddings (ada-002) to build 1,536-dimensional vectors for document similarity and QA.

OpenAI vector embeddings and cosine similarity help you match queries with the most relevant summaries. Inject context from embeddings into prompts to enable accurate, cost-effective answers without fine-tuning.